This is an archive of the discontinued LLVM Phabricator instance.

Differential D101313

[AMDGPU] Drop the GCNRegBankReassign pass
ClosedPublic

Authored by bsaleil on Apr 26 2021, 11:14 AM.

Details

Reviewers
piotr
foad
rampitec
Commits
rGcaf1294d9578: [AMDGPU] Experiments show that the GCNRegBankReassign pass significantly impacts
Summary

Experiments show that the GCNRegBankReassign pass significantly impacts the compilation time and there is no case for which we see any improvement in performance.
This patch removes this pass and its associated test cases from the tree.

Diff Detail

Unit TestsFailed

TimeTest
200 msx64 windows > lld.MachO::reproduce.s

Event Timeline

bsaleil created this revision.Apr 26 2021, 11:14 AM
Herald added subscribers: wenlei, kerbowa, asbirlea and 6 others. · View Herald TranscriptApr 26 2021, 11:14 AM
bsaleil requested review of this revision.Apr 26 2021, 11:14 AM
Herald added a subscriber: llvm-commits. · View Herald TranscriptApr 26 2021, 11:14 AM
Harbormaster completed remote builds in B100984: Diff 340587.Apr 26 2021, 11:15 AM
bsaleil added reviewers: piotr, foad, rampitec.Apr 26 2021, 11:17 AM
rampitec accepted this revision.Apr 26 2021, 11:20 AM
This revision is now accepted and ready to land.Apr 26 2021, 11:20 AM
bsaleil updated this revision to Diff 340599.Apr 26 2021, 11:41 AM

Update patch so it applies upstream.

Harbormaster completed remote builds in B100993: Diff 340599.Apr 26 2021, 12:31 PM
bsaleil retitled this revision from Drop the GCNRegBankReassign pass to [AMDGPU] Drop the GCNRegBankReassign pass.Apr 26 2021, 2:01 PM
Herald added subscribers: t-tye, tpr, dstuttard and 3 others. · View Herald TranscriptApr 26 2021, 2:01 PM
This revision was landed with ongoing or failed builds.Apr 26 2021, 2:22 PM
Closed by commit rGcaf1294d9578: [AMDGPU] Experiments show that the GCNRegBankReassign pass significantly impacts (authored by bsaleil). · Explain Why
This revision was automatically updated to reflect the committed changes.
bsaleil added a commit: rGcaf1294d9578: [AMDGPU] Experiments show that the GCNRegBankReassign pass significantly impacts.

Revision Contents

PathSize
llvm/
lib/
Target/
AMDGPU/
13 lines
5 lines
1 line
test/
CodeGen/
AMDGPU/
GlobalISel/
8 lines
96 lines
4 lines
28 lines
208 lines
12 lines
156 lines
422 lines
8 lines
326 lines
318 lines
48 lines
428 lines
184 lines
8 lines
40 lines
4 lines
44 lines
86 lines
4 lines
4 lines
20 lines
8 lines
23 lines
31 lines
8 lines
130 lines
6 lines
178 lines
8 lines
178 lines
16 lines
12 lines
78 lines
60 lines
72 lines
8 lines
4 lines
8 lines
8 lines
4 lines
176 lines
68 lines
32 lines
4 lines
20 lines
46 lines
144 lines
120 lines
110 lines
68 lines
4 lines
8 lines
40 lines
40 lines
20 lines
16 lines
8 lines
12 lines
20 lines
14 lines
8 lines
14 lines
20 lines
8 lines
24 lines
27 lines
32 lines
24 lines
24 lines
50 lines

Diff 340599

llvm/lib/Target/AMDGPU/AMDGPU.h

Show First 20 Lines • Show All 68 Lines • ▼ Show 20 Lines
FunctionPass *createAMDGPULateCodeGenPreparePass(); FunctionPass *createAMDGPULateCodeGenPreparePass();
FunctionPass *createAMDGPUMachineCFGStructurizerPass(); FunctionPass *createAMDGPUMachineCFGStructurizerPass();
FunctionPass *createAMDGPUPropagateAttributesEarlyPass(const TargetMachine *); FunctionPass *createAMDGPUPropagateAttributesEarlyPass(const TargetMachine *);
ModulePass *createAMDGPUPropagateAttributesLatePass(const TargetMachine *); ModulePass *createAMDGPUPropagateAttributesLatePass(const TargetMachine *);
FunctionPass *createAMDGPURewriteOutArgumentsPass(); FunctionPass *createAMDGPURewriteOutArgumentsPass();
ModulePass *createAMDGPULowerModuleLDSPass(); ModulePass *createAMDGPULowerModuleLDSPass();
FunctionPass *createSIModeRegisterPass(); FunctionPass *createSIModeRegisterPass();
namespace AMDGPU {
enum RegBankReassignMode {
RM_VGPR = 1,
RM_SGPR = 2,
RM_BOTH = RM_VGPR | RM_SGPR
};
}
MachineFunctionPass *
createGCNRegBankReassignPass(AMDGPU::RegBankReassignMode Mode);
struct AMDGPUSimplifyLibCallsPass : PassInfoMixin<AMDGPUSimplifyLibCallsPass> { struct AMDGPUSimplifyLibCallsPass : PassInfoMixin<AMDGPUSimplifyLibCallsPass> {
AMDGPUSimplifyLibCallsPass(TargetMachine &TM) : TM(TM) {} AMDGPUSimplifyLibCallsPass(TargetMachine &TM) : TM(TM) {}
PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM); PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM);
private: private:
TargetMachine &TM; TargetMachine &TM;
}; };
▲ Show 20 Lines • Show All 242 Lines • ▼ Show 20 Lines
void initializeAMDGPUExternalAAWrapperPass(PassRegistry&); void initializeAMDGPUExternalAAWrapperPass(PassRegistry&);
void initializeAMDGPUArgumentUsageInfoPass(PassRegistry &); void initializeAMDGPUArgumentUsageInfoPass(PassRegistry &);
ModulePass *createAMDGPUOpenCLEnqueuedBlockLoweringPass(); ModulePass *createAMDGPUOpenCLEnqueuedBlockLoweringPass();
void initializeAMDGPUOpenCLEnqueuedBlockLoweringPass(PassRegistry &); void initializeAMDGPUOpenCLEnqueuedBlockLoweringPass(PassRegistry &);
extern char &AMDGPUOpenCLEnqueuedBlockLoweringID; extern char &AMDGPUOpenCLEnqueuedBlockLoweringID;
void initializeGCNRegBankReassignPass(PassRegistry &);
extern char &GCNRegBankReassignID;
void initializeGCNNSAReassignPass(PassRegistry &); void initializeGCNNSAReassignPass(PassRegistry &);
extern char &GCNNSAReassignID; extern char &GCNNSAReassignID;
namespace AMDGPU { namespace AMDGPU {
enum TargetIndex { enum TargetIndex {
TI_CONSTDATA_START, TI_CONSTDATA_START,
TI_SCRATCH_RSRC_DWORD0, TI_SCRATCH_RSRC_DWORD0,
TI_SCRATCH_RSRC_DWORD1, TI_SCRATCH_RSRC_DWORD1,
▲ Show 20 Lines • Show All 75 Lines • Show Last 20 Lines

llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

Show First 20 Lines • Show All 256 Lines • ▼ Show 20 Linesextern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeAMDGPUTarget() {
initializeSIFormMemoryClausesPass(*PR); initializeSIFormMemoryClausesPass(*PR);
initializeSIPostRABundlerPass(*PR); initializeSIPostRABundlerPass(*PR);
initializeAMDGPUUnifyDivergentExitNodesPass(*PR); initializeAMDGPUUnifyDivergentExitNodesPass(*PR);
initializeAMDGPUAAWrapperPassPass(*PR); initializeAMDGPUAAWrapperPassPass(*PR);
initializeAMDGPUExternalAAWrapperPass(*PR); initializeAMDGPUExternalAAWrapperPass(*PR);
initializeAMDGPUUseNativeCallsPass(*PR); initializeAMDGPUUseNativeCallsPass(*PR);
initializeAMDGPUSimplifyLibCallsPass(*PR); initializeAMDGPUSimplifyLibCallsPass(*PR);
initializeAMDGPUPrintfRuntimeBindingPass(*PR); initializeAMDGPUPrintfRuntimeBindingPass(*PR);
initializeGCNRegBankReassignPass(*PR);
initializeGCNNSAReassignPass(*PR); initializeGCNNSAReassignPass(*PR);
} }
static std::unique_ptr<TargetLoweringObjectFile> createTLOF(const Triple &TT) { static std::unique_ptr<TargetLoweringObjectFile> createTLOF(const Triple &TT) {
return std::make_unique<AMDGPUTargetObjectFile>(); return std::make_unique<AMDGPUTargetObjectFile>();
} }
static ScheduleDAGInstrs *createR600MachineScheduler(MachineSchedContext *C) { static ScheduleDAGInstrs *createR600MachineScheduler(MachineSchedContext *C) {
▲ Show 20 Lines • Show All 898 Lines • ▼ Show 20 Linesvoid GCNPassConfig::addOptimizedRegAlloc() {
if (EnableDCEInRA) if (EnableDCEInRA)
insertPass(&DetectDeadLanesID, &DeadMachineInstructionElimID); insertPass(&DetectDeadLanesID, &DeadMachineInstructionElimID);
TargetPassConfig::addOptimizedRegAlloc(); TargetPassConfig::addOptimizedRegAlloc();
} }
bool GCNPassConfig::addPreRewrite() { bool GCNPassConfig::addPreRewrite() {
if (EnableRegReassign) { if (EnableRegReassign)
addPass(&GCNNSAReassignID); addPass(&GCNNSAReassignID);
addPass(createGCNRegBankReassignPass(AMDGPU::RM_BOTH));
}
return true; return true;
} }
void GCNPassConfig::addPostRegAlloc() { void GCNPassConfig::addPostRegAlloc() {
addPass(&SIFixVGPRCopiesID); addPass(&SIFixVGPRCopiesID);
if (getOptLevel() > CodeGenOpt::None) if (getOptLevel() > CodeGenOpt::None)
addPass(&SIOptimizeExecMaskingID); addPass(&SIOptimizeExecMaskingID);
TargetPassConfig::addPostRegAlloc(); TargetPassConfig::addPostRegAlloc();
▲ Show 20 Lines • Show All 197 Lines • Show Last 20 Lines

llvm/lib/Target/AMDGPU/CMakeLists.txt

Show First 20 Lines • Show All 133 Lines • ▼ Show 20 Linesadd_llvm_target(AMDGPUCodeGen
SIPeepholeSDWA.cpp SIPeepholeSDWA.cpp
SIPostRABundler.cpp SIPostRABundler.cpp
SIPreEmitPeephole.cpp SIPreEmitPeephole.cpp
SIProgramInfo.cpp SIProgramInfo.cpp
SIRegisterInfo.cpp SIRegisterInfo.cpp
SIShrinkInstructions.cpp SIShrinkInstructions.cpp
SIWholeQuadMode.cpp SIWholeQuadMode.cpp
GCNILPSched.cpp GCNILPSched.cpp
GCNRegBankReassign.cpp
GCNNSAReassign.cpp GCNNSAReassign.cpp
GCNDPPCombine.cpp GCNDPPCombine.cpp
SIModeRegister.cpp SIModeRegister.cpp
LINK_COMPONENTS LINK_COMPONENTS
Analysis Analysis
AsmPrinter AsmPrinter
CodeGen CodeGen
Show All 26 Lines

llvm/lib/Target/AMDGPU/GCNRegBankReassign.cpp

  • This file was deleted.
//===-- GCNRegBankReassign.cpp - Reassign registers after regalloc --------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
/// \file
/// \brief Try to reassign registers on GFX10+ to reduce register bank
/// conflicts.
///
/// On GFX10 registers are organized in banks. VGPRs have 4 banks assigned in
/// a round-robin fashion: v0, v4, v8... belong to bank 0. v1, v5, v9... to
/// bank 1, etc. SGPRs have 8 banks and allocated in pairs, so that s0:s1,
/// s16:s17, s32:s33 are at bank 0. s2:s3, s18:s19, s34:s35 are at bank 1 etc.
///
/// The shader can read one dword from each of these banks once per cycle.
/// If an instruction has to read more register operands from the same bank
/// an additional cycle is needed. HW attempts to pre-load registers through
/// input operand gathering, but a stall cycle may occur if that fails. For
/// example V_FMA_F32 V111 = V0 + V4 * V8 will need 3 cycles to read operands,
/// potentially incuring 2 stall cycles.
///
/// The pass tries to reassign registers to reduce bank conflicts.
///
/// In this pass bank numbers 0-3 are VGPR banks and 4-11 are SGPR banks, so
/// that 4 has to be subtracted from an SGPR bank number to get the real value.
/// This also corresponds to bit numbers in bank masks used in the pass.
///
//===----------------------------------------------------------------------===//
#include "AMDGPU.h"
#include "GCNSubtarget.h"
#include "SIMachineFunctionInfo.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/CodeGen/LiveIntervals.h"
#include "llvm/CodeGen/LiveRegMatrix.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/CodeGen/MachineLoopInfo.h"
#include "llvm/InitializePasses.h"
using namespace llvm;
using namespace AMDGPU;
static cl::opt<unsigned> VerifyStallCycles("amdgpu-verify-regbanks-reassign",
cl::desc("Verify stall cycles in the regbanks reassign pass"),
cl::value_desc("0|1|2"),
cl::init(0), cl::Hidden);
// Threshold to keep compile time reasonable.
static cl::opt<unsigned> VRegThresh("amdgpu-regbanks-reassign-threshold",
cl::desc("Max number of vregs to run the regbanks reassign pass"),
cl::init(15000), cl::Hidden);
#define DEBUG_TYPE "amdgpu-regbanks-reassign"
#define NUM_VGPR_BANKS 4
#define NUM_SGPR_BANKS 8
#define NUM_BANKS (NUM_VGPR_BANKS + NUM_SGPR_BANKS)
#define SGPR_BANK_OFFSET NUM_VGPR_BANKS
#define VGPR_BANK_MASK 0xf
#define SGPR_BANK_MASK 0xff0
#define SGPR_BANK_SHIFTED_MASK (SGPR_BANK_MASK >> SGPR_BANK_OFFSET)
STATISTIC(NumStallsDetected,
"Number of operand read stalls detected");
STATISTIC(NumStallsRecovered,
"Number of operand read stalls recovered");
namespace {
class GCNRegBankReassign : public MachineFunctionPass {
class OperandMask {
public:
OperandMask(unsigned r, unsigned s, unsigned m)
: Reg(r), SubReg(s), Mask(m) {}
Register Reg;
unsigned SubReg;
unsigned Mask;
};
class Candidate {
public:
Candidate(MachineInstr *mi, Register reg, unsigned subreg,
unsigned freebanks)
: MI(mi), Reg(reg), SubReg(subreg), FreeBanks(freebanks) {}
#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void dump(const GCNRegBankReassign *P) const {
MI->dump();
dbgs() << P->printReg(Reg) << " to banks ";
dumpFreeBanks(FreeBanks);
dbgs() << '\n';
}
#endif
MachineInstr *MI;
Register Reg;
unsigned SubReg;
unsigned FreeBanks;
};
class CandidateList : public std::map<unsigned, std::list<Candidate>> {
public:
void push(unsigned Weight, const Candidate&& C) {
operator[](Weight).push_front(C);
}
Candidate &back() {
return rbegin()->second.back();
}
void pop_back() {
rbegin()->second.pop_back();
if (rbegin()->second.empty())
erase(rbegin()->first);
}
#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void dump(const GCNRegBankReassign *P) const {
dbgs() << "\nCandidates:\n\n";
for (auto &B : *this) {
dbgs() << " Weight " << B.first << ":\n";
for (auto &C : B.second)
C.dump(P);
}
dbgs() << "\n\n";
}
#endif
};
public:
static char ID;
public:
GCNRegBankReassign(RegBankReassignMode Mode = RM_BOTH)
: MachineFunctionPass(ID), Mode(Mode) {
initializeGCNRegBankReassignPass(*PassRegistry::getPassRegistry());
}
bool runOnMachineFunction(MachineFunction &MF) override;
StringRef getPassName() const override { return "GCN RegBank Reassign"; }
void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.addRequired<MachineLoopInfo>();
AU.addRequired<LiveIntervals>();
AU.addRequired<VirtRegMap>();
AU.addRequired<LiveRegMatrix>();
AU.setPreservesAll();
MachineFunctionPass::getAnalysisUsage(AU);
}
private:
const GCNSubtarget *ST;
const MachineRegisterInfo *MRI;
const SIRegisterInfo *TRI;
MachineLoopInfo *MLI;
VirtRegMap *VRM;
LiveRegMatrix *LRM;
LiveIntervals *LIS;
RegBankReassignMode Mode;
unsigned MaxNumVGPRs;
unsigned MaxNumSGPRs;
BitVector RegsUsed;
SmallVector<OperandMask, 8> OperandMasks;
CandidateList Candidates;
const MCPhysReg *CSRegs;
// Returns bank for a phys reg.
unsigned getPhysRegBank(Register Reg, unsigned SubReg) const;
// Return a bit set for each register bank used. 4 banks for VGPRs and
// 8 banks for SGPRs.
// Registers already processed and recorded in RegsUsed are excluded.
// If Bank is not -1 assume Reg:SubReg to belong to that Bank.
uint32_t getRegBankMask(Register Reg, unsigned SubReg, int Bank);
// Analyze one instruction returning the number of stalls and a mask of the
// banks used by all operands.
// If Reg and Bank are provided, assume all uses of Reg will be replaced with
// a register chosen from Bank.
std::pair<unsigned, unsigned> analyzeInst(const MachineInstr &MI,
Register Reg = Register(),
unsigned SubReg = 0, int Bank = -1);
// Return true if register is regular VGPR or SGPR or their tuples.
// Returns false for special registers like m0, vcc etc.
bool isReassignable(Register Reg) const;
// Check if registers' defs are old and may be pre-loaded.
// Returns 0 if both registers are old enough, 1 or 2 if one or both
// registers will not likely be pre-loaded.
unsigned getOperandGatherWeight(const MachineInstr& MI,
Register Reg1,
Register Reg2,
unsigned StallCycles) const;
// Find all bank bits in UsedBanks where Mask can be relocated to.
unsigned getFreeBanks(unsigned Mask, unsigned UsedBanks) const;
// Find all bank bits in UsedBanks where Mask can be relocated to.
// Bank is relative to the register and not its subregister component.
// Returns 0 is a register is not reassignable.
unsigned getFreeBanks(Register Reg, unsigned SubReg, unsigned Mask,
unsigned UsedBanks) const;
// Add cadidate instruction to the work list.
void collectCandidates(MachineInstr& MI, unsigned UsedBanks,
unsigned StallCycles);
// Collect cadidate instructions across function. Returns a number stall
// cycles detected. Only counts stalls if Collect is false.
unsigned collectCandidates(MachineFunction &MF, bool Collect = true);
// Remove all candidates that read specified register.
void removeCandidates(Register Reg);
// Compute stalls within the uses of SrcReg replaced by a register from
// Bank. If Bank is -1 does not perform substitution. If Collect is set
// candidates are collected and added to work list.
unsigned computeStallCycles(Register SrcReg,
Register Reg = Register(),
unsigned SubReg = 0, int Bank = -1,
bool Collect = false);
// Search for a register in Bank unused within LI.
// Returns phys reg or NoRegister.
MCRegister scavengeReg(LiveInterval &LI, unsigned Bank,
unsigned SubReg) const;
// Try to reassign candidate. Returns number or stall cycles saved.
unsigned tryReassign(Candidate &C);
bool verifyCycles(MachineFunction &MF,
unsigned OriginalCycles, unsigned CyclesSaved);
#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
public:
Printable printReg(Register Reg, unsigned SubReg = 0) const {
return Printable([Reg, SubReg, this](raw_ostream &OS) {
if (Reg.isPhysical()) {
OS << llvm::printReg(Reg, TRI);
return;
}
if (!VRM->isAssignedReg(Reg))
OS << "<unassigned> " << llvm::printReg(Reg, TRI);
else
OS << llvm::printReg(Reg, TRI) << '('
<< llvm::printReg(VRM->getPhys(Reg), TRI) << ')';
if (SubReg)
OS << ':' << TRI->getSubRegIndexName(SubReg);
});
}
static Printable printBank(unsigned Bank) {
return Printable([Bank](raw_ostream &OS) {
OS << ((Bank >= SGPR_BANK_OFFSET) ? Bank - SGPR_BANK_OFFSET : Bank);
});
}
static void dumpFreeBanks(unsigned FreeBanks) {
for (unsigned L = 0; L < NUM_BANKS; ++L)
if (FreeBanks & (1 << L))
dbgs() << printBank(L) << ' ';
}
#endif
};
} // End anonymous namespace.
INITIALIZE_PASS_BEGIN(GCNRegBankReassign, DEBUG_TYPE, "GCN RegBank Reassign",
false, false)
INITIALIZE_PASS_DEPENDENCY(LiveIntervals)
INITIALIZE_PASS_DEPENDENCY(MachineLoopInfo)
INITIALIZE_PASS_DEPENDENCY(VirtRegMap)
INITIALIZE_PASS_DEPENDENCY(LiveRegMatrix)
INITIALIZE_PASS_END(GCNRegBankReassign, DEBUG_TYPE, "GCN RegBank Reassign",
false, false)
char GCNRegBankReassign::ID = 0;
char &llvm::GCNRegBankReassignID = GCNRegBankReassign::ID;
unsigned GCNRegBankReassign::getPhysRegBank(Register Reg,
unsigned SubReg) const {
assert(Reg.isPhysical());
const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg);
unsigned Size = TRI->getRegSizeInBits(*RC);
if (Size == 16)
Reg = TRI->get32BitRegister(Reg);
else if (Size > 32) {
if (SubReg) {
const TargetRegisterClass *SubRC = TRI->getSubRegClass(RC, SubReg);
Reg = TRI->getSubReg(Reg, SubReg);
if (TRI->getRegSizeInBits(*SubRC) > 32)
Reg = TRI->getSubReg(Reg, AMDGPU::sub0);
} else {
Reg = TRI->getSubReg(Reg, AMDGPU::sub0);
}
}
if (TRI->hasVGPRs(RC)) {
unsigned RegNo = Reg - AMDGPU::VGPR0;
return RegNo % NUM_VGPR_BANKS;
}
unsigned RegNo = TRI->getEncodingValue(AMDGPU::getMCReg(Reg, *ST)) / 2;
return RegNo % NUM_SGPR_BANKS + SGPR_BANK_OFFSET;
}
uint32_t GCNRegBankReassign::getRegBankMask(Register Reg, unsigned SubReg,
int Bank) {
if (Reg.isVirtual()) {
if (!VRM->isAssignedReg(Reg))
return 0;
Reg = VRM->getPhys(Reg);
if (!Reg)
return 0;
if (SubReg)
Reg = TRI->getSubReg(Reg, SubReg);
}
const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg);
unsigned Size = TRI->getRegSizeInBits(*RC);
if (Size == 16) {
Reg = TRI->get32BitRegister(Reg);
Size = 1;
} else {
Size /= 32;
if (Size > 1)
Reg = TRI->getSubReg(Reg, AMDGPU::sub0);
}
if (TRI->hasVGPRs(RC)) {
// VGPRs have 4 banks assigned in a round-robin fashion.
unsigned RegNo = Reg - AMDGPU::VGPR0;
uint32_t Mask = maskTrailingOnes<uint32_t>(Size);
unsigned Used = 0;
// Bitmask lacks an extract method
for (unsigned I = 0; I < Size; ++I)
if (RegsUsed.test(RegNo + I))
Used |= 1 << I;
RegsUsed.set(RegNo, RegNo + Size);
Mask &= ~Used;
Mask <<= (Bank == -1) ? RegNo % NUM_VGPR_BANKS : uint32_t(Bank);
return (Mask | (Mask >> NUM_VGPR_BANKS)) & VGPR_BANK_MASK;
}
// SGPRs have 8 banks holding 2 consequitive registers each.
unsigned RegNo = TRI->getEncodingValue(AMDGPU::getMCReg(Reg, *ST)) / 2;
unsigned StartBit = AMDGPU::VGPR_32RegClass.getNumRegs();
if (RegNo + StartBit >= RegsUsed.size())
return 0;
if (Size > 1)
Size /= 2;
unsigned Mask = (1 << Size) - 1;
unsigned Used = 0;
for (unsigned I = 0; I < Size; ++I)
if (RegsUsed.test(StartBit + RegNo + I))
Used |= 1 << I;
RegsUsed.set(StartBit + RegNo, StartBit + RegNo + Size);
Mask &= ~Used;
Mask <<= (Bank == -1) ? RegNo % NUM_SGPR_BANKS
: unsigned(Bank - SGPR_BANK_OFFSET);
Mask = (Mask | (Mask >> NUM_SGPR_BANKS)) & SGPR_BANK_SHIFTED_MASK;
// Reserve 4 bank ids for VGPRs.
return Mask << SGPR_BANK_OFFSET;
}
std::pair<unsigned, unsigned>
GCNRegBankReassign::analyzeInst(const MachineInstr &MI, Register Reg,
unsigned SubReg, int Bank) {
unsigned StallCycles = 0;
unsigned UsedBanks = 0;
if (MI.isMetaInstruction())
return std::make_pair(StallCycles, UsedBanks);
if (!(Mode & RM_SGPR) &&
MI.getDesc().TSFlags & (SIInstrFlags::SMRD | SIInstrFlags::SALU))
return std::make_pair(StallCycles, UsedBanks);
RegsUsed.reset();
OperandMasks.clear();
for (const auto& Op : MI.explicit_uses()) {
// Undef can be assigned to any register, so two vregs can be assigned
// the same phys reg within the same instruction.
if (!Op.isReg() || Op.isUndef())
continue;
const Register R = Op.getReg();
const TargetRegisterClass *RC = TRI->getRegClassForReg(*MRI, R);
// Do not compute stalls for AGPRs
if (TRI->hasAGPRs(RC))
continue;
if ((Mode != RM_BOTH) && !(Mode & (TRI->hasVGPRs(RC) ? RM_VGPR : RM_SGPR)))
continue;
// Do not compute stalls if sub-register covers all banks
if (Op.getSubReg()) {
LaneBitmask LM = TRI->getSubRegIndexLaneMask(Op.getSubReg());
if (TRI->hasVGPRs(RC)) {
if (TRI->getNumCoveredRegs(LM) >= NUM_VGPR_BANKS)
continue;
} else {
if (TRI->getNumCoveredRegs(LM) / 2 >= NUM_SGPR_BANKS)
continue;
}
}
unsigned ShiftedBank = Bank;
if (Bank != -1 && R == Reg && (Op.getSubReg() || SubReg)) {
unsigned RegOffset =
TRI->getChannelFromSubReg(SubReg ? SubReg : (unsigned)AMDGPU::sub0);
unsigned Offset = TRI->getChannelFromSubReg(
Op.getSubReg() ? Op.getSubReg() : (unsigned)AMDGPU::sub0);
if (Bank < NUM_VGPR_BANKS) {
unsigned Shift = ((NUM_VGPR_BANKS + Offset) - RegOffset);
ShiftedBank = (Bank + Shift) % NUM_VGPR_BANKS;
} else if (Bank >= SGPR_BANK_OFFSET) {
unsigned Shift = (NUM_SGPR_BANKS + (Offset >> 1)) - (RegOffset >> 1);
ShiftedBank = SGPR_BANK_OFFSET +
(Bank - SGPR_BANK_OFFSET + Shift) % NUM_SGPR_BANKS;
}
}
uint32_t Mask = getRegBankMask(R, Op.getSubReg(),
(Reg == R) ? ShiftedBank : -1);
StallCycles += countPopulation(UsedBanks & Mask);
UsedBanks |= Mask;
OperandMasks.push_back(OperandMask(Op.getReg(), Op.getSubReg(), Mask));
}
return std::make_pair(StallCycles, UsedBanks);
}
unsigned GCNRegBankReassign::getOperandGatherWeight(const MachineInstr& MI,
Register Reg1,
Register Reg2,
unsigned StallCycles) const
{
unsigned Defs = 0;
MachineBasicBlock::const_instr_iterator Def(MI.getIterator());
MachineBasicBlock::const_instr_iterator B(MI.getParent()->instr_begin());
for (unsigned S = StallCycles; S && Def != B && Defs != 3; --S) {
if (MI.isDebugInstr())
continue;
--Def;
if (Def->getOpcode() == TargetOpcode::IMPLICIT_DEF)
continue;
if (Def->modifiesRegister(Reg1, TRI))
Defs |= 1;
if (Def->modifiesRegister(Reg2, TRI))
Defs |= 2;
}
return countPopulation(Defs);
}
bool GCNRegBankReassign::isReassignable(Register Reg) const {
if (Reg.isPhysical() || !VRM->isAssignedReg(Reg))
return false;
// InlineSpiller does not call LRM::assign() after an LI split leaving it
// in an inconsistent state, so we cannot call LRM::unassign().
// See llvm bug #48911.
// Skip reassign if a register has originated from such split.
// FIXME: Remove the workaround when bug #48911 is fixed.
if (VRM->getPreSplitReg(Reg))
return false;
const MachineInstr *Def = MRI->getUniqueVRegDef(Reg);
Register PhysReg = VRM->getPhys(Reg);
if (Def && Def->isCopy() && Def->getOperand(1).getReg() == PhysReg)
return false;
for (auto U : MRI->use_nodbg_operands(Reg)) {
if (U.isImplicit())
return false;
const MachineInstr *UseInst = U.getParent();
if (UseInst->isCopy() && UseInst->getOperand(0).getReg() == PhysReg)
return false;
}
const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(PhysReg);
unsigned Size = TRI->getRegSizeInBits(*RC);
// TODO: Support 16 bit registers. Those needs to be moved with their
// parent VGPR_32 and potentially a sibling 16 bit sub-register.
if (Size < 32)
return false;
if (TRI->hasVGPRs(RC))
return true;
if (Size == 16)
return AMDGPU::SGPR_LO16RegClass.contains(PhysReg);
if (Size > 32)
PhysReg = TRI->getSubReg(PhysReg, AMDGPU::sub0);
return AMDGPU::SGPR_32RegClass.contains(PhysReg);
}
unsigned GCNRegBankReassign::getFreeBanks(unsigned Mask,
unsigned UsedBanks) const {
unsigned Size = countPopulation(Mask);
unsigned FreeBanks = 0;
unsigned Bank = findFirstSet(Mask);
UsedBanks &= ~Mask;
// Find free VGPR banks
if ((Mask & VGPR_BANK_MASK) && (Size < NUM_VGPR_BANKS)) {
for (unsigned I = 0; I < NUM_VGPR_BANKS; ++I) {
if (Bank == I)
continue;
unsigned NewMask = ((1 << Size) - 1) << I;
NewMask = (NewMask | (NewMask >> NUM_VGPR_BANKS)) & VGPR_BANK_MASK;
if (!(UsedBanks & NewMask))
FreeBanks |= 1 << I;
}
return FreeBanks;
}
// Find free SGPR banks
// SGPR tuples must be aligned, so step is size in banks it
// crosses.
Bank -= SGPR_BANK_OFFSET;
for (unsigned I = 0; I < NUM_SGPR_BANKS; I += Size) {
if (Bank == I)
continue;
unsigned NewMask = ((1 << Size) - 1) << I;
NewMask = (NewMask | (NewMask >> NUM_SGPR_BANKS)) & SGPR_BANK_SHIFTED_MASK;
if (!(UsedBanks & (NewMask << SGPR_BANK_OFFSET)))
FreeBanks |= (1 << SGPR_BANK_OFFSET) << I;
}
return FreeBanks;
}
unsigned GCNRegBankReassign::getFreeBanks(Register Reg,
unsigned SubReg,
unsigned Mask,
unsigned UsedBanks) const {
if (!isReassignable(Reg))
return 0;
unsigned FreeBanks = getFreeBanks(Mask, UsedBanks);
unsigned Offset = TRI->getChannelFromSubReg(SubReg);
if (Offset && (Mask & VGPR_BANK_MASK)) {
unsigned Shift = Offset;
if (Shift >= NUM_VGPR_BANKS)
return 0;
unsigned VB = FreeBanks & VGPR_BANK_MASK;
FreeBanks = ((VB >> Shift) | (VB << (NUM_VGPR_BANKS - Shift))) &
VGPR_BANK_MASK;
} else if (Offset > 1 && (Mask & SGPR_BANK_MASK)) {
unsigned Shift = Offset >> 1;
if (Shift >= NUM_SGPR_BANKS)
return 0;
unsigned SB = FreeBanks >> SGPR_BANK_OFFSET;
FreeBanks = ((SB >> Shift) | (SB << (NUM_SGPR_BANKS - Shift))) &
SGPR_BANK_SHIFTED_MASK;
FreeBanks <<= SGPR_BANK_OFFSET;
}
LLVM_DEBUG(if (FreeBanks) {
dbgs() << "Potential reassignments of " << printReg(Reg, SubReg)
<< " to banks: "; dumpFreeBanks(FreeBanks);
dbgs() << '\n'; });
return FreeBanks;
}
void GCNRegBankReassign::collectCandidates(MachineInstr& MI,
unsigned UsedBanks,
unsigned StallCycles) {
LLVM_DEBUG(MI.dump());
if (!StallCycles)
return;
LLVM_DEBUG(dbgs() << "Stall cycles = " << StallCycles << '\n');
for (unsigned I = 0, E = OperandMasks.size(); I + 1 < E; ++I) {
for (unsigned J = I + 1; J != E; ++J) {
if (!(OperandMasks[I].Mask & OperandMasks[J].Mask))
continue;
Register Reg1 = OperandMasks[I].Reg;
Register Reg2 = OperandMasks[J].Reg;
unsigned SubReg1 = OperandMasks[I].SubReg;
unsigned SubReg2 = OperandMasks[J].SubReg;
unsigned Mask1 = OperandMasks[I].Mask;
unsigned Mask2 = OperandMasks[J].Mask;
unsigned Size1 = countPopulation(Mask1);
unsigned Size2 = countPopulation(Mask2);
LLVM_DEBUG(dbgs() << "Conflicting operands: " << printReg(Reg1, SubReg1) <<
" and " << printReg(Reg2, SubReg2) << '\n');
unsigned Weight = getOperandGatherWeight(MI, Reg1, Reg2, StallCycles);
Weight += MLI->getLoopDepth(MI.getParent()) * 10;
LLVM_DEBUG(dbgs() << "Stall weight = " << Weight << '\n');
unsigned FreeBanks1 = getFreeBanks(Reg1, SubReg1, Mask1, UsedBanks);
unsigned FreeBanks2 = getFreeBanks(Reg2, SubReg2, Mask2, UsedBanks);
if (FreeBanks1)
Candidates.push(Weight + ((Size2 > Size1) ? 1 : 0),
Candidate(&MI, Reg1, SubReg1, FreeBanks1));
if (FreeBanks2)
Candidates.push(Weight + ((Size1 > Size2) ? 1 : 0),
Candidate(&MI, Reg2, SubReg2, FreeBanks2));
}
}
}
unsigned GCNRegBankReassign::computeStallCycles(Register SrcReg, Register Reg,
unsigned SubReg, int Bank,
bool Collect) {
unsigned TotalStallCycles = 0;
SmallSet<const MachineInstr *, 16> Visited;
for (auto &MI : MRI->use_nodbg_instructions(SrcReg)) {
if (MI.isBundle())
continue;
if (!Visited.insert(&MI).second)
continue;
unsigned StallCycles;
unsigned UsedBanks;
std::tie(StallCycles, UsedBanks) = analyzeInst(MI, Reg, SubReg, Bank);
TotalStallCycles += StallCycles;
if (Collect)
collectCandidates(MI, UsedBanks, StallCycles);
}
return TotalStallCycles;
}
MCRegister GCNRegBankReassign::scavengeReg(LiveInterval &LI, unsigned Bank,
unsigned SubReg) const {
const TargetRegisterClass *RC = MRI->getRegClass(LI.reg());
unsigned MaxNumRegs = (Bank < NUM_VGPR_BANKS) ? MaxNumVGPRs
: MaxNumSGPRs;
unsigned MaxReg = MaxNumRegs + (Bank < NUM_VGPR_BANKS ? AMDGPU::VGPR0
: AMDGPU::SGPR0);
for (MCRegister Reg : RC->getRegisters()) {
// Check occupancy limit.
if (TRI->isSubRegisterEq(Reg, MaxReg))
break;
if (!MRI->isAllocatable(Reg) || getPhysRegBank(Reg, SubReg) != Bank)
continue;
for (unsigned I = 0; CSRegs[I]; ++I)
if (TRI->isSubRegisterEq(Reg, CSRegs[I]) &&
!LRM->isPhysRegUsed(CSRegs[I]))
return MCRegister::from(AMDGPU::NoRegister);
LLVM_DEBUG(dbgs() << "Trying register " << printReg(Reg) << '\n');
if (!LRM->checkInterference(LI, Reg))
return Reg;
}
return MCRegister::from(AMDGPU::NoRegister);
}
unsigned GCNRegBankReassign::tryReassign(Candidate &C) {
if (!LIS->hasInterval(C.Reg))
return 0;
LiveInterval &LI = LIS->getInterval(C.Reg);
LLVM_DEBUG(dbgs() << "Try reassign " << printReg(C.Reg) << " in "; C.MI->dump();
LI.dump());
// For each candidate bank walk all instructions in the range of live
// interval and check if replacing the register with one belonging to
// the candidate bank reduces conflicts.
unsigned OrigStalls = computeStallCycles(C.Reg);
LLVM_DEBUG(dbgs() << "--- Stall cycles in range = " << OrigStalls << '\n');
if (!OrigStalls)
return 0;
struct BankStall {
BankStall(unsigned b, unsigned s) : Bank(b), Stalls(s) {};
bool operator<(const BankStall &RHS) const {
if (Stalls == RHS.Stalls)
return Bank < RHS.Bank;
return Stalls > RHS.Stalls;
}
unsigned Bank;
unsigned Stalls;
};
SmallVector<BankStall, 8> BankStalls;
for (int Bank = 0; Bank < NUM_BANKS; ++Bank) {
if (C.FreeBanks & (1 << Bank)) {
LLVM_DEBUG(dbgs() << "Trying bank " << printBank(Bank) << '\n');
unsigned Stalls = computeStallCycles(C.Reg, C.Reg, C.SubReg, Bank);
if (Stalls < OrigStalls) {
LLVM_DEBUG(dbgs() << "With bank " << printBank(Bank) << " -> "
<< Stalls << '\n');
BankStalls.push_back(BankStall((unsigned)Bank, Stalls));
}
}
}
llvm::sort(BankStalls);
MCRegister OrigReg = VRM->getPhys(C.Reg);
LRM->unassign(LI);
while (!BankStalls.empty()) {
BankStall BS = BankStalls.pop_back_val();
MCRegister Reg = scavengeReg(LI, BS.Bank, C.SubReg);
if (Reg == AMDGPU::NoRegister) {
LLVM_DEBUG(dbgs() << "No free registers in bank " << printBank(BS.Bank)
<< '\n');
continue;
}
LLVM_DEBUG(dbgs() << "Found free register " << printReg(Reg)
<< (LRM->isPhysRegUsed(Reg) ? "" : " (new)")
<< " in bank " << printBank(BS.Bank) << '\n');
LRM->assign(LI, Reg);
LLVM_DEBUG(dbgs() << "--- Cycles saved: " << OrigStalls - BS.Stalls << '\n');
return OrigStalls - BS.Stalls;
}
LRM->assign(LI, OrigReg);
return 0;
}
unsigned GCNRegBankReassign::collectCandidates(MachineFunction &MF,
bool Collect) {
unsigned TotalStallCycles = 0;
for (MachineBasicBlock &MBB : MF) {
LLVM_DEBUG(if (Collect) {
if (MBB.getName().empty()) dbgs() << "bb." << MBB.getNumber();
else dbgs() << MBB.getName(); dbgs() << ":\n";
});
for (MachineInstr &MI : MBB.instrs()) {
if (MI.isBundle())
continue; // we analyze the instructions inside the bundle individually
unsigned StallCycles;
unsigned UsedBanks;
std::tie(StallCycles, UsedBanks) = analyzeInst(MI);
if (Collect)
collectCandidates(MI, UsedBanks, StallCycles);
TotalStallCycles += StallCycles;
}
LLVM_DEBUG(if (Collect) { dbgs() << '\n'; });
}
return TotalStallCycles;
}
void GCNRegBankReassign::removeCandidates(Register Reg) {
typename CandidateList::iterator Next;
for (auto I = Candidates.begin(), E = Candidates.end(); I != E; I = Next) {
Next = std::next(I);
I->second.remove_if([Reg, this](const Candidate& C) {
return C.MI->readsRegister(Reg, TRI);
});
if (I->second.empty())
Candidates.erase(I);
}
}
bool GCNRegBankReassign::verifyCycles(MachineFunction &MF,
unsigned OriginalCycles,
unsigned CyclesSaved) {
unsigned StallCycles = collectCandidates(MF, false);
LLVM_DEBUG(dbgs() << "=== After the pass " << StallCycles
<< " stall cycles left\n");
return StallCycles + CyclesSaved == OriginalCycles;
}
bool GCNRegBankReassign::runOnMachineFunction(MachineFunction &MF) {
ST = &MF.getSubtarget<GCNSubtarget>();
if (!ST->hasRegisterBanking() || skipFunction(MF.getFunction()))
return false;
MRI = &MF.getRegInfo();
LLVM_DEBUG(dbgs() << "=== RegBanks reassign analysis on function "
<< MF.getName() << '\n'
<< ((Mode & RM_VGPR) ? "VGPR " : "")
<< ((Mode & RM_SGPR) ? "SGPR " : "") << "mode\n"
<< "NumVirtRegs = " << MRI->getNumVirtRegs() << "\n\n");
if (MRI->getNumVirtRegs() > VRegThresh) {
LLVM_DEBUG(dbgs() << "NumVirtRegs > " << VRegThresh
<< " threshold, skipping function.\n\n");
return false;
}
TRI = ST->getRegisterInfo();
MLI = &getAnalysis<MachineLoopInfo>();
VRM = &getAnalysis<VirtRegMap>();
LRM = &getAnalysis<LiveRegMatrix>();
LIS = &getAnalysis<LiveIntervals>();
const SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
unsigned Occupancy = MFI->getOccupancy();
MaxNumVGPRs = ST->getMaxNumVGPRs(MF);
MaxNumSGPRs = ST->getMaxNumSGPRs(MF);
MaxNumVGPRs = std::min(ST->getMaxNumVGPRs(Occupancy), MaxNumVGPRs);
MaxNumSGPRs = std::min(ST->getMaxNumSGPRs(Occupancy, true), MaxNumSGPRs);
CSRegs = MRI->getCalleeSavedRegs();
unsigned NumRegBanks = AMDGPU::VGPR_32RegClass.getNumRegs() +
// Not a tight bound
AMDGPU::SReg_32RegClass.getNumRegs() / 2 + 1;
RegsUsed.resize(NumRegBanks);
unsigned StallCycles = collectCandidates(MF);
NumStallsDetected += StallCycles;
LLVM_DEBUG(dbgs() << "=== " << StallCycles << " stall cycles detected in "
"function " << MF.getName() << '\n');
LLVM_DEBUG(Candidates.dump(this));
unsigned CyclesSaved = 0;
while (!Candidates.empty()) {
Candidate C = Candidates.back();
unsigned LocalCyclesSaved = tryReassign(C);
CyclesSaved += LocalCyclesSaved;
if (VerifyStallCycles > 1 && !verifyCycles(MF, StallCycles, CyclesSaved))
report_fatal_error("RegBank reassign stall cycles verification failed.");
Candidates.pop_back();
if (LocalCyclesSaved) {
removeCandidates(C.Reg);
computeStallCycles(C.Reg, AMDGPU::NoRegister, 0, -1, true);
LLVM_DEBUG(Candidates.dump(this));
}
}
NumStallsRecovered += CyclesSaved;
LLVM_DEBUG(dbgs() << "=== After the pass " << CyclesSaved
<< " cycles saved in function " << MF.getName() << '\n');
Candidates.clear();
if (VerifyStallCycles == 1 && !verifyCycles(MF, StallCycles, CyclesSaved))
report_fatal_error("RegBank reassign stall cycles verification failed.");
RegsUsed.clear();
return CyclesSaved > 0;
}
MachineFunctionPass *
llvm::createGCNRegBankReassignPass(RegBankReassignMode Mode) {
return new GCNRegBankReassign(Mode);
}

llvm/test/CodeGen/AMDGPU/GlobalISel/ashr.ll

Show First 20 Lines • Show All 1,637 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_ashrrev_i64 v[0:1], v4, v[0:1] ; GFX9-NEXT: v_ashrrev_i64 v[0:1], v4, v[0:1]
; GFX9-NEXT: v_ashrrev_i64 v[2:3], v6, v[2:3] ; GFX9-NEXT: v_ashrrev_i64 v[2:3], v6, v[2:3]
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_ashr_v2i64: ; GFX10-LABEL: v_ashr_v2i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v10, v0 ; GFX10-NEXT: v_ashrrev_i64 v[0:1], v4, v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v11, v1 ; GFX10-NEXT: v_ashrrev_i64 v[2:3], v6, v[2:3]
; GFX10-NEXT: v_mov_b32_e32 v7, v2
; GFX10-NEXT: v_mov_b32_e32 v8, v3
; GFX10-NEXT: v_ashrrev_i64 v[0:1], v4, v[10:11]
; GFX10-NEXT: v_ashrrev_i64 v[2:3], v6, v[7:8]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = ashr <2 x i64> %value, %amount %result = ashr <2 x i64> %value, %amount
ret <2 x i64> %result ret <2 x i64> %result
} }
define <2 x i64> @v_ashr_v2i64_31(<2 x i64> %value) { define <2 x i64> @v_ashr_v2i64_31(<2 x i64> %value) {
; GFX6-LABEL: v_ashr_v2i64_31: ; GFX6-LABEL: v_ashr_v2i64_31:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 45 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i128.ll

Show First 20 Lines • Show All 308 Lines • ▼ Show 20 Lines
; GFX10-NEXT: global_load_dwordx4 v[3:6], v[0:1], off ; GFX10-NEXT: global_load_dwordx4 v[3:6], v[0:1], off
; GFX10-NEXT: global_load_dwordx4 v[7:10], v[0:1], off offset:16 ; GFX10-NEXT: global_load_dwordx4 v[7:10], v[0:1], off offset:16
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 1, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 1, v2
; GFX10-NEXT: global_load_dwordx4 v[11:14], v[0:1], off offset:32 ; GFX10-NEXT: global_load_dwordx4 v[11:14], v[0:1], off offset:32
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2
; GFX10-NEXT: v_add_nc_u32_e32 v19, 1, v2 ; GFX10-NEXT: v_add_nc_u32_e32 v19, 1, v2
; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 1, v19 ; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 1, v19
; GFX10-NEXT: s_waitcnt vmcnt(2) ; GFX10-NEXT: s_waitcnt vmcnt(2)
; GFX10-NEXT: v_cndmask_b32_e32 v18, v3, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v15, v3, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v15, v4, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v16, v4, v6, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v2 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v2
; GFX10-NEXT: v_cndmask_b32_e64 v23, v4, v6, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v5, s4
; GFX10-NEXT: v_cndmask_b32_e64 v22, v3, v5, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, v6, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 2, v19 ; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 2, v19
; GFX10-NEXT: s_waitcnt vmcnt(1) ; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_cndmask_b32_e32 v27, v18, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v15, v7, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v15, v15, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v6, v16, v8, vcc_lo
; GFX10-NEXT: global_load_dwordx4 v[15:18], v[0:1], off offset:48
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v2 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v2
; GFX10-NEXT: v_cndmask_b32_e64 v3, v22, v7, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v7, s4
; GFX10-NEXT: v_cndmask_b32_e64 v4, v23, v8, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, v8, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 3, v19 ; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 3, v19
; GFX10-NEXT: v_cndmask_b32_e32 v6, v15, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v9, vcc_lo
; GFX10-NEXT: global_load_dwordx4 v[15:18], v[0:1], off offset:48 ; GFX10-NEXT: v_cndmask_b32_e32 v6, v6, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v5, v27, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v2 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v2
; GFX10-NEXT: v_cndmask_b32_e64 v22, v3, v9, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v9, s4
; GFX10-NEXT: v_cndmask_b32_e64 v7, v4, v10, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, v10, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 4, v19 ; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 4, v19
; GFX10-NEXT: s_waitcnt vmcnt(1) ; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_cndmask_b32_e32 v0, v5, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v5, v11, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v6, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v6, v12, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v2 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v2
; GFX10-NEXT: v_cndmask_b32_e64 v3, v22, v11, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v11, s4
; GFX10-NEXT: v_cndmask_b32_e64 v4, v7, v12, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, v12, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 5, v19 ; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 5, v19
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v13, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v14, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v2 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v2
; GFX10-NEXT: v_cndmask_b32_e64 v6, v3, v13, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v13, s4
; GFX10-NEXT: v_cndmask_b32_e64 v7, v4, v14, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, v14, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 6, v19 ; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 6, v19
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v15, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v15, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v16, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v16, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v2 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v2
; GFX10-NEXT: v_cndmask_b32_e64 v3, v6, v15, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v15, s4
; GFX10-NEXT: v_cndmask_b32_e64 v4, v7, v16, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, v16, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 7, v19 ; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 7, v19
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v17, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v17, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v18, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v18, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v3, v17, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v3, v17, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, v4, v18, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v4, v18, s4
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%vector = load <4 x i128>, <4 x i128> addrspace(1)* %ptr %vector = load <4 x i128>, <4 x i128> addrspace(1)* %ptr
%element = extractelement <4 x i128> %vector, i32 %idx %element = extractelement <4 x i128> %vector, i32 %idx
▲ Show 20 Lines • Show All 208 Lines • ▼ Show 20 Lines
; GFX7-NEXT: v_readfirstlane_b32 s0, v0 ; GFX7-NEXT: v_readfirstlane_b32 s0, v0
; GFX7-NEXT: v_readfirstlane_b32 s1, v1 ; GFX7-NEXT: v_readfirstlane_b32 s1, v1
; GFX7-NEXT: v_readfirstlane_b32 s2, v2 ; GFX7-NEXT: v_readfirstlane_b32 s2, v2
; GFX7-NEXT: v_readfirstlane_b32 s3, v3 ; GFX7-NEXT: v_readfirstlane_b32 s3, v3
; GFX7-NEXT: ; return to shader part epilog ; GFX7-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: extractelement_sgpr_v4i128_vgpr_idx: ; GFX10-LABEL: extractelement_sgpr_v4i128_vgpr_idx:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx16 s[8:23], s[2:3], 0x0 ; GFX10-NEXT: s_load_dwordx16 s[4:19], s[2:3], 0x0
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 1, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX10-NEXT: v_add_nc_u32_e32 v1, 1, v0 ; GFX10-NEXT: v_add_nc_u32_e32 v1, 1, v0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 1, v1 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 1, v1
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v2, s10 ; GFX10-NEXT: v_mov_b32_e32 v2, s6
; GFX10-NEXT: v_mov_b32_e32 v3, s11 ; GFX10-NEXT: v_mov_b32_e32 v3, s7
; GFX10-NEXT: v_cndmask_b32_e32 v4, s8, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, s4, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v5, s9, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, s5, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: v_cndmask_b32_e64 v2, s8, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, s4, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, s9, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, s5, v3, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v1 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v1
; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, s8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, s9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s8, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s9, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 3, v1
; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, s10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, s11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s10, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s11, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 4, v1
; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, s12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, s12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, s13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, s13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s12, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s12, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s13, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s13, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 3, v1 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 5, v1
; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, s14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, s14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, s15, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, s15, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s14, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s14, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s15, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s15, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s14, 4, v1 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 6, v1
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 5, v1
; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, s16, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, s16, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, s17, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, s17, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s16, s14
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s17, s14
; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, s18, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, s19, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s18, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s19, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 6, v1
; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, s20, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, s21, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s20, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s16, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s21, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s17, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 7, v1 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 7, v1
; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, s23, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, s19, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, s22, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, s18, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s22, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s18, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s23, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s19, s0
; GFX10-NEXT: v_readfirstlane_b32 s1, v1 ; GFX10-NEXT: v_readfirstlane_b32 s1, v1
; GFX10-NEXT: v_readfirstlane_b32 s0, v0 ; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: v_readfirstlane_b32 s2, v2 ; GFX10-NEXT: v_readfirstlane_b32 s2, v2
; GFX10-NEXT: v_readfirstlane_b32 s3, v3 ; GFX10-NEXT: v_readfirstlane_b32 s3, v3
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%vector = load <4 x i128>, <4 x i128> addrspace(4)* %ptr %vector = load <4 x i128>, <4 x i128> addrspace(4)* %ptr
%element = extractelement <4 x i128> %vector, i32 %idx %element = extractelement <4 x i128> %vector, i32 %idx
ret i128 %element ret i128 %element
▲ Show 20 Lines • Show All 295 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i16.ll

Show First 20 Lines • Show All 575 Lines • ▼ Show 20 Lines
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: global_load_dwordx4 v[3:6], v[0:1], off ; GFX10-NEXT: global_load_dwordx4 v[3:6], v[0:1], off
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 1, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v0, 1, v2
; GFX10-NEXT: v_and_b32_e32 v2, 1, v2 ; GFX10-NEXT: v_and_b32_e32 v2, 1, v2
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v6, vcc_lo
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 4, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 4, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v0, v1, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%vector = load <8 x i16>, <8 x i16> addrspace(1)* %ptr %vector = load <8 x i16>, <8 x i16> addrspace(1)* %ptr
%element = extractelement <8 x i16> %vector, i32 %idx %element = extractelement <8 x i16> %vector, i32 %idx
ret i16 %element ret i16 %element
▲ Show 20 Lines • Show All 519 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.i8.ll

Show First 20 Lines • Show All 217 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_mov_b32 s4, 8 ; GFX10-NEXT: s_mov_b32 s4, 8
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 24, v0
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_movk_i32 s4, 0xff ; GFX10-NEXT: s_movk_i32 s4, 0xff
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; GFX10-NEXT: v_and_b32_sdwa v4, v0, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v4, v0, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_or_b32 v6, v0, s4, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, s4, v1
; GFX10-NEXT: v_and_b32_e32 v1, 3, v2 ; GFX10-NEXT: v_and_b32_e32 v1, 3, v2
; GFX10-NEXT: v_or3_b32 v0, v6, v4, v3 ; GFX10-NEXT: v_or3_b32 v0, v0, v4, v3
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 3, v1 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 3, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v0, v1, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%vector = load <4 x i8>, <4 x i8> addrspace(1)* %ptr %vector = load <4 x i8>, <4 x i8> addrspace(1)* %ptr
%element = extractelement <4 x i8> %vector, i32 %idx %element = extractelement <4 x i8> %vector, i32 %idx
ret i8 %element ret i8 %element
} }
▲ Show 20 Lines • Show All 794 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v1
; GFX10-NEXT: v_and_b32_sdwa v6, v0, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v6, v0, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v7, v1, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v7, v1, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5
; GFX10-NEXT: v_and_or_b32 v11, v0, s1, v2 ; GFX10-NEXT: v_and_or_b32 v0, v0, s1, v2
; GFX10-NEXT: v_and_or_b32 v2, v1, s1, v3 ; GFX10-NEXT: v_and_or_b32 v1, v1, s1, v3
; GFX10-NEXT: s_lshr_b32 s0, s2, 2 ; GFX10-NEXT: s_lshr_b32 s0, s2, 2
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 1 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 1
; GFX10-NEXT: v_or3_b32 v0, v11, v6, v4 ; GFX10-NEXT: v_or3_b32 v0, v0, v6, v4
; GFX10-NEXT: v_or3_b32 v1, v2, v7, v5 ; GFX10-NEXT: v_or3_b32 v1, v1, v7, v5
; GFX10-NEXT: s_and_b32 s0, s2, 3 ; GFX10-NEXT: s_and_b32 s0, s2, 3
; GFX10-NEXT: s_lshl_b32 s0, s0, 3 ; GFX10-NEXT: s_lshl_b32 s0, s0, 3
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_lshrrev_b32_e32 v0, s0, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX10-NEXT: v_readfirstlane_b32 s0, v0 ; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%vector = load <8 x i8>, <8 x i8> addrspace(1)* %ptr %vector = load <8 x i8>, <8 x i8> addrspace(1)* %ptr
%element = extractelement <8 x i8> %vector, i32 %idx %element = extractelement <8 x i8> %vector, i32 %idx
▲ Show 20 Lines • Show All 1,555 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v9, 24, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 8, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 8, v2
; GFX10-NEXT: v_lshlrev_b32_sdwa v6, s0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v6, s0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v5, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v5, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 24, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 24, v2
; GFX10-NEXT: v_and_b32_sdwa v13, v0, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v13, v0, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 8, v3 ; GFX10-NEXT: v_and_b32_sdwa v14, v1, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_sdwa v19, v5, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_or_b32 v0, v0, s1, v6 ; GFX10-NEXT: v_and_or_b32 v0, v0, s1, v6
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7 ; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7
; GFX10-NEXT: v_and_b32_sdwa v14, v1, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_lshrrev_b32_e32 v12, 8, v3
; GFX10-NEXT: v_lshlrev_b32_sdwa v10, v5, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9 ; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9
; GFX10-NEXT: v_and_or_b32 v23, v1, s1, v8 ; GFX10-NEXT: v_and_or_b32 v1, v1, s1, v8
; GFX10-NEXT: s_lshr_b32 s0, s2, 2 ; GFX10-NEXT: s_lshr_b32 s0, s2, 2
; GFX10-NEXT: v_lshlrev_b32_sdwa v5, v5, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v5, v5, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v17, v2, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v15, v2, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v11 ; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v11
; GFX10-NEXT: v_and_or_b32 v2, v2, s1, v19 ; GFX10-NEXT: v_and_or_b32 v2, v2, s1, v10
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v3
; GFX10-NEXT: v_or3_b32 v0, v0, v13, v7 ; GFX10-NEXT: v_or3_b32 v0, v0, v13, v7
; GFX10-NEXT: v_or3_b32 v1, v23, v14, v9 ; GFX10-NEXT: v_or3_b32 v1, v1, v14, v9
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 1 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 1
; GFX10-NEXT: v_and_or_b32 v5, v3, v4, v5 ; GFX10-NEXT: v_and_or_b32 v5, v3, v4, v5
; GFX10-NEXT: v_and_b32_sdwa v3, v3, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v3, v3, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_or3_b32 v2, v2, v17, v11 ; GFX10-NEXT: v_or3_b32 v2, v2, v15, v11
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v6 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v6
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 2 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 2
; GFX10-NEXT: v_or3_b32 v1, v5, v3, v4 ; GFX10-NEXT: v_or3_b32 v1, v5, v3, v4
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 3 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 3
; GFX10-NEXT: s_and_b32 s0, s2, 3 ; GFX10-NEXT: s_and_b32 s0, s2, 3
; GFX10-NEXT: s_lshl_b32 s0, s0, 3 ; GFX10-NEXT: s_lshl_b32 s0, s0, 3
▲ Show 20 Lines • Show All 1,690 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/extractelement.ll

Show First 20 Lines • Show All 180 Lines • ▼ Show 20 Lines
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v8
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
Show All 22 Lines
; ;
; GFX10-LABEL: dyn_extract_v8f32_v_s: ; GFX10-LABEL: dyn_extract_v8f32_v_s:
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 2 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 2
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 3 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 3
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 4 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 4
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 5 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 5
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 6 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 6
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 7 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 7
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
entry: entry:
▲ Show 20 Lines • Show All 100 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: s_mov_b64 s[6:7], 2 ; GFX10-NEXT: s_mov_b64 s[6:7], 2
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10-NEXT: v_mov_b32_e32 v1, s6 ; GFX10-NEXT: v_mov_b32_e32 v1, s6
; GFX10-NEXT: v_mov_b32_e32 v2, s7 ; GFX10-NEXT: v_mov_b32_e32 v2, s7
; GFX10-NEXT: s_mov_b64 s[4:5], 1 ; GFX10-NEXT: s_mov_b64 s[4:5], 1
; GFX10-NEXT: s_mov_b64 s[8:9], 3 ; GFX10-NEXT: s_mov_b64 s[8:9], 3
; GFX10-NEXT: s_mov_b64 s[14:15], 4 ; GFX10-NEXT: s_mov_b64 s[10:11], 4
; GFX10-NEXT: v_cndmask_b32_e32 v1, s4, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, s4, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, s5, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, s5, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: s_mov_b64 s[12:13], 5 ; GFX10-NEXT: s_mov_b64 s[12:13], 5
; GFX10-NEXT: s_mov_b64 s[14:15], 6
; GFX10-NEXT: s_mov_b64 s[16:17], 7 ; GFX10-NEXT: s_mov_b64 s[16:17], 7
; GFX10-NEXT: s_mov_b64 s[18:19], 8 ; GFX10-NEXT: s_mov_b64 s[18:19], 8
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s15, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10-NEXT: s_mov_b64 s[14:15], 6
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s15, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s15, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s16, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s16, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s17, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s17, vcc_lo
▲ Show 20 Lines • Show All 185 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_mov_b32 s5, s7 ; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8 ; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: s_mov_b32 s7, s9 ; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: s_mov_b32 s8, s10 ; GFX10-NEXT: s_mov_b32 s8, s10
; GFX10-NEXT: s_mov_b32 s9, s11 ; GFX10-NEXT: s_mov_b32 s9, s11
; GFX10-NEXT: s_mov_b32 s46, s12 ; GFX10-NEXT: s_mov_b32 s10, s12
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: s_mov_b32 s47, s13 ; GFX10-NEXT: s_mov_b32 s11, s13
; GFX10-NEXT: s_mov_b32 s12, s14 ; GFX10-NEXT: s_mov_b32 s12, s14
; GFX10-NEXT: s_mov_b32 s13, s15 ; GFX10-NEXT: s_mov_b32 s13, s15
; GFX10-NEXT: s_mov_b32 s14, s16 ; GFX10-NEXT: s_mov_b32 s14, s16
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10-NEXT: s_mov_b32 s15, s17 ; GFX10-NEXT: s_mov_b32 s15, s17
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s46, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s47, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s15, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s15, vcc_lo
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off ; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
Show All 30 Lines
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc ; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: dyn_extract_v8i64_v_v: ; GFX10-LABEL: dyn_extract_v8i64_v_v:
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16
; GFX10-NEXT: v_cndmask_b32_e32 v19, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v1, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v19, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, v1, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v7, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, v1, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v7, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
%ext = extractelement <8 x i64> %vec, i32 %sel %ext = extractelement <8 x i64> %vec, i32 %sel
ret i64 %ext ret i64 %ext
} }
▲ Show 20 Lines • Show All 203 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_add_nc_u32_e32 v8, 3, v8 ; GFX10-NEXT: v_add_nc_u32_e32 v8, 3, v8
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v8
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
▲ Show 20 Lines • Show All 481 Lines • ▼ Show 20 Lines
; MOVREL-NEXT: s_setpc_b64 s[30:31] ; MOVREL-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: dyn_extract_v8f64_v_v_offset3: ; GFX10-LABEL: dyn_extract_v8f64_v_v_offset3:
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_add_nc_u32_e32 v16, 3, v16 ; GFX10-NEXT: v_add_nc_u32_e32 v16, 3, v16
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16
; GFX10-NEXT: v_cndmask_b32_e32 v19, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v1, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v19, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, v1, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v7, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, v1, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v7, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
%add = add i32 %sel, 3 %add = add i32 %sel, 3
%ext = extractelement <8 x double> %vec, i32 %add %ext = extractelement <8 x double> %vec, i32 %add
ret double %ext ret double %ext
Show All 23 Lines
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v8
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v8 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v8
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
▲ Show 20 Lines • Show All 95 Lines • ▼ Show 20 Lines
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc ; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: dyn_extract_v8p1_v_v: ; GFX10-LABEL: dyn_extract_v8p1_v_v:
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16
; GFX10-NEXT: v_cndmask_b32_e32 v19, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v1, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v19, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, v1, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v7, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, v1, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v7, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
%ext = extractelement <8 x i8 addrspace(1)*> %vec, i32 %idx %ext = extractelement <8 x i8 addrspace(1)*> %vec, i32 %idx
ret i8 addrspace(1)* %ext ret i8 addrspace(1)* %ext
} }
▲ Show 20 Lines • Show All 438 Lines • ▼ Show 20 Lines
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v6 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v6
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v6 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v6
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v6 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v6
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v6 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v6
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v6 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v6
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
%ext = extractelement <6 x float> %vec, i32 %sel %ext = extractelement <6 x float> %vec, i32 %sel
ret float %ext ret float %ext
} }
Show All 14 Lines
; ;
; GFX10-LABEL: dyn_extract_v6f32_v_s: ; GFX10-LABEL: dyn_extract_v6f32_v_s:
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 2 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 2
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 3 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 3
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 4 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 4
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 5 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 5
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
entry: entry:
%ext = extractelement <6 x float> %vec, i32 %sel %ext = extractelement <6 x float> %vec, i32 %sel
ret float %ext ret float %ext
} }
▲ Show 20 Lines • Show All 109 Lines • ▼ Show 20 Lines
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v7 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v7
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v7 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v7
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v7 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v7
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v7 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v7
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v7 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v7
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v7 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v7
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
%ext = extractelement <7 x float> %vec, i32 %sel %ext = extractelement <7 x float> %vec, i32 %sel
ret float %ext ret float %ext
Show All 18 Lines
; ;
; GFX10-LABEL: dyn_extract_v7f32_v_s: ; GFX10-LABEL: dyn_extract_v7f32_v_s:
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 2 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 2
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 3 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 3
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 4 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 4
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 5 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 5
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 6 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 6
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
entry: entry:
%ext = extractelement <7 x float> %vec, i32 %sel %ext = extractelement <7 x float> %vec, i32 %sel
ret float %ext ret float %ext
▲ Show 20 Lines • Show All 93 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_mov_b32 s5, s7 ; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8 ; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: s_mov_b32 s7, s9 ; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: s_mov_b32 s8, s10 ; GFX10-NEXT: s_mov_b32 s8, s10
; GFX10-NEXT: s_mov_b32 s9, s11 ; GFX10-NEXT: s_mov_b32 s9, s11
; GFX10-NEXT: s_mov_b32 s14, s12 ; GFX10-NEXT: s_mov_b32 s10, s12
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: s_mov_b32 s47, s13 ; GFX10-NEXT: s_mov_b32 s11, s13
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s47, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s11, vcc_lo
; GFX10-NEXT: v_readfirstlane_b32 s0, v0 ; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: v_readfirstlane_b32 s1, v1 ; GFX10-NEXT: v_readfirstlane_b32 s1, v1
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
entry: entry:
%ext = extractelement <6 x double> %vec, i32 %sel %ext = extractelement <6 x double> %vec, i32 %sel
ret double %ext ret double %ext
} }
Show All 18 Lines
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc ; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: dyn_extract_v6f64_v_v: ; GFX10-LABEL: dyn_extract_v6f64_v_v:
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v12 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v12
; GFX10-NEXT: v_cndmask_b32_e32 v15, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v1, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v12 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v12
; GFX10-NEXT: v_cndmask_b32_e32 v0, v15, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v12 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v12
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, v1, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v12 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v12
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v7, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v12 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v12
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
%ext = extractelement <6 x double> %vec, i32 %sel %ext = extractelement <6 x double> %vec, i32 %sel
ret double %ext ret double %ext
} }
▲ Show 20 Lines • Show All 135 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_mov_b32 s5, s7 ; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8 ; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: s_mov_b32 s7, s9 ; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: s_mov_b32 s8, s10 ; GFX10-NEXT: s_mov_b32 s8, s10
; GFX10-NEXT: s_mov_b32 s9, s11 ; GFX10-NEXT: s_mov_b32 s9, s11
; GFX10-NEXT: s_mov_b32 s46, s12 ; GFX10-NEXT: s_mov_b32 s10, s12
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: s_mov_b32 s47, s13 ; GFX10-NEXT: s_mov_b32 s11, s13
; GFX10-NEXT: s_mov_b32 s12, s14 ; GFX10-NEXT: s_mov_b32 s12, s14
; GFX10-NEXT: s_mov_b32 s13, s15 ; GFX10-NEXT: s_mov_b32 s13, s15
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s46, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s47, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s13, vcc_lo
; GFX10-NEXT: v_readfirstlane_b32 s0, v0 ; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: v_readfirstlane_b32 s1, v1 ; GFX10-NEXT: v_readfirstlane_b32 s1, v1
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
entry: entry:
%ext = extractelement <7 x double> %vec, i32 %sel %ext = extractelement <7 x double> %vec, i32 %sel
Show All 24 Lines
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc ; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: dyn_extract_v7f64_v_v: ; GFX10-LABEL: dyn_extract_v7f64_v_v:
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v14 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v14
; GFX10-NEXT: v_cndmask_b32_e32 v15, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v1, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v14 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v14
; GFX10-NEXT: v_cndmask_b32_e32 v0, v15, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v14 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v14
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, v1, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v14 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v14
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v7, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v14 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v14
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, v1, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v14 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v14
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v7, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
%ext = extractelement <7 x double> %vec, i32 %sel %ext = extractelement <7 x double> %vec, i32 %sel
ret double %ext ret double %ext
} }
define amdgpu_ps double @dyn_extract_v7f64_v_s(<7 x double> %vec, i32 inreg %sel) { define amdgpu_ps double @dyn_extract_v7f64_v_s(<7 x double> %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v7f64_v_s: ; GPRIDX-LABEL: dyn_extract_v7f64_v_s:
▲ Show 20 Lines • Show All 560 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: s_mov_b32 s5, s7 ; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8 ; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: s_mov_b32 s7, s9 ; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: s_mov_b32 s8, s10 ; GFX10-NEXT: s_mov_b32 s8, s10
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: s_mov_b32 s9, s11 ; GFX10-NEXT: s_mov_b32 s9, s11
; GFX10-NEXT: s_mov_b32 s46, s12 ; GFX10-NEXT: s_mov_b32 s10, s12
; GFX10-NEXT: s_mov_b32 s47, s13 ; GFX10-NEXT: s_mov_b32 s11, s13
; GFX10-NEXT: s_mov_b32 s12, s14 ; GFX10-NEXT: s_mov_b32 s12, s14
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10-NEXT: s_mov_b32 s13, s15 ; GFX10-NEXT: s_mov_b32 s13, s15
; GFX10-NEXT: s_mov_b32 s14, s16 ; GFX10-NEXT: s_mov_b32 s14, s16
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 8, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 8, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 9, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 9, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 10, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 10, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s46, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s10, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 11, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 11, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s47, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 12, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 12, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s12, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 13, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 13, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 14, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 14, v0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s14, vcc_lo
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
entry: entry:
Show All 39 Lines
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v15
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v15
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 8, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 8, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 9, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 9, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 10, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 10, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 11, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 11, v15
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 12, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 12, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 13, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 13, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 14, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 14, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
%ext = extractelement <15 x float> %vec, i32 %sel %ext = extractelement <15 x float> %vec, i32 %sel
ret float %ext ret float %ext
▲ Show 20 Lines • Show All 196 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_add_nc_u32_e32 v15, 3, v15 ; GFX10-NEXT: v_add_nc_u32_e32 v15, 3, v15
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v15
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v15
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 8, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 8, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 9, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 9, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 10, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 10, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 11, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 11, v15
; GFX10-NEXT: v_cndmask_b32_e32 v3, v0, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 12, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 12, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 13, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 13, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 14, v15 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 14, v15
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
entry: entry:
%add = add i32 %sel, 3 %add = add i32 %sel, 3
%ext = extractelement <15 x float> %vec, i32 %add %ext = extractelement <15 x float> %vec, i32 %add
▲ Show 20 Lines • Show All 682 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/fdiv.f16.ll

Show First 20 Lines • Show All 631 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 16, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 16, v0
; GFX10-NEXT: v_cvt_f32_f16_e32 v7, v0 ; GFX10-NEXT: v_cvt_f32_f16_e32 v7, v0
; GFX10-NEXT: v_cvt_f32_f16_e32 v3, v2 ; GFX10-NEXT: v_cvt_f32_f16_e32 v3, v2
; GFX10-NEXT: v_rcp_f32_e32 v4, v4 ; GFX10-NEXT: v_rcp_f32_e32 v4, v4
; GFX10-NEXT: v_cvt_f32_f16_e32 v6, v5 ; GFX10-NEXT: v_cvt_f32_f16_e32 v6, v5
; GFX10-NEXT: v_rcp_f32_e32 v3, v3 ; GFX10-NEXT: v_rcp_f32_e32 v3, v3
; GFX10-NEXT: v_mul_f32_e32 v4, v7, v4 ; GFX10-NEXT: v_mul_f32_e32 v4, v7, v4
; GFX10-NEXT: v_mul_f32_e32 v3, v6, v3 ; GFX10-NEXT: v_mul_f32_e32 v3, v6, v3
; GFX10-NEXT: v_cvt_f16_f32_e32 v7, v4 ; GFX10-NEXT: v_cvt_f16_f32_e32 v4, v4
; GFX10-NEXT: v_cvt_f16_f32_e32 v3, v3 ; GFX10-NEXT: v_cvt_f16_f32_e32 v3, v3
; GFX10-NEXT: v_div_fixup_f16 v0, v7, v1, v0 ; GFX10-NEXT: v_div_fixup_f16 v0, v4, v1, v0
; GFX10-NEXT: v_div_fixup_f16 v2, v3, v2, v5 ; GFX10-NEXT: v_div_fixup_f16 v2, v3, v2, v5
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v2
; GFX10-NEXT: v_and_or_b32 v0, v0, 0xffff, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, 0xffff, v1
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv <2 x half> %a, %b %fdiv = fdiv <2 x half> %a, %b
ret <2 x half> %fdiv ret <2 x half> %fdiv
} }
▲ Show 20 Lines • Show All 193 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 16, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 16, v0
; GFX10-NEXT: v_cvt_f32_f16_e32 v7, v0 ; GFX10-NEXT: v_cvt_f32_f16_e32 v7, v0
; GFX10-NEXT: v_cvt_f32_f16_e32 v3, v2 ; GFX10-NEXT: v_cvt_f32_f16_e32 v3, v2
; GFX10-NEXT: v_rcp_f32_e32 v4, v4 ; GFX10-NEXT: v_rcp_f32_e32 v4, v4
; GFX10-NEXT: v_cvt_f32_f16_e32 v6, v5 ; GFX10-NEXT: v_cvt_f32_f16_e32 v6, v5
; GFX10-NEXT: v_rcp_f32_e32 v3, v3 ; GFX10-NEXT: v_rcp_f32_e32 v3, v3
; GFX10-NEXT: v_mul_f32_e32 v4, v7, v4 ; GFX10-NEXT: v_mul_f32_e32 v4, v7, v4
; GFX10-NEXT: v_mul_f32_e32 v3, v6, v3 ; GFX10-NEXT: v_mul_f32_e32 v3, v6, v3
; GFX10-NEXT: v_cvt_f16_f32_e32 v7, v4 ; GFX10-NEXT: v_cvt_f16_f32_e32 v4, v4
; GFX10-NEXT: v_cvt_f16_f32_e32 v3, v3 ; GFX10-NEXT: v_cvt_f16_f32_e32 v3, v3
; GFX10-NEXT: v_div_fixup_f16 v0, v7, v1, v0 ; GFX10-NEXT: v_div_fixup_f16 v0, v4, v1, v0
; GFX10-NEXT: v_div_fixup_f16 v2, v3, v2, v5 ; GFX10-NEXT: v_div_fixup_f16 v2, v3, v2, v5
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v2
; GFX10-NEXT: v_and_or_b32 v0, v0, 0xffff, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, 0xffff, v1
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv <2 x half> %a, %b, !fpmath !0 %fdiv = fdiv <2 x half> %a, %b, !fpmath !0
ret <2 x half> %fdiv ret <2 x half> %fdiv
} }
▲ Show 20 Lines • Show All 647 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 16, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 16, v0
; GFX10-NEXT: v_cvt_f32_f16_e32 v7, v0 ; GFX10-NEXT: v_cvt_f32_f16_e32 v7, v0
; GFX10-NEXT: v_cvt_f32_f16_e32 v3, v2 ; GFX10-NEXT: v_cvt_f32_f16_e32 v3, v2
; GFX10-NEXT: v_rcp_f32_e32 v4, v4 ; GFX10-NEXT: v_rcp_f32_e32 v4, v4
; GFX10-NEXT: v_cvt_f32_f16_e32 v6, v5 ; GFX10-NEXT: v_cvt_f32_f16_e32 v6, v5
; GFX10-NEXT: v_rcp_f32_e32 v3, v3 ; GFX10-NEXT: v_rcp_f32_e32 v3, v3
; GFX10-NEXT: v_mul_f32_e32 v4, v7, v4 ; GFX10-NEXT: v_mul_f32_e32 v4, v7, v4
; GFX10-NEXT: v_mul_f32_e32 v3, v6, v3 ; GFX10-NEXT: v_mul_f32_e32 v3, v6, v3
; GFX10-NEXT: v_cvt_f16_f32_e32 v7, v4 ; GFX10-NEXT: v_cvt_f16_f32_e32 v4, v4
; GFX10-NEXT: v_cvt_f16_f32_e32 v3, v3 ; GFX10-NEXT: v_cvt_f16_f32_e32 v3, v3
; GFX10-NEXT: v_div_fixup_f16 v0, v7, v1, v0 ; GFX10-NEXT: v_div_fixup_f16 v0, v4, v1, v0
; GFX10-NEXT: v_div_fixup_f16 v2, v3, v2, v5 ; GFX10-NEXT: v_div_fixup_f16 v2, v3, v2, v5
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v2
; GFX10-NEXT: v_and_or_b32 v0, v0, 0xffff, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, 0xffff, v1
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv arcp <2 x half> %a, %b, !fpmath !0 %fdiv = fdiv arcp <2 x half> %a, %b, !fpmath !0
ret <2 x half> %fdiv ret <2 x half> %fdiv
} }
▲ Show 20 Lines • Show All 60 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/fdiv.f32.ll

Show First 20 Lines • Show All 706 Lines • ▼ Show 20 Lines
; GFX89-FLUSH-NEXT: s_setpc_b64 s[30:31] ; GFX89-FLUSH-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-IEEE-LABEL: v_fdiv_v2f32: ; GFX10-IEEE-LABEL: v_fdiv_v2f32:
; GFX10-IEEE: ; %bb.0: ; GFX10-IEEE: ; %bb.0:
; GFX10-IEEE-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-IEEE-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-IEEE-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-IEEE-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-IEEE-NEXT: v_div_scale_f32 v4, s4, v2, v2, v0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v4, s4, v2, v2, v0
; GFX10-IEEE-NEXT: v_div_scale_f32 v5, s4, v3, v3, v1 ; GFX10-IEEE-NEXT: v_div_scale_f32 v5, s4, v3, v3, v1
; GFX10-IEEE-NEXT: v_div_scale_f32 v15, vcc_lo, v0, v2, v0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v10, vcc_lo, v0, v2, v0
; GFX10-IEEE-NEXT: v_rcp_f32_e32 v6, v4 ; GFX10-IEEE-NEXT: v_rcp_f32_e32 v6, v4
; GFX10-IEEE-NEXT: v_rcp_f32_e32 v10, v5 ; GFX10-IEEE-NEXT: v_rcp_f32_e32 v7, v5
; GFX10-IEEE-NEXT: v_fma_f32 v8, -v4, v6, 1.0 ; GFX10-IEEE-NEXT: v_fma_f32 v8, -v4, v6, 1.0
; GFX10-IEEE-NEXT: v_fma_f32 v9, -v5, v10, 1.0 ; GFX10-IEEE-NEXT: v_fma_f32 v9, -v5, v7, 1.0
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v6, v8, v6 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v6, v8, v6
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v10, v9, v10 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v7, v9, v7
; GFX10-IEEE-NEXT: v_div_scale_f32 v8, s4, v1, v3, v1 ; GFX10-IEEE-NEXT: v_div_scale_f32 v8, s4, v1, v3, v1
; GFX10-IEEE-NEXT: v_mul_f32_e32 v9, v15, v6 ; GFX10-IEEE-NEXT: v_mul_f32_e32 v9, v10, v6
; GFX10-IEEE-NEXT: v_mul_f32_e32 v11, v8, v10 ; GFX10-IEEE-NEXT: v_mul_f32_e32 v11, v8, v7
; GFX10-IEEE-NEXT: v_fma_f32 v12, v9, -v4, v15 ; GFX10-IEEE-NEXT: v_fma_f32 v12, v9, -v4, v10
; GFX10-IEEE-NEXT: v_fma_f32 v13, v11, -v5, v8 ; GFX10-IEEE-NEXT: v_fma_f32 v13, v11, -v5, v8
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v9, v12, v6 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v9, v12, v6
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v11, v13, v10 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v11, v13, v7
; GFX10-IEEE-NEXT: v_fmac_f32_e64 v15, -v4, v9 ; GFX10-IEEE-NEXT: v_fmac_f32_e64 v10, -v4, v9
; GFX10-IEEE-NEXT: v_fmac_f32_e64 v8, -v5, v11 ; GFX10-IEEE-NEXT: v_fmac_f32_e64 v8, -v5, v11
; GFX10-IEEE-NEXT: v_div_fmas_f32 v7, v15, v6, v9 ; GFX10-IEEE-NEXT: v_div_fmas_f32 v4, v10, v6, v9
; GFX10-IEEE-NEXT: s_mov_b32 vcc_lo, s4 ; GFX10-IEEE-NEXT: s_mov_b32 vcc_lo, s4
; GFX10-IEEE-NEXT: v_div_fmas_f32 v6, v8, v10, v11 ; GFX10-IEEE-NEXT: v_div_fmas_f32 v5, v8, v7, v11
; GFX10-IEEE-NEXT: v_div_fixup_f32 v0, v7, v2, v0 ; GFX10-IEEE-NEXT: v_div_fixup_f32 v0, v4, v2, v0
; GFX10-IEEE-NEXT: v_div_fixup_f32 v1, v6, v3, v1 ; GFX10-IEEE-NEXT: v_div_fixup_f32 v1, v5, v3, v1
; GFX10-IEEE-NEXT: s_setpc_b64 s[30:31] ; GFX10-IEEE-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-FLUSH-LABEL: v_fdiv_v2f32: ; GFX10-FLUSH-LABEL: v_fdiv_v2f32:
; GFX10-FLUSH: ; %bb.0: ; GFX10-FLUSH: ; %bb.0:
; GFX10-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-FLUSH-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-FLUSH-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-FLUSH-NEXT: v_div_scale_f32 v4, s4, v2, v2, v0 ; GFX10-FLUSH-NEXT: v_div_scale_f32 v4, s4, v2, v2, v0
; GFX10-FLUSH-NEXT: v_div_scale_f32 v6, vcc_lo, v0, v2, v0 ; GFX10-FLUSH-NEXT: v_div_scale_f32 v6, vcc_lo, v0, v2, v0
; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v5, v4 ; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v5, v4
; GFX10-FLUSH-NEXT: s_denorm_mode 3 ; GFX10-FLUSH-NEXT: s_denorm_mode 3
; GFX10-FLUSH-NEXT: v_fma_f32 v7, -v4, v5, 1.0 ; GFX10-FLUSH-NEXT: v_fma_f32 v7, -v4, v5, 1.0
; GFX10-FLUSH-NEXT: v_fmac_f32_e32 v5, v7, v5 ; GFX10-FLUSH-NEXT: v_fmac_f32_e32 v5, v7, v5
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v7, v6, v5 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v7, v6, v5
; GFX10-FLUSH-NEXT: v_fma_f32 v8, v7, -v4, v6 ; GFX10-FLUSH-NEXT: v_fma_f32 v8, v7, -v4, v6
; GFX10-FLUSH-NEXT: v_fmac_f32_e32 v7, v8, v5 ; GFX10-FLUSH-NEXT: v_fmac_f32_e32 v7, v8, v5
; GFX10-FLUSH-NEXT: v_fmac_f32_e64 v6, -v4, v7 ; GFX10-FLUSH-NEXT: v_fmac_f32_e64 v6, -v4, v7
; GFX10-FLUSH-NEXT: s_denorm_mode 0 ; GFX10-FLUSH-NEXT: s_denorm_mode 0
; GFX10-FLUSH-NEXT: v_div_scale_f32 v4, s4, v3, v3, v1 ; GFX10-FLUSH-NEXT: v_div_scale_f32 v4, s4, v3, v3, v1
; GFX10-FLUSH-NEXT: v_div_fmas_f32 v5, v6, v5, v7 ; GFX10-FLUSH-NEXT: v_div_fmas_f32 v5, v6, v5, v7
; GFX10-FLUSH-NEXT: v_div_scale_f32 v11, vcc_lo, v1, v3, v1
; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v6, v4 ; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v6, v4
; GFX10-FLUSH-NEXT: v_div_fixup_f32 v0, v5, v2, v0 ; GFX10-FLUSH-NEXT: v_div_fixup_f32 v0, v5, v2, v0
; GFX10-FLUSH-NEXT: v_div_scale_f32 v2, vcc_lo, v1, v3, v1
; GFX10-FLUSH-NEXT: s_denorm_mode 3 ; GFX10-FLUSH-NEXT: s_denorm_mode 3
; GFX10-FLUSH-NEXT: v_fma_f32 v5, -v4, v6, 1.0 ; GFX10-FLUSH-NEXT: v_fma_f32 v5, -v4, v6, 1.0
; GFX10-FLUSH-NEXT: v_fmac_f32_e32 v6, v5, v6 ; GFX10-FLUSH-NEXT: v_fmac_f32_e32 v6, v5, v6
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v5, v11, v6 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v5, v2, v6
; GFX10-FLUSH-NEXT: v_fma_f32 v7, v5, -v4, v11 ; GFX10-FLUSH-NEXT: v_fma_f32 v7, v5, -v4, v2
; GFX10-FLUSH-NEXT: v_fmac_f32_e32 v5, v7, v6 ; GFX10-FLUSH-NEXT: v_fmac_f32_e32 v5, v7, v6
; GFX10-FLUSH-NEXT: v_fmac_f32_e64 v11, -v4, v5 ; GFX10-FLUSH-NEXT: v_fmac_f32_e64 v2, -v4, v5
; GFX10-FLUSH-NEXT: s_denorm_mode 0 ; GFX10-FLUSH-NEXT: s_denorm_mode 0
; GFX10-FLUSH-NEXT: v_div_fmas_f32 v2, v11, v6, v5 ; GFX10-FLUSH-NEXT: v_div_fmas_f32 v2, v2, v6, v5
; GFX10-FLUSH-NEXT: v_div_fixup_f32 v1, v2, v3, v1 ; GFX10-FLUSH-NEXT: v_div_fixup_f32 v1, v2, v3, v1
; GFX10-FLUSH-NEXT: s_setpc_b64 s[30:31] ; GFX10-FLUSH-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv <2 x float> %a, %b %fdiv = fdiv <2 x float> %a, %b
ret <2 x float> %fdiv ret <2 x float> %fdiv
} }
define <2 x float> @v_fdiv_v2f32_afn(<2 x float> %a, <2 x float> %b) { define <2 x float> @v_fdiv_v2f32_afn(<2 x float> %a, <2 x float> %b) {
; GCN-LABEL: v_fdiv_v2f32_afn: ; GCN-LABEL: v_fdiv_v2f32_afn:
▲ Show 20 Lines • Show All 94 Lines • ▼ Show 20 Lines
; GFX89-IEEE-NEXT: s_setpc_b64 s[30:31] ; GFX89-IEEE-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-IEEE-LABEL: v_fdiv_v2f32_ulp25: ; GFX10-IEEE-LABEL: v_fdiv_v2f32_ulp25:
; GFX10-IEEE: ; %bb.0: ; GFX10-IEEE: ; %bb.0:
; GFX10-IEEE-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-IEEE-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-IEEE-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-IEEE-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-IEEE-NEXT: v_div_scale_f32 v4, s4, v2, v2, v0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v4, s4, v2, v2, v0
; GFX10-IEEE-NEXT: v_div_scale_f32 v5, s4, v3, v3, v1 ; GFX10-IEEE-NEXT: v_div_scale_f32 v5, s4, v3, v3, v1
; GFX10-IEEE-NEXT: v_div_scale_f32 v15, vcc_lo, v0, v2, v0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v10, vcc_lo, v0, v2, v0
; GFX10-IEEE-NEXT: v_rcp_f32_e32 v6, v4 ; GFX10-IEEE-NEXT: v_rcp_f32_e32 v6, v4
; GFX10-IEEE-NEXT: v_rcp_f32_e32 v10, v5 ; GFX10-IEEE-NEXT: v_rcp_f32_e32 v7, v5
; GFX10-IEEE-NEXT: v_fma_f32 v8, -v4, v6, 1.0 ; GFX10-IEEE-NEXT: v_fma_f32 v8, -v4, v6, 1.0
; GFX10-IEEE-NEXT: v_fma_f32 v9, -v5, v10, 1.0 ; GFX10-IEEE-NEXT: v_fma_f32 v9, -v5, v7, 1.0
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v6, v8, v6 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v6, v8, v6
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v10, v9, v10 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v7, v9, v7
; GFX10-IEEE-NEXT: v_div_scale_f32 v8, s4, v1, v3, v1 ; GFX10-IEEE-NEXT: v_div_scale_f32 v8, s4, v1, v3, v1
; GFX10-IEEE-NEXT: v_mul_f32_e32 v9, v15, v6 ; GFX10-IEEE-NEXT: v_mul_f32_e32 v9, v10, v6
; GFX10-IEEE-NEXT: v_mul_f32_e32 v11, v8, v10 ; GFX10-IEEE-NEXT: v_mul_f32_e32 v11, v8, v7
; GFX10-IEEE-NEXT: v_fma_f32 v12, v9, -v4, v15 ; GFX10-IEEE-NEXT: v_fma_f32 v12, v9, -v4, v10
; GFX10-IEEE-NEXT: v_fma_f32 v13, v11, -v5, v8 ; GFX10-IEEE-NEXT: v_fma_f32 v13, v11, -v5, v8
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v9, v12, v6 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v9, v12, v6
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v11, v13, v10 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v11, v13, v7
; GFX10-IEEE-NEXT: v_fmac_f32_e64 v15, -v4, v9 ; GFX10-IEEE-NEXT: v_fmac_f32_e64 v10, -v4, v9
; GFX10-IEEE-NEXT: v_fmac_f32_e64 v8, -v5, v11 ; GFX10-IEEE-NEXT: v_fmac_f32_e64 v8, -v5, v11
; GFX10-IEEE-NEXT: v_div_fmas_f32 v7, v15, v6, v9 ; GFX10-IEEE-NEXT: v_div_fmas_f32 v4, v10, v6, v9
; GFX10-IEEE-NEXT: s_mov_b32 vcc_lo, s4 ; GFX10-IEEE-NEXT: s_mov_b32 vcc_lo, s4
; GFX10-IEEE-NEXT: v_div_fmas_f32 v6, v8, v10, v11 ; GFX10-IEEE-NEXT: v_div_fmas_f32 v5, v8, v7, v11
; GFX10-IEEE-NEXT: v_div_fixup_f32 v0, v7, v2, v0 ; GFX10-IEEE-NEXT: v_div_fixup_f32 v0, v4, v2, v0
; GFX10-IEEE-NEXT: v_div_fixup_f32 v1, v6, v3, v1 ; GFX10-IEEE-NEXT: v_div_fixup_f32 v1, v5, v3, v1
; GFX10-IEEE-NEXT: s_setpc_b64 s[30:31] ; GFX10-IEEE-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-FLUSH-LABEL: v_fdiv_v2f32_ulp25: ; GFX10-FLUSH-LABEL: v_fdiv_v2f32_ulp25:
; GFX10-FLUSH: ; %bb.0: ; GFX10-FLUSH: ; %bb.0:
; GFX10-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-FLUSH-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-FLUSH-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-FLUSH-NEXT: s_mov_b32 s4, 0x6f800000 ; GFX10-FLUSH-NEXT: s_mov_b32 s4, 0x6f800000
; GFX10-FLUSH-NEXT: s_mov_b32 s5, 0x2f800000 ; GFX10-FLUSH-NEXT: s_mov_b32 s5, 0x2f800000
; GFX10-FLUSH-NEXT: v_cmp_gt_f32_e64 s6, |v2|, s4 ; GFX10-FLUSH-NEXT: v_cmp_gt_f32_e64 s6, |v2|, s4
; GFX10-FLUSH-NEXT: v_cmp_gt_f32_e64 s4, |v3|, s4 ; GFX10-FLUSH-NEXT: v_cmp_gt_f32_e64 s4, |v3|, s4
; GFX10-FLUSH-NEXT: v_cndmask_b32_e64 v7, 1.0, s5, s6 ; GFX10-FLUSH-NEXT: v_cndmask_b32_e64 v4, 1.0, s5, s6
; GFX10-FLUSH-NEXT: v_cndmask_b32_e64 v6, 1.0, s5, s4 ; GFX10-FLUSH-NEXT: v_cndmask_b32_e64 v5, 1.0, s5, s4
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v2, v2, v7 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v2, v2, v4
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v6 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v5
; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v2, v2 ; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v2, v2
; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v3, v3 ; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v3, v3
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v0, v0, v2 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v0, v0, v2
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v1, v1, v3 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v1, v1, v3
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v0, v7, v0 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v0, v4, v0
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v1, v6, v1 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v1, v5, v1
; GFX10-FLUSH-NEXT: s_setpc_b64 s[30:31] ; GFX10-FLUSH-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv <2 x float> %a, %b, !fpmath !0 %fdiv = fdiv <2 x float> %a, %b, !fpmath !0
ret <2 x float> %fdiv ret <2 x float> %fdiv
} }
define <2 x float> @v_rcp_v2f32(<2 x float> %x) { define <2 x float> @v_rcp_v2f32(<2 x float> %x) {
; GFX6-IEEE-LABEL: v_rcp_v2f32: ; GFX6-IEEE-LABEL: v_rcp_v2f32:
; GFX6-IEEE: ; %bb.0: ; GFX6-IEEE: ; %bb.0:
▲ Show 20 Lines • Show All 113 Lines • ▼ Show 20 Lines
; GFX89-FLUSH-NEXT: s_setpc_b64 s[30:31] ; GFX89-FLUSH-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-IEEE-LABEL: v_rcp_v2f32: ; GFX10-IEEE-LABEL: v_rcp_v2f32:
; GFX10-IEEE: ; %bb.0: ; GFX10-IEEE: ; %bb.0:
; GFX10-IEEE-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-IEEE-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-IEEE-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-IEEE-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-IEEE-NEXT: v_div_scale_f32 v2, s4, v0, v0, 1.0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v2, s4, v0, v0, 1.0
; GFX10-IEEE-NEXT: v_div_scale_f32 v3, s4, v1, v1, 1.0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v3, s4, v1, v1, 1.0
; GFX10-IEEE-NEXT: v_div_scale_f32 v13, vcc_lo, 1.0, v0, 1.0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v8, vcc_lo, 1.0, v0, 1.0
; GFX10-IEEE-NEXT: v_rcp_f32_e32 v4, v2 ; GFX10-IEEE-NEXT: v_rcp_f32_e32 v4, v2
; GFX10-IEEE-NEXT: v_rcp_f32_e32 v8, v3 ; GFX10-IEEE-NEXT: v_rcp_f32_e32 v5, v3
; GFX10-IEEE-NEXT: v_fma_f32 v6, -v2, v4, 1.0 ; GFX10-IEEE-NEXT: v_fma_f32 v6, -v2, v4, 1.0
; GFX10-IEEE-NEXT: v_fma_f32 v7, -v3, v8, 1.0 ; GFX10-IEEE-NEXT: v_fma_f32 v7, -v3, v5, 1.0
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v4, v6, v4 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v4, v6, v4
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v8, v7, v8 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v5, v7, v5
; GFX10-IEEE-NEXT: v_div_scale_f32 v6, s4, 1.0, v1, 1.0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v6, s4, 1.0, v1, 1.0
; GFX10-IEEE-NEXT: v_mul_f32_e32 v7, v13, v4 ; GFX10-IEEE-NEXT: v_mul_f32_e32 v7, v8, v4
; GFX10-IEEE-NEXT: v_mul_f32_e32 v9, v6, v8 ; GFX10-IEEE-NEXT: v_mul_f32_e32 v9, v6, v5
; GFX10-IEEE-NEXT: v_fma_f32 v10, v7, -v2, v13 ; GFX10-IEEE-NEXT: v_fma_f32 v10, v7, -v2, v8
; GFX10-IEEE-NEXT: v_fma_f32 v11, v9, -v3, v6 ; GFX10-IEEE-NEXT: v_fma_f32 v11, v9, -v3, v6
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v7, v10, v4 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v7, v10, v4
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v9, v11, v8 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v9, v11, v5
; GFX10-IEEE-NEXT: v_fmac_f32_e64 v13, -v2, v7 ; GFX10-IEEE-NEXT: v_fmac_f32_e64 v8, -v2, v7
; GFX10-IEEE-NEXT: v_fmac_f32_e64 v6, -v3, v9 ; GFX10-IEEE-NEXT: v_fmac_f32_e64 v6, -v3, v9
; GFX10-IEEE-NEXT: v_div_fmas_f32 v2, v13, v4, v7 ; GFX10-IEEE-NEXT: v_div_fmas_f32 v2, v8, v4, v7
; GFX10-IEEE-NEXT: s_mov_b32 vcc_lo, s4 ; GFX10-IEEE-NEXT: s_mov_b32 vcc_lo, s4
; GFX10-IEEE-NEXT: v_div_fmas_f32 v3, v6, v8, v9 ; GFX10-IEEE-NEXT: v_div_fmas_f32 v3, v6, v5, v9
; GFX10-IEEE-NEXT: v_div_fixup_f32 v0, v2, v0, 1.0 ; GFX10-IEEE-NEXT: v_div_fixup_f32 v0, v2, v0, 1.0
; GFX10-IEEE-NEXT: v_div_fixup_f32 v1, v3, v1, 1.0 ; GFX10-IEEE-NEXT: v_div_fixup_f32 v1, v3, v1, 1.0
; GFX10-IEEE-NEXT: s_setpc_b64 s[30:31] ; GFX10-IEEE-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-FLUSH-LABEL: v_rcp_v2f32: ; GFX10-FLUSH-LABEL: v_rcp_v2f32:
; GFX10-FLUSH: ; %bb.0: ; GFX10-FLUSH: ; %bb.0:
; GFX10-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-FLUSH-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-FLUSH-NEXT: s_waitcnt_vscnt null, 0x0
▲ Show 20 Lines • Show All 147 Lines • ▼ Show 20 Lines
; GFX89-FLUSH-NEXT: s_setpc_b64 s[30:31] ; GFX89-FLUSH-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-IEEE-LABEL: v_rcp_v2f32_arcp: ; GFX10-IEEE-LABEL: v_rcp_v2f32_arcp:
; GFX10-IEEE: ; %bb.0: ; GFX10-IEEE: ; %bb.0:
; GFX10-IEEE-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-IEEE-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-IEEE-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-IEEE-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-IEEE-NEXT: v_div_scale_f32 v2, s4, v0, v0, 1.0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v2, s4, v0, v0, 1.0
; GFX10-IEEE-NEXT: v_div_scale_f32 v3, s4, v1, v1, 1.0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v3, s4, v1, v1, 1.0
; GFX10-IEEE-NEXT: v_div_scale_f32 v13, vcc_lo, 1.0, v0, 1.0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v8, vcc_lo, 1.0, v0, 1.0
; GFX10-IEEE-NEXT: v_rcp_f32_e32 v4, v2 ; GFX10-IEEE-NEXT: v_rcp_f32_e32 v4, v2
; GFX10-IEEE-NEXT: v_rcp_f32_e32 v8, v3 ; GFX10-IEEE-NEXT: v_rcp_f32_e32 v5, v3
; GFX10-IEEE-NEXT: v_fma_f32 v6, -v2, v4, 1.0 ; GFX10-IEEE-NEXT: v_fma_f32 v6, -v2, v4, 1.0
; GFX10-IEEE-NEXT: v_fma_f32 v7, -v3, v8, 1.0 ; GFX10-IEEE-NEXT: v_fma_f32 v7, -v3, v5, 1.0
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v4, v6, v4 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v4, v6, v4
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v8, v7, v8 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v5, v7, v5
; GFX10-IEEE-NEXT: v_div_scale_f32 v6, s4, 1.0, v1, 1.0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v6, s4, 1.0, v1, 1.0
; GFX10-IEEE-NEXT: v_mul_f32_e32 v7, v13, v4 ; GFX10-IEEE-NEXT: v_mul_f32_e32 v7, v8, v4
; GFX10-IEEE-NEXT: v_mul_f32_e32 v9, v6, v8 ; GFX10-IEEE-NEXT: v_mul_f32_e32 v9, v6, v5
; GFX10-IEEE-NEXT: v_fma_f32 v10, v7, -v2, v13 ; GFX10-IEEE-NEXT: v_fma_f32 v10, v7, -v2, v8
; GFX10-IEEE-NEXT: v_fma_f32 v11, v9, -v3, v6 ; GFX10-IEEE-NEXT: v_fma_f32 v11, v9, -v3, v6
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v7, v10, v4 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v7, v10, v4
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v9, v11, v8 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v9, v11, v5
; GFX10-IEEE-NEXT: v_fmac_f32_e64 v13, -v2, v7 ; GFX10-IEEE-NEXT: v_fmac_f32_e64 v8, -v2, v7
; GFX10-IEEE-NEXT: v_fmac_f32_e64 v6, -v3, v9 ; GFX10-IEEE-NEXT: v_fmac_f32_e64 v6, -v3, v9
; GFX10-IEEE-NEXT: v_div_fmas_f32 v2, v13, v4, v7 ; GFX10-IEEE-NEXT: v_div_fmas_f32 v2, v8, v4, v7
; GFX10-IEEE-NEXT: s_mov_b32 vcc_lo, s4 ; GFX10-IEEE-NEXT: s_mov_b32 vcc_lo, s4
; GFX10-IEEE-NEXT: v_div_fmas_f32 v3, v6, v8, v9 ; GFX10-IEEE-NEXT: v_div_fmas_f32 v3, v6, v5, v9
; GFX10-IEEE-NEXT: v_div_fixup_f32 v0, v2, v0, 1.0 ; GFX10-IEEE-NEXT: v_div_fixup_f32 v0, v2, v0, 1.0
; GFX10-IEEE-NEXT: v_div_fixup_f32 v1, v3, v1, 1.0 ; GFX10-IEEE-NEXT: v_div_fixup_f32 v1, v3, v1, 1.0
; GFX10-IEEE-NEXT: s_setpc_b64 s[30:31] ; GFX10-IEEE-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-FLUSH-LABEL: v_rcp_v2f32_arcp: ; GFX10-FLUSH-LABEL: v_rcp_v2f32_arcp:
; GFX10-FLUSH: ; %bb.0: ; GFX10-FLUSH: ; %bb.0:
; GFX10-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-FLUSH-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-FLUSH-NEXT: s_waitcnt_vscnt null, 0x0
▲ Show 20 Lines • Show All 204 Lines • ▼ Show 20 Lines
; GFX89-IEEE-NEXT: s_setpc_b64 s[30:31] ; GFX89-IEEE-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-IEEE-LABEL: v_fdiv_v2f32_arcp_ulp25: ; GFX10-IEEE-LABEL: v_fdiv_v2f32_arcp_ulp25:
; GFX10-IEEE: ; %bb.0: ; GFX10-IEEE: ; %bb.0:
; GFX10-IEEE-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-IEEE-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-IEEE-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-IEEE-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-IEEE-NEXT: v_div_scale_f32 v4, s4, v2, v2, v0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v4, s4, v2, v2, v0
; GFX10-IEEE-NEXT: v_div_scale_f32 v5, s4, v3, v3, v1 ; GFX10-IEEE-NEXT: v_div_scale_f32 v5, s4, v3, v3, v1
; GFX10-IEEE-NEXT: v_div_scale_f32 v15, vcc_lo, v0, v2, v0 ; GFX10-IEEE-NEXT: v_div_scale_f32 v10, vcc_lo, v0, v2, v0
; GFX10-IEEE-NEXT: v_rcp_f32_e32 v6, v4 ; GFX10-IEEE-NEXT: v_rcp_f32_e32 v6, v4
; GFX10-IEEE-NEXT: v_rcp_f32_e32 v10, v5 ; GFX10-IEEE-NEXT: v_rcp_f32_e32 v7, v5
; GFX10-IEEE-NEXT: v_fma_f32 v8, -v4, v6, 1.0 ; GFX10-IEEE-NEXT: v_fma_f32 v8, -v4, v6, 1.0
; GFX10-IEEE-NEXT: v_fma_f32 v9, -v5, v10, 1.0 ; GFX10-IEEE-NEXT: v_fma_f32 v9, -v5, v7, 1.0
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v6, v8, v6 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v6, v8, v6
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v10, v9, v10 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v7, v9, v7
; GFX10-IEEE-NEXT: v_div_scale_f32 v8, s4, v1, v3, v1 ; GFX10-IEEE-NEXT: v_div_scale_f32 v8, s4, v1, v3, v1
; GFX10-IEEE-NEXT: v_mul_f32_e32 v9, v15, v6 ; GFX10-IEEE-NEXT: v_mul_f32_e32 v9, v10, v6
; GFX10-IEEE-NEXT: v_mul_f32_e32 v11, v8, v10 ; GFX10-IEEE-NEXT: v_mul_f32_e32 v11, v8, v7
; GFX10-IEEE-NEXT: v_fma_f32 v12, v9, -v4, v15 ; GFX10-IEEE-NEXT: v_fma_f32 v12, v9, -v4, v10
; GFX10-IEEE-NEXT: v_fma_f32 v13, v11, -v5, v8 ; GFX10-IEEE-NEXT: v_fma_f32 v13, v11, -v5, v8
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v9, v12, v6 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v9, v12, v6
; GFX10-IEEE-NEXT: v_fmac_f32_e32 v11, v13, v10 ; GFX10-IEEE-NEXT: v_fmac_f32_e32 v11, v13, v7
; GFX10-IEEE-NEXT: v_fmac_f32_e64 v15, -v4, v9 ; GFX10-IEEE-NEXT: v_fmac_f32_e64 v10, -v4, v9
; GFX10-IEEE-NEXT: v_fmac_f32_e64 v8, -v5, v11 ; GFX10-IEEE-NEXT: v_fmac_f32_e64 v8, -v5, v11
; GFX10-IEEE-NEXT: v_div_fmas_f32 v7, v15, v6, v9 ; GFX10-IEEE-NEXT: v_div_fmas_f32 v4, v10, v6, v9
; GFX10-IEEE-NEXT: s_mov_b32 vcc_lo, s4 ; GFX10-IEEE-NEXT: s_mov_b32 vcc_lo, s4
; GFX10-IEEE-NEXT: v_div_fmas_f32 v6, v8, v10, v11 ; GFX10-IEEE-NEXT: v_div_fmas_f32 v5, v8, v7, v11
; GFX10-IEEE-NEXT: v_div_fixup_f32 v0, v7, v2, v0 ; GFX10-IEEE-NEXT: v_div_fixup_f32 v0, v4, v2, v0
; GFX10-IEEE-NEXT: v_div_fixup_f32 v1, v6, v3, v1 ; GFX10-IEEE-NEXT: v_div_fixup_f32 v1, v5, v3, v1
; GFX10-IEEE-NEXT: s_setpc_b64 s[30:31] ; GFX10-IEEE-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-FLUSH-LABEL: v_fdiv_v2f32_arcp_ulp25: ; GFX10-FLUSH-LABEL: v_fdiv_v2f32_arcp_ulp25:
; GFX10-FLUSH: ; %bb.0: ; GFX10-FLUSH: ; %bb.0:
; GFX10-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-FLUSH-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-FLUSH-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-FLUSH-NEXT: s_mov_b32 s4, 0x6f800000 ; GFX10-FLUSH-NEXT: s_mov_b32 s4, 0x6f800000
; GFX10-FLUSH-NEXT: s_mov_b32 s5, 0x2f800000 ; GFX10-FLUSH-NEXT: s_mov_b32 s5, 0x2f800000
; GFX10-FLUSH-NEXT: v_cmp_gt_f32_e64 s6, |v2|, s4 ; GFX10-FLUSH-NEXT: v_cmp_gt_f32_e64 s6, |v2|, s4
; GFX10-FLUSH-NEXT: v_cmp_gt_f32_e64 s4, |v3|, s4 ; GFX10-FLUSH-NEXT: v_cmp_gt_f32_e64 s4, |v3|, s4
; GFX10-FLUSH-NEXT: v_cndmask_b32_e64 v7, 1.0, s5, s6 ; GFX10-FLUSH-NEXT: v_cndmask_b32_e64 v4, 1.0, s5, s6
; GFX10-FLUSH-NEXT: v_cndmask_b32_e64 v6, 1.0, s5, s4 ; GFX10-FLUSH-NEXT: v_cndmask_b32_e64 v5, 1.0, s5, s4
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v2, v2, v7 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v2, v2, v4
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v6 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v5
; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v2, v2 ; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v2, v2
; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v3, v3 ; GFX10-FLUSH-NEXT: v_rcp_f32_e32 v3, v3
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v0, v0, v2 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v0, v0, v2
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v1, v1, v3 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v1, v1, v3
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v0, v7, v0 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v0, v4, v0
; GFX10-FLUSH-NEXT: v_mul_f32_e32 v1, v6, v1 ; GFX10-FLUSH-NEXT: v_mul_f32_e32 v1, v5, v1
; GFX10-FLUSH-NEXT: s_setpc_b64 s[30:31] ; GFX10-FLUSH-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv arcp <2 x float> %a, %b, !fpmath !0 %fdiv = fdiv arcp <2 x float> %a, %b, !fpmath !0
ret <2 x float> %fdiv ret <2 x float> %fdiv
} }
define <2 x float> @v_fdiv_v2f32_arcp_afn_ulp25(<2 x float> %a, <2 x float> %b) { define <2 x float> @v_fdiv_v2f32_arcp_afn_ulp25(<2 x float> %a, <2 x float> %b) {
; GCN-LABEL: v_fdiv_v2f32_arcp_afn_ulp25: ; GCN-LABEL: v_fdiv_v2f32_arcp_afn_ulp25:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
Show All 21 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/fdiv.f64.ll

Show First 20 Lines • Show All 99 Lines • ▼ Show 20 Lines
; GFX10-LABEL: v_fdiv_f64_afn: ; GFX10-LABEL: v_fdiv_f64_afn:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_rcp_f64_e32 v[4:5], v[2:3] ; GFX10-NEXT: v_rcp_f64_e32 v[4:5], v[2:3]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0 ; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5] ; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0 ; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0
; GFX10-NEXT: v_fma_f64 v[10:11], v[6:7], v[4:5], v[4:5] ; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5]
; GFX10-NEXT: v_mul_f64 v[6:7], v[0:1], v[10:11] ; GFX10-NEXT: v_mul_f64 v[6:7], v[0:1], v[4:5]
; GFX10-NEXT: v_fma_f64 v[0:1], -v[2:3], v[6:7], v[0:1] ; GFX10-NEXT: v_fma_f64 v[0:1], -v[2:3], v[6:7], v[0:1]
; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[10:11], v[6:7] ; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[6:7]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv afn double %a, %b %fdiv = fdiv afn double %a, %b
ret double %fdiv ret double %fdiv
} }
define double @v_fdiv_f64_ulp25(double %a, double %b) { define double @v_fdiv_f64_ulp25(double %a, double %b) {
; GFX6-LABEL: v_fdiv_f64_ulp25: ; GFX6-LABEL: v_fdiv_f64_ulp25:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 230 Lines • ▼ Show 20 Lines
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_rcp_f64_e32 v[2:3], v[0:1] ; GFX10-NEXT: v_rcp_f64_e32 v[2:3], v[0:1]
; GFX10-NEXT: v_fma_f64 v[4:5], -v[0:1], v[2:3], 1.0 ; GFX10-NEXT: v_fma_f64 v[4:5], -v[0:1], v[2:3], 1.0
; GFX10-NEXT: v_fma_f64 v[2:3], v[4:5], v[2:3], v[2:3] ; GFX10-NEXT: v_fma_f64 v[2:3], v[4:5], v[2:3], v[2:3]
; GFX10-NEXT: v_fma_f64 v[4:5], -v[0:1], v[2:3], 1.0 ; GFX10-NEXT: v_fma_f64 v[4:5], -v[0:1], v[2:3], 1.0
; GFX10-NEXT: v_fma_f64 v[2:3], v[4:5], v[2:3], v[2:3] ; GFX10-NEXT: v_fma_f64 v[2:3], v[4:5], v[2:3], v[2:3]
; GFX10-NEXT: v_mul_f64 v[6:7], 1.0, v[2:3] ; GFX10-NEXT: v_mul_f64 v[4:5], 1.0, v[2:3]
; GFX10-NEXT: v_fma_f64 v[0:1], -v[0:1], v[6:7], 1.0 ; GFX10-NEXT: v_fma_f64 v[0:1], -v[0:1], v[4:5], 1.0
; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], v[6:7] ; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], v[4:5]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv arcp afn double 1.0, %x %fdiv = fdiv arcp afn double 1.0, %x
ret double %fdiv ret double %fdiv
} }
define double @v_rcp_f64_ulp25(double %x) { define double @v_rcp_f64_ulp25(double %x) {
; GFX6-LABEL: v_rcp_f64_ulp25: ; GFX6-LABEL: v_rcp_f64_ulp25:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 84 Lines • ▼ Show 20 Lines
; GFX10-LABEL: v_fdiv_f64_afn_ulp25: ; GFX10-LABEL: v_fdiv_f64_afn_ulp25:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_rcp_f64_e32 v[4:5], v[2:3] ; GFX10-NEXT: v_rcp_f64_e32 v[4:5], v[2:3]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0 ; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5] ; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0 ; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0
; GFX10-NEXT: v_fma_f64 v[10:11], v[6:7], v[4:5], v[4:5] ; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5]
; GFX10-NEXT: v_mul_f64 v[6:7], v[0:1], v[10:11] ; GFX10-NEXT: v_mul_f64 v[6:7], v[0:1], v[4:5]
; GFX10-NEXT: v_fma_f64 v[0:1], -v[2:3], v[6:7], v[0:1] ; GFX10-NEXT: v_fma_f64 v[0:1], -v[2:3], v[6:7], v[0:1]
; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[10:11], v[6:7] ; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[6:7]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv afn double %a, %b, !fpmath !0 %fdiv = fdiv afn double %a, %b, !fpmath !0
ret double %fdiv ret double %fdiv
} }
define double @v_fdiv_f64_arcp_ulp25(double %a, double %b) { define double @v_fdiv_f64_arcp_ulp25(double %a, double %b) {
; GFX6-LABEL: v_fdiv_f64_arcp_ulp25: ; GFX6-LABEL: v_fdiv_f64_arcp_ulp25:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 156 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_div_fmas_f64 v[10:11], v[10:11], v[14:15], v[20:21] ; GFX9-NEXT: v_div_fmas_f64 v[10:11], v[10:11], v[14:15], v[20:21]
; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[10:11], v[6:7], v[2:3] ; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[10:11], v[6:7], v[2:3]
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fdiv_v2f64: ; GFX10-LABEL: v_fdiv_v2f64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v30, v4 ; GFX10-NEXT: v_div_scale_f64 v[8:9], s4, v[4:5], v[4:5], v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v31, v5 ; GFX10-NEXT: v_div_scale_f64 v[10:11], s4, v[6:7], v[6:7], v[2:3]
; GFX10-NEXT: v_mov_b32_e32 v4, v6 ; GFX10-NEXT: v_div_scale_f64 v[20:21], vcc_lo, v[0:1], v[4:5], v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v5, v7 ; GFX10-NEXT: v_rcp_f64_e32 v[12:13], v[8:9]
; GFX10-NEXT: v_div_scale_f64 v[26:27], s4, v[30:31], v[30:31], v[0:1] ; GFX10-NEXT: v_rcp_f64_e32 v[14:15], v[10:11]
; GFX10-NEXT: v_div_scale_f64 v[20:21], vcc_lo, v[0:1], v[30:31], v[0:1] ; GFX10-NEXT: v_fma_f64 v[16:17], -v[8:9], v[12:13], 1.0
; GFX10-NEXT: v_div_scale_f64 v[24:25], s4, v[4:5], v[4:5], v[2:3] ; GFX10-NEXT: v_fma_f64 v[18:19], -v[10:11], v[14:15], 1.0
; GFX10-NEXT: v_div_scale_f64 v[16:17], s4, v[2:3], v[4:5], v[2:3] ; GFX10-NEXT: v_fma_f64 v[12:13], v[12:13], v[16:17], v[12:13]
; GFX10-NEXT: v_rcp_f64_e32 v[12:13], v[26:27] ; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[18:19], v[14:15]
; GFX10-NEXT: v_rcp_f64_e32 v[14:15], v[24:25] ; GFX10-NEXT: v_fma_f64 v[16:17], -v[8:9], v[12:13], 1.0
; GFX10-NEXT: v_fma_f64 v[10:11], -v[26:27], v[12:13], 1.0 ; GFX10-NEXT: v_fma_f64 v[18:19], -v[10:11], v[14:15], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], -v[24:25], v[14:15], 1.0 ; GFX10-NEXT: v_fma_f64 v[12:13], v[12:13], v[16:17], v[12:13]
; GFX10-NEXT: v_fma_f64 v[12:13], v[12:13], v[10:11], v[12:13] ; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[18:19], v[14:15]
; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[8:9], v[14:15] ; GFX10-NEXT: v_div_scale_f64 v[16:17], s4, v[2:3], v[6:7], v[2:3]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[26:27], v[12:13], 1.0 ; GFX10-NEXT: v_mul_f64 v[18:19], v[20:21], v[12:13]
; GFX10-NEXT: v_fma_f64 v[8:9], -v[24:25], v[14:15], 1.0
; GFX10-NEXT: v_fma_f64 v[6:7], v[12:13], v[6:7], v[12:13]
; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[8:9], v[14:15]
; GFX10-NEXT: v_mul_f64 v[18:19], v[20:21], v[6:7]
; GFX10-NEXT: v_mul_f64 v[22:23], v[16:17], v[14:15] ; GFX10-NEXT: v_mul_f64 v[22:23], v[16:17], v[14:15]
; GFX10-NEXT: v_fma_f64 v[8:9], -v[26:27], v[18:19], v[20:21] ; GFX10-NEXT: v_fma_f64 v[8:9], -v[8:9], v[18:19], v[20:21]
; GFX10-NEXT: v_fma_f64 v[16:17], -v[24:25], v[22:23], v[16:17] ; GFX10-NEXT: v_fma_f64 v[10:11], -v[10:11], v[22:23], v[16:17]
; GFX10-NEXT: v_div_fmas_f64 v[18:19], v[8:9], v[6:7], v[18:19] ; GFX10-NEXT: v_div_fmas_f64 v[8:9], v[8:9], v[12:13], v[18:19]
; GFX10-NEXT: s_mov_b32 vcc_lo, s4 ; GFX10-NEXT: s_mov_b32 vcc_lo, s4
; GFX10-NEXT: v_div_fmas_f64 v[8:9], v[16:17], v[14:15], v[22:23] ; GFX10-NEXT: v_div_fmas_f64 v[10:11], v[10:11], v[14:15], v[22:23]
; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[18:19], v[30:31], v[0:1] ; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[8:9], v[4:5], v[0:1]
; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[8:9], v[4:5], v[2:3] ; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[10:11], v[6:7], v[2:3]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv <2 x double> %a, %b %fdiv = fdiv <2 x double> %a, %b
ret <2 x double> %fdiv ret <2 x double> %fdiv
} }
define <2 x double> @v_fdiv_v2f64_afn(<2 x double> %a, <2 x double> %b) { define <2 x double> @v_fdiv_v2f64_afn(<2 x double> %a, <2 x double> %b) {
; GCN-LABEL: v_fdiv_v2f64_afn: ; GCN-LABEL: v_fdiv_v2f64_afn:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
Show All 15 Lines
; GCN-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[12:13] ; GCN-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[12:13]
; GCN-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[14:15] ; GCN-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[14:15]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fdiv_v2f64_afn: ; GFX10-LABEL: v_fdiv_v2f64_afn:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v18, v4 ; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v19, v5 ; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v4, v6 ; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_mov_b32_e32 v5, v7 ; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_mov_b32_e32 v22, v0 ; GFX10-NEXT: v_fma_f64 v[8:9], v[12:13], v[8:9], v[8:9]
; GFX10-NEXT: v_mov_b32_e32 v23, v1 ; GFX10-NEXT: v_fma_f64 v[10:11], v[14:15], v[10:11], v[10:11]
; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[18:19] ; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_mov_b32_e32 v0, v2 ; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[4:5] ; GFX10-NEXT: v_fma_f64 v[8:9], v[12:13], v[8:9], v[8:9]
; GFX10-NEXT: v_mov_b32_e32 v1, v3 ; GFX10-NEXT: v_fma_f64 v[10:11], v[14:15], v[10:11], v[10:11]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[18:19], v[8:9], 1.0 ; GFX10-NEXT: v_mul_f64 v[12:13], v[0:1], v[8:9]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[10:11], 1.0 ; GFX10-NEXT: v_mul_f64 v[14:15], v[2:3], v[10:11]
; GFX10-NEXT: v_fma_f64 v[8:9], v[6:7], v[8:9], v[8:9] ; GFX10-NEXT: v_fma_f64 v[0:1], -v[4:5], v[12:13], v[0:1]
; GFX10-NEXT: v_fma_f64 v[10:11], v[12:13], v[10:11], v[10:11] ; GFX10-NEXT: v_fma_f64 v[2:3], -v[6:7], v[14:15], v[2:3]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[18:19], v[8:9], 1.0 ; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[12:13]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[10:11], 1.0 ; GFX10-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[14:15]
; GFX10-NEXT: v_fma_f64 v[8:9], v[6:7], v[8:9], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[12:13], v[10:11], v[10:11]
; GFX10-NEXT: v_mul_f64 v[12:13], v[22:23], v[8:9]
; GFX10-NEXT: v_mul_f64 v[14:15], v[0:1], v[10:11]
; GFX10-NEXT: v_fma_f64 v[18:19], -v[18:19], v[12:13], v[22:23]
; GFX10-NEXT: v_fma_f64 v[4:5], -v[4:5], v[14:15], v[0:1]
; GFX10-NEXT: v_fma_f64 v[0:1], v[18:19], v[8:9], v[12:13]
; GFX10-NEXT: v_fma_f64 v[2:3], v[4:5], v[10:11], v[14:15]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv afn <2 x double> %a, %b %fdiv = fdiv afn <2 x double> %a, %b
ret <2 x double> %fdiv ret <2 x double> %fdiv
} }
define <2 x double> @v_fdiv_v2f64_ulp25(<2 x double> %a, <2 x double> %b) { define <2 x double> @v_fdiv_v2f64_ulp25(<2 x double> %a, <2 x double> %b) {
; GFX6-LABEL: v_fdiv_v2f64_ulp25: ; GFX6-LABEL: v_fdiv_v2f64_ulp25:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 84 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_div_fmas_f64 v[10:11], v[10:11], v[14:15], v[20:21] ; GFX9-NEXT: v_div_fmas_f64 v[10:11], v[10:11], v[14:15], v[20:21]
; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[10:11], v[6:7], v[2:3] ; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[10:11], v[6:7], v[2:3]
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fdiv_v2f64_ulp25: ; GFX10-LABEL: v_fdiv_v2f64_ulp25:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v30, v4 ; GFX10-NEXT: v_div_scale_f64 v[8:9], s4, v[4:5], v[4:5], v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v31, v5 ; GFX10-NEXT: v_div_scale_f64 v[10:11], s4, v[6:7], v[6:7], v[2:3]
; GFX10-NEXT: v_mov_b32_e32 v4, v6 ; GFX10-NEXT: v_div_scale_f64 v[20:21], vcc_lo, v[0:1], v[4:5], v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v5, v7 ; GFX10-NEXT: v_rcp_f64_e32 v[12:13], v[8:9]
; GFX10-NEXT: v_div_scale_f64 v[26:27], s4, v[30:31], v[30:31], v[0:1] ; GFX10-NEXT: v_rcp_f64_e32 v[14:15], v[10:11]
; GFX10-NEXT: v_div_scale_f64 v[20:21], vcc_lo, v[0:1], v[30:31], v[0:1] ; GFX10-NEXT: v_fma_f64 v[16:17], -v[8:9], v[12:13], 1.0
; GFX10-NEXT: v_div_scale_f64 v[24:25], s4, v[4:5], v[4:5], v[2:3] ; GFX10-NEXT: v_fma_f64 v[18:19], -v[10:11], v[14:15], 1.0
; GFX10-NEXT: v_div_scale_f64 v[16:17], s4, v[2:3], v[4:5], v[2:3] ; GFX10-NEXT: v_fma_f64 v[12:13], v[12:13], v[16:17], v[12:13]
; GFX10-NEXT: v_rcp_f64_e32 v[12:13], v[26:27] ; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[18:19], v[14:15]
; GFX10-NEXT: v_rcp_f64_e32 v[14:15], v[24:25] ; GFX10-NEXT: v_fma_f64 v[16:17], -v[8:9], v[12:13], 1.0
; GFX10-NEXT: v_fma_f64 v[10:11], -v[26:27], v[12:13], 1.0 ; GFX10-NEXT: v_fma_f64 v[18:19], -v[10:11], v[14:15], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], -v[24:25], v[14:15], 1.0 ; GFX10-NEXT: v_fma_f64 v[12:13], v[12:13], v[16:17], v[12:13]
; GFX10-NEXT: v_fma_f64 v[12:13], v[12:13], v[10:11], v[12:13] ; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[18:19], v[14:15]
; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[8:9], v[14:15] ; GFX10-NEXT: v_div_scale_f64 v[16:17], s4, v[2:3], v[6:7], v[2:3]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[26:27], v[12:13], 1.0 ; GFX10-NEXT: v_mul_f64 v[18:19], v[20:21], v[12:13]
; GFX10-NEXT: v_fma_f64 v[8:9], -v[24:25], v[14:15], 1.0
; GFX10-NEXT: v_fma_f64 v[6:7], v[12:13], v[6:7], v[12:13]
; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[8:9], v[14:15]
; GFX10-NEXT: v_mul_f64 v[18:19], v[20:21], v[6:7]
; GFX10-NEXT: v_mul_f64 v[22:23], v[16:17], v[14:15] ; GFX10-NEXT: v_mul_f64 v[22:23], v[16:17], v[14:15]
; GFX10-NEXT: v_fma_f64 v[8:9], -v[26:27], v[18:19], v[20:21] ; GFX10-NEXT: v_fma_f64 v[8:9], -v[8:9], v[18:19], v[20:21]
; GFX10-NEXT: v_fma_f64 v[16:17], -v[24:25], v[22:23], v[16:17] ; GFX10-NEXT: v_fma_f64 v[10:11], -v[10:11], v[22:23], v[16:17]
; GFX10-NEXT: v_div_fmas_f64 v[18:19], v[8:9], v[6:7], v[18:19] ; GFX10-NEXT: v_div_fmas_f64 v[8:9], v[8:9], v[12:13], v[18:19]
; GFX10-NEXT: s_mov_b32 vcc_lo, s4 ; GFX10-NEXT: s_mov_b32 vcc_lo, s4
; GFX10-NEXT: v_div_fmas_f64 v[8:9], v[16:17], v[14:15], v[22:23] ; GFX10-NEXT: v_div_fmas_f64 v[10:11], v[10:11], v[14:15], v[22:23]
; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[18:19], v[30:31], v[0:1] ; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[8:9], v[4:5], v[0:1]
; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[8:9], v[4:5], v[2:3] ; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[10:11], v[6:7], v[2:3]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv <2 x double> %a, %b, !fpmath !0 %fdiv = fdiv <2 x double> %a, %b, !fpmath !0
ret <2 x double> %fdiv ret <2 x double> %fdiv
} }
define <2 x double> @v_rcp_v2f64(<2 x double> %x) { define <2 x double> @v_rcp_v2f64(<2 x double> %x) {
; GFX6-LABEL: v_rcp_v2f64: ; GFX6-LABEL: v_rcp_v2f64:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 84 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_div_fixup_f64 v[0:1], v[4:5], v[0:1], 1.0 ; GFX9-NEXT: v_div_fixup_f64 v[0:1], v[4:5], v[0:1], 1.0
; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[6:7], v[2:3], 1.0 ; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[6:7], v[2:3], 1.0
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_rcp_v2f64: ; GFX10-LABEL: v_rcp_v2f64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_div_scale_f64 v[22:23], s4, v[0:1], v[0:1], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[4:5], s4, v[0:1], v[0:1], 1.0
; GFX10-NEXT: v_div_scale_f64 v[20:21], s4, v[2:3], v[2:3], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[6:7], s4, v[2:3], v[2:3], 1.0
; GFX10-NEXT: v_div_scale_f64 v[16:17], vcc_lo, 1.0, v[0:1], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[16:17], vcc_lo, 1.0, v[0:1], 1.0
; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[4:5]
; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[6:7]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[12:13], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[14:15], v[10:11]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[12:13], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[14:15], v[10:11]
; GFX10-NEXT: v_div_scale_f64 v[12:13], s4, 1.0, v[2:3], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[12:13], s4, 1.0, v[2:3], 1.0
; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[22:23] ; GFX10-NEXT: v_mul_f64 v[14:15], v[16:17], v[8:9]
; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[20:21]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[22:23], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], -v[20:21], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[6:7], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[4:5], v[10:11]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[22:23], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], -v[20:21], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[6:7], v[8:9], v[6:7], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[4:5], v[10:11]
; GFX10-NEXT: v_mul_f64 v[14:15], v[16:17], v[6:7]
; GFX10-NEXT: v_mul_f64 v[18:19], v[12:13], v[10:11] ; GFX10-NEXT: v_mul_f64 v[18:19], v[12:13], v[10:11]
; GFX10-NEXT: v_fma_f64 v[4:5], -v[22:23], v[14:15], v[16:17] ; GFX10-NEXT: v_fma_f64 v[4:5], -v[4:5], v[14:15], v[16:17]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[20:21], v[18:19], v[12:13] ; GFX10-NEXT: v_fma_f64 v[6:7], -v[6:7], v[18:19], v[12:13]
; GFX10-NEXT: v_div_fmas_f64 v[6:7], v[4:5], v[6:7], v[14:15] ; GFX10-NEXT: v_div_fmas_f64 v[4:5], v[4:5], v[8:9], v[14:15]
; GFX10-NEXT: s_mov_b32 vcc_lo, s4 ; GFX10-NEXT: s_mov_b32 vcc_lo, s4
; GFX10-NEXT: v_div_fmas_f64 v[8:9], v[12:13], v[10:11], v[18:19] ; GFX10-NEXT: v_div_fmas_f64 v[6:7], v[6:7], v[10:11], v[18:19]
; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[6:7], v[0:1], 1.0 ; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[4:5], v[0:1], 1.0
; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[8:9], v[2:3], 1.0 ; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[6:7], v[2:3], 1.0
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv <2 x double> <double 1.0, double 1.0>, %x %fdiv = fdiv <2 x double> <double 1.0, double 1.0>, %x
ret <2 x double> %fdiv ret <2 x double> %fdiv
} }
define <2 x double> @v_rcp_v2f64_arcp(<2 x double> %x) { define <2 x double> @v_rcp_v2f64_arcp(<2 x double> %x) {
; GFX6-LABEL: v_rcp_v2f64_arcp: ; GFX6-LABEL: v_rcp_v2f64_arcp:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 84 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_div_fixup_f64 v[0:1], v[4:5], v[0:1], 1.0 ; GFX9-NEXT: v_div_fixup_f64 v[0:1], v[4:5], v[0:1], 1.0
; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[6:7], v[2:3], 1.0 ; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[6:7], v[2:3], 1.0
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_rcp_v2f64_arcp: ; GFX10-LABEL: v_rcp_v2f64_arcp:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_div_scale_f64 v[22:23], s4, v[0:1], v[0:1], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[4:5], s4, v[0:1], v[0:1], 1.0
; GFX10-NEXT: v_div_scale_f64 v[20:21], s4, v[2:3], v[2:3], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[6:7], s4, v[2:3], v[2:3], 1.0
; GFX10-NEXT: v_div_scale_f64 v[16:17], vcc_lo, 1.0, v[0:1], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[16:17], vcc_lo, 1.0, v[0:1], 1.0
; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[4:5]
; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[6:7]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[12:13], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[14:15], v[10:11]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[12:13], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[14:15], v[10:11]
; GFX10-NEXT: v_div_scale_f64 v[12:13], s4, 1.0, v[2:3], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[12:13], s4, 1.0, v[2:3], 1.0
; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[22:23] ; GFX10-NEXT: v_mul_f64 v[14:15], v[16:17], v[8:9]
; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[20:21]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[22:23], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], -v[20:21], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[6:7], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[4:5], v[10:11]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[22:23], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], -v[20:21], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[6:7], v[8:9], v[6:7], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[4:5], v[10:11]
; GFX10-NEXT: v_mul_f64 v[14:15], v[16:17], v[6:7]
; GFX10-NEXT: v_mul_f64 v[18:19], v[12:13], v[10:11] ; GFX10-NEXT: v_mul_f64 v[18:19], v[12:13], v[10:11]
; GFX10-NEXT: v_fma_f64 v[4:5], -v[22:23], v[14:15], v[16:17] ; GFX10-NEXT: v_fma_f64 v[4:5], -v[4:5], v[14:15], v[16:17]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[20:21], v[18:19], v[12:13] ; GFX10-NEXT: v_fma_f64 v[6:7], -v[6:7], v[18:19], v[12:13]
; GFX10-NEXT: v_div_fmas_f64 v[6:7], v[4:5], v[6:7], v[14:15] ; GFX10-NEXT: v_div_fmas_f64 v[4:5], v[4:5], v[8:9], v[14:15]
; GFX10-NEXT: s_mov_b32 vcc_lo, s4 ; GFX10-NEXT: s_mov_b32 vcc_lo, s4
; GFX10-NEXT: v_div_fmas_f64 v[8:9], v[12:13], v[10:11], v[18:19] ; GFX10-NEXT: v_div_fmas_f64 v[6:7], v[6:7], v[10:11], v[18:19]
; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[6:7], v[0:1], 1.0 ; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[4:5], v[0:1], 1.0
; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[8:9], v[2:3], 1.0 ; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[6:7], v[2:3], 1.0
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv arcp <2 x double> <double 1.0, double 1.0>, %x %fdiv = fdiv arcp <2 x double> <double 1.0, double 1.0>, %x
ret <2 x double> %fdiv ret <2 x double> %fdiv
} }
define <2 x double> @v_rcp_v2f64_arcp_afn(<2 x double> %x) { define <2 x double> @v_rcp_v2f64_arcp_afn(<2 x double> %x) {
; GCN-LABEL: v_rcp_v2f64_arcp_afn: ; GCN-LABEL: v_rcp_v2f64_arcp_afn:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
Show All 15 Lines
; GCN-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9] ; GCN-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9]
; GCN-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11] ; GCN-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_rcp_v2f64_arcp_afn: ; GFX10-LABEL: v_rcp_v2f64_arcp_afn:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v14, v0 ; GFX10-NEXT: v_rcp_f64_e32 v[4:5], v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v15, v1 ; GFX10-NEXT: v_rcp_f64_e32 v[6:7], v[2:3]
; GFX10-NEXT: v_mov_b32_e32 v0, v2 ; GFX10-NEXT: v_fma_f64 v[8:9], -v[0:1], v[4:5], 1.0
; GFX10-NEXT: v_mov_b32_e32 v1, v3 ; GFX10-NEXT: v_fma_f64 v[10:11], -v[2:3], v[6:7], 1.0
; GFX10-NEXT: v_rcp_f64_e32 v[4:5], v[14:15] ; GFX10-NEXT: v_fma_f64 v[4:5], v[8:9], v[4:5], v[4:5]
; GFX10-NEXT: v_rcp_f64_e32 v[6:7], v[0:1] ; GFX10-NEXT: v_fma_f64 v[6:7], v[10:11], v[6:7], v[6:7]
; GFX10-NEXT: v_fma_f64 v[2:3], -v[14:15], v[4:5], 1.0 ; GFX10-NEXT: v_fma_f64 v[8:9], -v[0:1], v[4:5], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], -v[0:1], v[6:7], 1.0 ; GFX10-NEXT: v_fma_f64 v[10:11], -v[2:3], v[6:7], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], v[2:3], v[4:5], v[4:5] ; GFX10-NEXT: v_fma_f64 v[4:5], v[8:9], v[4:5], v[4:5]
; GFX10-NEXT: v_fma_f64 v[6:7], v[8:9], v[6:7], v[6:7] ; GFX10-NEXT: v_fma_f64 v[6:7], v[10:11], v[6:7], v[6:7]
; GFX10-NEXT: v_fma_f64 v[2:3], -v[14:15], v[4:5], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], -v[0:1], v[6:7], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], v[2:3], v[4:5], v[4:5]
; GFX10-NEXT: v_fma_f64 v[6:7], v[8:9], v[6:7], v[6:7]
; GFX10-NEXT: v_mul_f64 v[8:9], 1.0, v[4:5] ; GFX10-NEXT: v_mul_f64 v[8:9], 1.0, v[4:5]
; GFX10-NEXT: v_mul_f64 v[10:11], 1.0, v[6:7] ; GFX10-NEXT: v_mul_f64 v[10:11], 1.0, v[6:7]
; GFX10-NEXT: v_fma_f64 v[14:15], -v[14:15], v[8:9], 1.0 ; GFX10-NEXT: v_fma_f64 v[0:1], -v[0:1], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[12:13], -v[0:1], v[10:11], 1.0 ; GFX10-NEXT: v_fma_f64 v[2:3], -v[2:3], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[0:1], v[14:15], v[4:5], v[8:9] ; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9]
; GFX10-NEXT: v_fma_f64 v[2:3], v[12:13], v[6:7], v[10:11] ; GFX10-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv arcp afn <2 x double> <double 1.0, double 1.0>, %x %fdiv = fdiv arcp afn <2 x double> <double 1.0, double 1.0>, %x
ret <2 x double> %fdiv ret <2 x double> %fdiv
} }
define <2 x double> @v_rcp_v2f64_ulp25(<2 x double> %x) { define <2 x double> @v_rcp_v2f64_ulp25(<2 x double> %x) {
; GFX6-LABEL: v_rcp_v2f64_ulp25: ; GFX6-LABEL: v_rcp_v2f64_ulp25:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 84 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_div_fixup_f64 v[0:1], v[4:5], v[0:1], 1.0 ; GFX9-NEXT: v_div_fixup_f64 v[0:1], v[4:5], v[0:1], 1.0
; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[6:7], v[2:3], 1.0 ; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[6:7], v[2:3], 1.0
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_rcp_v2f64_ulp25: ; GFX10-LABEL: v_rcp_v2f64_ulp25:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_div_scale_f64 v[22:23], s4, v[0:1], v[0:1], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[4:5], s4, v[0:1], v[0:1], 1.0
; GFX10-NEXT: v_div_scale_f64 v[20:21], s4, v[2:3], v[2:3], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[6:7], s4, v[2:3], v[2:3], 1.0
; GFX10-NEXT: v_div_scale_f64 v[16:17], vcc_lo, 1.0, v[0:1], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[16:17], vcc_lo, 1.0, v[0:1], 1.0
; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[4:5]
; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[6:7]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[12:13], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[14:15], v[10:11]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[12:13], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[14:15], v[10:11]
; GFX10-NEXT: v_div_scale_f64 v[12:13], s4, 1.0, v[2:3], 1.0 ; GFX10-NEXT: v_div_scale_f64 v[12:13], s4, 1.0, v[2:3], 1.0
; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[22:23] ; GFX10-NEXT: v_mul_f64 v[14:15], v[16:17], v[8:9]
; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[20:21]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[22:23], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], -v[20:21], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[6:7], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[4:5], v[10:11]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[22:23], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], -v[20:21], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[6:7], v[8:9], v[6:7], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[4:5], v[10:11]
; GFX10-NEXT: v_mul_f64 v[14:15], v[16:17], v[6:7]
; GFX10-NEXT: v_mul_f64 v[18:19], v[12:13], v[10:11] ; GFX10-NEXT: v_mul_f64 v[18:19], v[12:13], v[10:11]
; GFX10-NEXT: v_fma_f64 v[4:5], -v[22:23], v[14:15], v[16:17] ; GFX10-NEXT: v_fma_f64 v[4:5], -v[4:5], v[14:15], v[16:17]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[20:21], v[18:19], v[12:13] ; GFX10-NEXT: v_fma_f64 v[6:7], -v[6:7], v[18:19], v[12:13]
; GFX10-NEXT: v_div_fmas_f64 v[6:7], v[4:5], v[6:7], v[14:15] ; GFX10-NEXT: v_div_fmas_f64 v[4:5], v[4:5], v[8:9], v[14:15]
; GFX10-NEXT: s_mov_b32 vcc_lo, s4 ; GFX10-NEXT: s_mov_b32 vcc_lo, s4
; GFX10-NEXT: v_div_fmas_f64 v[8:9], v[12:13], v[10:11], v[18:19] ; GFX10-NEXT: v_div_fmas_f64 v[6:7], v[6:7], v[10:11], v[18:19]
; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[6:7], v[0:1], 1.0 ; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[4:5], v[0:1], 1.0
; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[8:9], v[2:3], 1.0 ; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[6:7], v[2:3], 1.0
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv <2 x double> <double 1.0, double 1.0>, %x, !fpmath !0 %fdiv = fdiv <2 x double> <double 1.0, double 1.0>, %x, !fpmath !0
ret <2 x double> %fdiv ret <2 x double> %fdiv
} }
define <2 x double> @v_fdiv_v2f64_afn_ulp25(<2 x double> %a, <2 x double> %b) { define <2 x double> @v_fdiv_v2f64_afn_ulp25(<2 x double> %a, <2 x double> %b) {
; GCN-LABEL: v_fdiv_v2f64_afn_ulp25: ; GCN-LABEL: v_fdiv_v2f64_afn_ulp25:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
Show All 15 Lines
; GCN-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[12:13] ; GCN-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[12:13]
; GCN-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[14:15] ; GCN-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[14:15]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fdiv_v2f64_afn_ulp25: ; GFX10-LABEL: v_fdiv_v2f64_afn_ulp25:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v18, v4 ; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v19, v5 ; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v4, v6 ; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_mov_b32_e32 v5, v7 ; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_mov_b32_e32 v22, v0 ; GFX10-NEXT: v_fma_f64 v[8:9], v[12:13], v[8:9], v[8:9]
; GFX10-NEXT: v_mov_b32_e32 v23, v1 ; GFX10-NEXT: v_fma_f64 v[10:11], v[14:15], v[10:11], v[10:11]
; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[18:19] ; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_mov_b32_e32 v0, v2 ; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[4:5] ; GFX10-NEXT: v_fma_f64 v[8:9], v[12:13], v[8:9], v[8:9]
; GFX10-NEXT: v_mov_b32_e32 v1, v3 ; GFX10-NEXT: v_fma_f64 v[10:11], v[14:15], v[10:11], v[10:11]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[18:19], v[8:9], 1.0 ; GFX10-NEXT: v_mul_f64 v[12:13], v[0:1], v[8:9]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[10:11], 1.0 ; GFX10-NEXT: v_mul_f64 v[14:15], v[2:3], v[10:11]
; GFX10-NEXT: v_fma_f64 v[8:9], v[6:7], v[8:9], v[8:9] ; GFX10-NEXT: v_fma_f64 v[0:1], -v[4:5], v[12:13], v[0:1]
; GFX10-NEXT: v_fma_f64 v[10:11], v[12:13], v[10:11], v[10:11] ; GFX10-NEXT: v_fma_f64 v[2:3], -v[6:7], v[14:15], v[2:3]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[18:19], v[8:9], 1.0 ; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[12:13]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[10:11], 1.0 ; GFX10-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[14:15]
; GFX10-NEXT: v_fma_f64 v[8:9], v[6:7], v[8:9], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[12:13], v[10:11], v[10:11]
; GFX10-NEXT: v_mul_f64 v[12:13], v[22:23], v[8:9]
; GFX10-NEXT: v_mul_f64 v[14:15], v[0:1], v[10:11]
; GFX10-NEXT: v_fma_f64 v[18:19], -v[18:19], v[12:13], v[22:23]
; GFX10-NEXT: v_fma_f64 v[4:5], -v[4:5], v[14:15], v[0:1]
; GFX10-NEXT: v_fma_f64 v[0:1], v[18:19], v[8:9], v[12:13]
; GFX10-NEXT: v_fma_f64 v[2:3], v[4:5], v[10:11], v[14:15]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv afn <2 x double> %a, %b, !fpmath !0 %fdiv = fdiv afn <2 x double> %a, %b, !fpmath !0
ret <2 x double> %fdiv ret <2 x double> %fdiv
} }
define <2 x double> @v_fdiv_v2f64_arcp_ulp25(<2 x double> %a, <2 x double> %b) { define <2 x double> @v_fdiv_v2f64_arcp_ulp25(<2 x double> %a, <2 x double> %b) {
; GFX6-LABEL: v_fdiv_v2f64_arcp_ulp25: ; GFX6-LABEL: v_fdiv_v2f64_arcp_ulp25:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 84 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_div_fmas_f64 v[10:11], v[10:11], v[14:15], v[20:21] ; GFX9-NEXT: v_div_fmas_f64 v[10:11], v[10:11], v[14:15], v[20:21]
; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[10:11], v[6:7], v[2:3] ; GFX9-NEXT: v_div_fixup_f64 v[2:3], v[10:11], v[6:7], v[2:3]
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fdiv_v2f64_arcp_ulp25: ; GFX10-LABEL: v_fdiv_v2f64_arcp_ulp25:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v30, v4 ; GFX10-NEXT: v_div_scale_f64 v[8:9], s4, v[4:5], v[4:5], v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v31, v5 ; GFX10-NEXT: v_div_scale_f64 v[10:11], s4, v[6:7], v[6:7], v[2:3]
; GFX10-NEXT: v_mov_b32_e32 v4, v6 ; GFX10-NEXT: v_div_scale_f64 v[20:21], vcc_lo, v[0:1], v[4:5], v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v5, v7 ; GFX10-NEXT: v_rcp_f64_e32 v[12:13], v[8:9]
; GFX10-NEXT: v_div_scale_f64 v[26:27], s4, v[30:31], v[30:31], v[0:1] ; GFX10-NEXT: v_rcp_f64_e32 v[14:15], v[10:11]
; GFX10-NEXT: v_div_scale_f64 v[20:21], vcc_lo, v[0:1], v[30:31], v[0:1] ; GFX10-NEXT: v_fma_f64 v[16:17], -v[8:9], v[12:13], 1.0
; GFX10-NEXT: v_div_scale_f64 v[24:25], s4, v[4:5], v[4:5], v[2:3] ; GFX10-NEXT: v_fma_f64 v[18:19], -v[10:11], v[14:15], 1.0
; GFX10-NEXT: v_div_scale_f64 v[16:17], s4, v[2:3], v[4:5], v[2:3] ; GFX10-NEXT: v_fma_f64 v[12:13], v[12:13], v[16:17], v[12:13]
; GFX10-NEXT: v_rcp_f64_e32 v[12:13], v[26:27] ; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[18:19], v[14:15]
; GFX10-NEXT: v_rcp_f64_e32 v[14:15], v[24:25] ; GFX10-NEXT: v_fma_f64 v[16:17], -v[8:9], v[12:13], 1.0
; GFX10-NEXT: v_fma_f64 v[10:11], -v[26:27], v[12:13], 1.0 ; GFX10-NEXT: v_fma_f64 v[18:19], -v[10:11], v[14:15], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], -v[24:25], v[14:15], 1.0 ; GFX10-NEXT: v_fma_f64 v[12:13], v[12:13], v[16:17], v[12:13]
; GFX10-NEXT: v_fma_f64 v[12:13], v[12:13], v[10:11], v[12:13] ; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[18:19], v[14:15]
; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[8:9], v[14:15] ; GFX10-NEXT: v_div_scale_f64 v[16:17], s4, v[2:3], v[6:7], v[2:3]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[26:27], v[12:13], 1.0 ; GFX10-NEXT: v_mul_f64 v[18:19], v[20:21], v[12:13]
; GFX10-NEXT: v_fma_f64 v[8:9], -v[24:25], v[14:15], 1.0
; GFX10-NEXT: v_fma_f64 v[6:7], v[12:13], v[6:7], v[12:13]
; GFX10-NEXT: v_fma_f64 v[14:15], v[14:15], v[8:9], v[14:15]
; GFX10-NEXT: v_mul_f64 v[18:19], v[20:21], v[6:7]
; GFX10-NEXT: v_mul_f64 v[22:23], v[16:17], v[14:15] ; GFX10-NEXT: v_mul_f64 v[22:23], v[16:17], v[14:15]
; GFX10-NEXT: v_fma_f64 v[8:9], -v[26:27], v[18:19], v[20:21] ; GFX10-NEXT: v_fma_f64 v[8:9], -v[8:9], v[18:19], v[20:21]
; GFX10-NEXT: v_fma_f64 v[16:17], -v[24:25], v[22:23], v[16:17] ; GFX10-NEXT: v_fma_f64 v[10:11], -v[10:11], v[22:23], v[16:17]
; GFX10-NEXT: v_div_fmas_f64 v[18:19], v[8:9], v[6:7], v[18:19] ; GFX10-NEXT: v_div_fmas_f64 v[8:9], v[8:9], v[12:13], v[18:19]
; GFX10-NEXT: s_mov_b32 vcc_lo, s4 ; GFX10-NEXT: s_mov_b32 vcc_lo, s4
; GFX10-NEXT: v_div_fmas_f64 v[8:9], v[16:17], v[14:15], v[22:23] ; GFX10-NEXT: v_div_fmas_f64 v[10:11], v[10:11], v[14:15], v[22:23]
; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[18:19], v[30:31], v[0:1] ; GFX10-NEXT: v_div_fixup_f64 v[0:1], v[8:9], v[4:5], v[0:1]
; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[8:9], v[4:5], v[2:3] ; GFX10-NEXT: v_div_fixup_f64 v[2:3], v[10:11], v[6:7], v[2:3]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv arcp <2 x double> %a, %b, !fpmath !0 %fdiv = fdiv arcp <2 x double> %a, %b, !fpmath !0
ret <2 x double> %fdiv ret <2 x double> %fdiv
} }
define <2 x double> @v_fdiv_v2f64_arcp_afn_ulp25(<2 x double> %a, <2 x double> %b) { define <2 x double> @v_fdiv_v2f64_arcp_afn_ulp25(<2 x double> %a, <2 x double> %b) {
; GCN-LABEL: v_fdiv_v2f64_arcp_afn_ulp25: ; GCN-LABEL: v_fdiv_v2f64_arcp_afn_ulp25:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
Show All 15 Lines
; GCN-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[12:13] ; GCN-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[12:13]
; GCN-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[14:15] ; GCN-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[14:15]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fdiv_v2f64_arcp_afn_ulp25: ; GFX10-LABEL: v_fdiv_v2f64_arcp_afn_ulp25:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v18, v4 ; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v19, v5 ; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v4, v6 ; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_mov_b32_e32 v5, v7 ; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_mov_b32_e32 v22, v0 ; GFX10-NEXT: v_fma_f64 v[8:9], v[12:13], v[8:9], v[8:9]
; GFX10-NEXT: v_mov_b32_e32 v23, v1 ; GFX10-NEXT: v_fma_f64 v[10:11], v[14:15], v[10:11], v[10:11]
; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[18:19] ; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[8:9], 1.0
; GFX10-NEXT: v_mov_b32_e32 v0, v2 ; GFX10-NEXT: v_fma_f64 v[14:15], -v[6:7], v[10:11], 1.0
; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[4:5] ; GFX10-NEXT: v_fma_f64 v[8:9], v[12:13], v[8:9], v[8:9]
; GFX10-NEXT: v_mov_b32_e32 v1, v3 ; GFX10-NEXT: v_fma_f64 v[10:11], v[14:15], v[10:11], v[10:11]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[18:19], v[8:9], 1.0 ; GFX10-NEXT: v_mul_f64 v[12:13], v[0:1], v[8:9]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[10:11], 1.0 ; GFX10-NEXT: v_mul_f64 v[14:15], v[2:3], v[10:11]
; GFX10-NEXT: v_fma_f64 v[8:9], v[6:7], v[8:9], v[8:9] ; GFX10-NEXT: v_fma_f64 v[0:1], -v[4:5], v[12:13], v[0:1]
; GFX10-NEXT: v_fma_f64 v[10:11], v[12:13], v[10:11], v[10:11] ; GFX10-NEXT: v_fma_f64 v[2:3], -v[6:7], v[14:15], v[2:3]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[18:19], v[8:9], 1.0 ; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[12:13]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[4:5], v[10:11], 1.0 ; GFX10-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[14:15]
; GFX10-NEXT: v_fma_f64 v[8:9], v[6:7], v[8:9], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], v[12:13], v[10:11], v[10:11]
; GFX10-NEXT: v_mul_f64 v[12:13], v[22:23], v[8:9]
; GFX10-NEXT: v_mul_f64 v[14:15], v[0:1], v[10:11]
; GFX10-NEXT: v_fma_f64 v[18:19], -v[18:19], v[12:13], v[22:23]
; GFX10-NEXT: v_fma_f64 v[4:5], -v[4:5], v[14:15], v[0:1]
; GFX10-NEXT: v_fma_f64 v[0:1], v[18:19], v[8:9], v[12:13]
; GFX10-NEXT: v_fma_f64 v[2:3], v[4:5], v[10:11], v[14:15]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fdiv = fdiv afn arcp <2 x double> %a, %b, !fpmath !0 %fdiv = fdiv afn arcp <2 x double> %a, %b, !fpmath !0
ret <2 x double> %fdiv ret <2 x double> %fdiv
} }
!0 = !{float 2.500000e+00} !0 = !{float 2.500000e+00}

llvm/test/CodeGen/AMDGPU/GlobalISel/fma.ll

Show First 20 Lines • Show All 473 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9] ; GFX9-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9]
; GFX9-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11] ; GFX9-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11]
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fma_v2f64: ; GFX10-LABEL: v_fma_v2f64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v14, v0 ; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9]
; GFX10-NEXT: v_mov_b32_e32 v15, v1 ; GFX10-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11]
; GFX10-NEXT: v_mov_b32_e32 v12, v2
; GFX10-NEXT: v_mov_b32_e32 v13, v3
; GFX10-NEXT: v_fma_f64 v[0:1], v[14:15], v[4:5], v[8:9]
; GFX10-NEXT: v_fma_f64 v[2:3], v[12:13], v[6:7], v[10:11]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%fma = call <2 x double> @llvm.fma.v2f64(<2 x double> %x, <2 x double> %y, <2 x double> %z) %fma = call <2 x double> @llvm.fma.v2f64(<2 x double> %x, <2 x double> %y, <2 x double> %z)
ret <2 x double> %fma ret <2 x double> %fma
} }
define float @v_fma_f32_fabs_lhs(float %x, float %y, float %z) { define float @v_fma_f32_fabs_lhs(float %x, float %y, float %z) {
; GFX6-LABEL: v_fma_f32_fabs_lhs: ; GFX6-LABEL: v_fma_f32_fabs_lhs:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 272 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/fshl.ll

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 1,153 Lines • ▼ Show 20 Lines
; GFX10-LABEL: v_fshl_v4i8: ; GFX10-LABEL: v_fshl_v4i8:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v2
; GFX10-NEXT: v_and_b32_e32 v11, 7, v2 ; GFX10-NEXT: v_and_b32_e32 v11, 7, v2
; GFX10-NEXT: v_xor_b32_e32 v10, -1, v2 ; GFX10-NEXT: v_xor_b32_e32 v10, -1, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 16, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v9, 16, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v2, 24, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 16, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 16, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v0
; GFX10-NEXT: v_lshlrev_b16 v0, v11, v0 ; GFX10-NEXT: v_lshlrev_b16 v0, v11, v0
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v8 ; GFX10-NEXT: v_xor_b32_e32 v11, -1, v8
; GFX10-NEXT: v_and_b32_e32 v8, 7, v8 ; GFX10-NEXT: v_and_b32_e32 v8, 7, v8
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 8, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 8, v1
; GFX10-NEXT: v_mov_b32_e32 v15, 0xff ; GFX10-NEXT: v_lshrrev_b32_e32 v2, 24, v2
; GFX10-NEXT: v_mov_b32_e32 v13, 0xff
; GFX10-NEXT: s_movk_i32 s4, 0xff ; GFX10-NEXT: s_movk_i32 s4, 0xff
; GFX10-NEXT: v_xor_b32_e32 v13, -1, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1
; GFX10-NEXT: v_and_b32_e32 v12, s4, v1 ; GFX10-NEXT: v_and_b32_e32 v12, s4, v1
; GFX10-NEXT: v_and_b32_e32 v6, s4, v6 ; GFX10-NEXT: v_and_b32_e32 v6, s4, v6
; GFX10-NEXT: v_and_b32_sdwa v1, v1, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v1, v1, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_xor_b32_e32 v13, -1, v2
; GFX10-NEXT: v_lshlrev_b16 v3, v8, v3 ; GFX10-NEXT: v_lshlrev_b16 v3, v8, v3
; GFX10-NEXT: v_xor_b32_e32 v8, -1, v9 ; GFX10-NEXT: v_xor_b32_e32 v8, -1, v9
; GFX10-NEXT: v_and_b32_e32 v2, 7, v2 ; GFX10-NEXT: v_and_b32_e32 v2, 7, v2
; GFX10-NEXT: v_and_b32_e32 v9, 7, v9 ; GFX10-NEXT: v_and_b32_e32 v9, 7, v9
; GFX10-NEXT: v_and_b32_e32 v11, 7, v11 ; GFX10-NEXT: v_and_b32_e32 v11, 7, v11
; GFX10-NEXT: v_lshrrev_b16 v6, 1, v6 ; GFX10-NEXT: v_lshrrev_b16 v6, 1, v6
; GFX10-NEXT: v_and_b32_e32 v8, 7, v8 ; GFX10-NEXT: v_and_b32_e32 v8, 7, v8
; GFX10-NEXT: v_lshrrev_b16 v1, 1, v1 ; GFX10-NEXT: v_lshrrev_b16 v1, 1, v1
▲ Show 20 Lines • Show All 1,000 Lines • ▼ Show 20 Lines
; ;
; GFX10-LABEL: v_fshl_v2i24: ; GFX10-LABEL: v_fshl_v2i24:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v6, 24 ; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v6, 24
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v7, 24 ; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v7, 24
; GFX10-NEXT: s_sub_i32 s4, 0, 24 ; GFX10-NEXT: s_sub_i32 s4, 0, 24
; GFX10-NEXT: v_mov_b32_e32 v12, 0xffffff ; GFX10-NEXT: v_mov_b32_e32 v10, 0xffffff
; GFX10-NEXT: v_and_b32_e32 v4, 0xffffff, v4 ; GFX10-NEXT: v_and_b32_e32 v4, 0xffffff, v4
; GFX10-NEXT: v_rcp_iflag_f32_e32 v6, v6 ; GFX10-NEXT: v_rcp_iflag_f32_e32 v6, v6
; GFX10-NEXT: v_rcp_iflag_f32_e32 v7, v7 ; GFX10-NEXT: v_rcp_iflag_f32_e32 v7, v7
; GFX10-NEXT: v_and_b32_e32 v5, v5, v12 ; GFX10-NEXT: v_and_b32_e32 v5, v5, v10
; GFX10-NEXT: v_and_b32_e32 v2, v2, v12 ; GFX10-NEXT: v_and_b32_e32 v2, v2, v10
; GFX10-NEXT: v_and_b32_e32 v3, v3, v12 ; GFX10-NEXT: v_and_b32_e32 v3, v3, v10
; GFX10-NEXT: v_lshrrev_b32_e32 v2, 1, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v2, 1, v2
; GFX10-NEXT: v_mul_f32_e32 v6, 0x4f7ffffe, v6 ; GFX10-NEXT: v_mul_f32_e32 v6, 0x4f7ffffe, v6
; GFX10-NEXT: v_mul_f32_e32 v7, 0x4f7ffffe, v7 ; GFX10-NEXT: v_mul_f32_e32 v7, 0x4f7ffffe, v7
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 1, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 1, v3
; GFX10-NEXT: v_cvt_u32_f32_e32 v6, v6 ; GFX10-NEXT: v_cvt_u32_f32_e32 v6, v6
; GFX10-NEXT: v_cvt_u32_f32_e32 v7, v7 ; GFX10-NEXT: v_cvt_u32_f32_e32 v7, v7
; GFX10-NEXT: v_mul_lo_u32 v8, s4, v6 ; GFX10-NEXT: v_mul_lo_u32 v8, s4, v6
; GFX10-NEXT: v_mul_lo_u32 v9, s4, v7 ; GFX10-NEXT: v_mul_lo_u32 v9, s4, v7
Show All 11 Lines
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v4 ; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v4
; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 24, v5 ; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 24, v5
; GFX10-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v5 ; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v5
; GFX10-NEXT: v_subrev_nc_u32_e32 v6, 24, v4 ; GFX10-NEXT: v_subrev_nc_u32_e32 v6, 24, v4
; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v4 ; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v4
; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 24, v5 ; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 24, v5
; GFX10-NEXT: v_cndmask_b32_e32 v15, v4, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v5 ; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v5
; GFX10-NEXT: v_sub_nc_u32_e32 v6, 23, v15 ; GFX10-NEXT: v_sub_nc_u32_e32 v6, 23, v4
; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc_lo
; GFX10-NEXT: v_and_b32_e32 v11, v6, v12 ; GFX10-NEXT: v_and_b32_e32 v4, v4, v10
; GFX10-NEXT: v_and_b32_e32 v6, v6, v10
; GFX10-NEXT: v_sub_nc_u32_e32 v7, 23, v5 ; GFX10-NEXT: v_sub_nc_u32_e32 v7, 23, v5
; GFX10-NEXT: v_and_b32_e32 v10, v5, v12 ; GFX10-NEXT: v_and_b32_e32 v5, v5, v10
; GFX10-NEXT: v_lshrrev_b32_e32 v2, v11, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v2, v6, v2
; GFX10-NEXT: v_and_b32_e32 v6, v7, v12 ; GFX10-NEXT: v_and_b32_e32 v7, v7, v10
; GFX10-NEXT: v_and_b32_e32 v7, v15, v12 ; GFX10-NEXT: v_lshl_or_b32 v0, v0, v4, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v3, v6, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, v7, v3
; GFX10-NEXT: v_lshl_or_b32 v0, v0, v7, v2 ; GFX10-NEXT: v_lshl_or_b32 v1, v1, v5, v3
; GFX10-NEXT: v_lshl_or_b32 v1, v1, v10, v3
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i24> @llvm.fshl.v2i24(<2 x i24> %lhs, <2 x i24> %rhs, <2 x i24> %amt) %result = call <2 x i24> @llvm.fshl.v2i24(<2 x i24> %lhs, <2 x i24> %rhs, <2 x i24> %amt)
ret <2 x i24> %result ret <2 x i24> %result
} }
define amdgpu_ps i32 @s_fshl_i32(i32 inreg %lhs, i32 inreg %rhs, i32 inreg %amt) { define amdgpu_ps i32 @s_fshl_i32(i32 inreg %lhs, i32 inreg %rhs, i32 inreg %amt) {
; GFX6-LABEL: s_fshl_i32: ; GFX6-LABEL: s_fshl_i32:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 364 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_alignbit_b32 v1, v1, v2, v3 ; GFX9-NEXT: v_alignbit_b32 v1, v1, v2, v3
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fshl_v2i32: ; GFX10-LABEL: v_fshl_v2i32:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_alignbit_b32 v2, v0, v2, 1 ; GFX10-NEXT: v_alignbit_b32 v2, v0, v2, 1
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 1, v0
; GFX10-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX10-NEXT: v_alignbit_b32 v3, v1, v3, 1 ; GFX10-NEXT: v_alignbit_b32 v3, v1, v3, 1
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 1, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v0, 1, v0
; GFX10-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 1, v1
; GFX10-NEXT: v_xor_b32_e32 v5, -1, v5 ; GFX10-NEXT: v_xor_b32_e32 v5, -1, v5
; GFX10-NEXT: v_alignbit_b32 v0, v7, v2, v4 ; GFX10-NEXT: v_alignbit_b32 v0, v0, v2, v4
; GFX10-NEXT: v_alignbit_b32 v1, v6, v3, v5 ; GFX10-NEXT: v_alignbit_b32 v1, v1, v3, v5
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %lhs, <2 x i32> %rhs, <2 x i32> %amt) %result = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %lhs, <2 x i32> %rhs, <2 x i32> %amt)
ret <2 x i32> %result ret <2 x i32> %result
} }
define <3 x i32> @v_fshl_v3i32(<3 x i32> %lhs, <3 x i32> %rhs, <3 x i32> %amt) { define <3 x i32> @v_fshl_v3i32(<3 x i32> %lhs, <3 x i32> %rhs, <3 x i32> %amt) {
; GFX6-LABEL: v_fshl_v3i32: ; GFX6-LABEL: v_fshl_v3i32:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 130 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_xor_b32_e32 v5, -1, v11 ; GFX9-NEXT: v_xor_b32_e32 v5, -1, v11
; GFX9-NEXT: v_alignbit_b32 v3, v3, v4, v5 ; GFX9-NEXT: v_alignbit_b32 v3, v3, v4, v5
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fshl_v4i32: ; GFX10-LABEL: v_fshl_v4i32:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_alignbit_b32 v22, v1, v5, 1 ; GFX10-NEXT: v_alignbit_b32 v4, v0, v4, 1
; GFX10-NEXT: v_alignbit_b32 v18, v0, v4, 1 ; GFX10-NEXT: v_alignbit_b32 v5, v1, v5, 1
; GFX10-NEXT: v_lshrrev_b32_e32 v15, 1, v0 ; GFX10-NEXT: v_alignbit_b32 v6, v2, v6, 1
; GFX10-NEXT: v_alignbit_b32 v7, v3, v7, 1
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 1, v0
; GFX10-NEXT: v_xor_b32_e32 v8, -1, v8 ; GFX10-NEXT: v_xor_b32_e32 v8, -1, v8
; GFX10-NEXT: v_lshrrev_b32_e32 v19, 1, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v1, 1, v1
; GFX10-NEXT: v_xor_b32_e32 v9, -1, v9 ; GFX10-NEXT: v_xor_b32_e32 v9, -1, v9
; GFX10-NEXT: v_alignbit_b32 v5, v2, v6, 1 ; GFX10-NEXT: v_lshrrev_b32_e32 v2, 1, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v23, 1, v2
; GFX10-NEXT: v_xor_b32_e32 v10, -1, v10 ; GFX10-NEXT: v_xor_b32_e32 v10, -1, v10
; GFX10-NEXT: v_alignbit_b32 v13, v3, v7, 1 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 1, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v14, 1, v3
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v11 ; GFX10-NEXT: v_xor_b32_e32 v11, -1, v11
; GFX10-NEXT: v_alignbit_b32 v0, v15, v18, v8 ; GFX10-NEXT: v_alignbit_b32 v0, v0, v4, v8
; GFX10-NEXT: v_alignbit_b32 v1, v19, v22, v9 ; GFX10-NEXT: v_alignbit_b32 v1, v1, v5, v9
; GFX10-NEXT: v_alignbit_b32 v2, v23, v5, v10 ; GFX10-NEXT: v_alignbit_b32 v2, v2, v6, v10
; GFX10-NEXT: v_alignbit_b32 v3, v14, v13, v11 ; GFX10-NEXT: v_alignbit_b32 v3, v3, v7, v11
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %lhs, <4 x i32> %rhs, <4 x i32> %amt) %result = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %lhs, <4 x i32> %rhs, <4 x i32> %amt)
ret <4 x i32> %result ret <4 x i32> %result
} }
define amdgpu_ps i16 @s_fshl_i16(i16 inreg %lhs, i16 inreg %rhs, i16 inreg %amt) { define amdgpu_ps i16 @s_fshl_i16(i16 inreg %lhs, i16 inreg %rhs, i16 inreg %amt) {
; GFX6-LABEL: s_fshl_i16: ; GFX6-LABEL: s_fshl_i16:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 1,374 Lines • ▼ Show 20 Lines
; GFX10-LABEL: v_fshl_v4i16: ; GFX10-LABEL: v_fshl_v4i16:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v6, -1, v4 ; GFX10-NEXT: v_xor_b32_e32 v6, -1, v4
; GFX10-NEXT: v_xor_b32_e32 v7, -1, v5 ; GFX10-NEXT: v_xor_b32_e32 v7, -1, v5
; GFX10-NEXT: s_mov_b32 s4, 0xf000f ; GFX10-NEXT: s_mov_b32 s4, 0xf000f
; GFX10-NEXT: v_pk_lshrrev_b16 v2, 1, v2 op_sel_hi:[0,1] ; GFX10-NEXT: v_pk_lshrrev_b16 v2, 1, v2 op_sel_hi:[0,1]
; GFX10-NEXT: v_and_b32_e32 v11, s4, v4 ; GFX10-NEXT: v_and_b32_e32 v4, s4, v4
; GFX10-NEXT: v_and_b32_e32 v15, s4, v6 ; GFX10-NEXT: v_and_b32_e32 v6, s4, v6
; GFX10-NEXT: v_and_b32_e32 v19, s4, v5 ; GFX10-NEXT: v_and_b32_e32 v5, s4, v5
; GFX10-NEXT: v_and_b32_e32 v6, s4, v7 ; GFX10-NEXT: v_and_b32_e32 v7, s4, v7
; GFX10-NEXT: v_pk_lshrrev_b16 v3, 1, v3 op_sel_hi:[0,1] ; GFX10-NEXT: v_pk_lshrrev_b16 v3, 1, v3 op_sel_hi:[0,1]
; GFX10-NEXT: v_pk_lshlrev_b16 v0, v11, v0 ; GFX10-NEXT: v_pk_lshlrev_b16 v0, v4, v0
; GFX10-NEXT: v_pk_lshrrev_b16 v2, v15, v2 ; GFX10-NEXT: v_pk_lshrrev_b16 v2, v6, v2
; GFX10-NEXT: v_pk_lshlrev_b16 v1, v19, v1 ; GFX10-NEXT: v_pk_lshlrev_b16 v1, v5, v1
; GFX10-NEXT: v_pk_lshrrev_b16 v3, v6, v3 ; GFX10-NEXT: v_pk_lshrrev_b16 v3, v7, v3
; GFX10-NEXT: v_or_b32_e32 v0, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v0, v0, v2
; GFX10-NEXT: v_or_b32_e32 v1, v1, v3 ; GFX10-NEXT: v_or_b32_e32 v1, v1, v3
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <4 x i16> @llvm.fshl.v4i16(<4 x i16> %lhs, <4 x i16> %rhs, <4 x i16> %amt) %result = call <4 x i16> @llvm.fshl.v4i16(<4 x i16> %lhs, <4 x i16> %rhs, <4 x i16> %amt)
%cast.result = bitcast <4 x i16> %result to <4 x half> %cast.result = bitcast <4 x i16> %result to <4 x half>
ret <4 x half> %cast.result ret <4 x half> %cast.result
} }
▲ Show 20 Lines • Show All 89 Lines • ▼ Show 20 Lines
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fshl_i64: ; GFX10-LABEL: v_fshl_i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v5, -1, v4 ; GFX10-NEXT: v_xor_b32_e32 v5, -1, v4
; GFX10-NEXT: v_lshrrev_b64 v[2:3], 1, v[2:3] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], 1, v[2:3]
; GFX10-NEXT: v_and_b32_e32 v7, 63, v4 ; GFX10-NEXT: v_and_b32_e32 v4, 63, v4
; GFX10-NEXT: v_and_b32_e32 v5, 63, v5 ; GFX10-NEXT: v_and_b32_e32 v5, 63, v5
; GFX10-NEXT: v_lshlrev_b64 v[0:1], v7, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], v4, v[0:1]
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v5, v[2:3] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], v5, v[2:3]
; GFX10-NEXT: v_or_b32_e32 v0, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v0, v0, v2
; GFX10-NEXT: v_or_b32_e32 v1, v1, v3 ; GFX10-NEXT: v_or_b32_e32 v1, v1, v3
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i64 @llvm.fshl.i64(i64 %lhs, i64 %rhs, i64 %amt) %result = call i64 @llvm.fshl.i64(i64 %lhs, i64 %rhs, i64 %amt)
ret i64 %result ret i64 %result
} }
▲ Show 20 Lines • Show All 394 Lines • ▼ Show 20 Lines
; GFX10-LABEL: v_fshl_v2i64: ; GFX10-LABEL: v_fshl_v2i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v9, -1, v8 ; GFX10-NEXT: v_xor_b32_e32 v9, -1, v8
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v10 ; GFX10-NEXT: v_xor_b32_e32 v11, -1, v10
; GFX10-NEXT: v_lshrrev_b64 v[4:5], 1, v[4:5] ; GFX10-NEXT: v_lshrrev_b64 v[4:5], 1, v[4:5]
; GFX10-NEXT: v_lshrrev_b64 v[6:7], 1, v[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[6:7], 1, v[6:7]
; GFX10-NEXT: v_and_b32_e32 v15, 63, v8 ; GFX10-NEXT: v_and_b32_e32 v8, 63, v8
; GFX10-NEXT: v_and_b32_e32 v19, 63, v9 ; GFX10-NEXT: v_and_b32_e32 v9, 63, v9
; GFX10-NEXT: v_and_b32_e32 v13, 63, v11 ; GFX10-NEXT: v_and_b32_e32 v10, 63, v10
; GFX10-NEXT: v_and_b32_e32 v9, 63, v10 ; GFX10-NEXT: v_and_b32_e32 v11, 63, v11
; GFX10-NEXT: v_lshlrev_b64 v[11:12], v15, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], v8, v[0:1]
; GFX10-NEXT: v_lshrrev_b64 v[4:5], v19, v[4:5] ; GFX10-NEXT: v_lshrrev_b64 v[4:5], v9, v[4:5]
; GFX10-NEXT: v_lshrrev_b64 v[6:7], v13, v[6:7] ; GFX10-NEXT: v_lshlrev_b64 v[2:3], v10, v[2:3]
; GFX10-NEXT: v_lshlrev_b64 v[15:16], v9, v[2:3] ; GFX10-NEXT: v_lshrrev_b64 v[6:7], v11, v[6:7]
; GFX10-NEXT: v_or_b32_e32 v0, v11, v4 ; GFX10-NEXT: v_or_b32_e32 v0, v0, v4
; GFX10-NEXT: v_or_b32_e32 v1, v12, v5 ; GFX10-NEXT: v_or_b32_e32 v1, v1, v5
; GFX10-NEXT: v_or_b32_e32 v2, v15, v6 ; GFX10-NEXT: v_or_b32_e32 v2, v2, v6
; GFX10-NEXT: v_or_b32_e32 v3, v16, v7 ; GFX10-NEXT: v_or_b32_e32 v3, v3, v7
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %lhs, <2 x i64> %rhs, <2 x i64> %amt) %result = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %lhs, <2 x i64> %rhs, <2 x i64> %amt)
ret <2 x i64> %result ret <2 x i64> %result
} }
define amdgpu_ps i128 @s_fshl_i128(i128 inreg %lhs, i128 inreg %rhs, i128 inreg %amt) { define amdgpu_ps i128 @s_fshl_i128(i128 inreg %lhs, i128 inreg %rhs, i128 inreg %amt) {
; GFX6-LABEL: s_fshl_i128: ; GFX6-LABEL: s_fshl_i128:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 447 Lines • ▼ Show 20 Lines
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fshl_i128: ; GFX10-LABEL: v_fshl_i128:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v9, -1, v8 ; GFX10-NEXT: v_xor_b32_e32 v9, -1, v8
; GFX10-NEXT: s_movk_i32 s4, 0x7f ; GFX10-NEXT: s_movk_i32 s4, 0x7f
; GFX10-NEXT: v_mov_b32_e32 v27, v2
; GFX10-NEXT: v_and_b32_e32 v18, s4, v8 ; GFX10-NEXT: v_and_b32_e32 v18, s4, v8
; GFX10-NEXT: v_mov_b32_e32 v28, v3
; GFX10-NEXT: v_and_b32_e32 v19, s4, v9 ; GFX10-NEXT: v_and_b32_e32 v19, s4, v9
; GFX10-NEXT: s_sub_i32 s4, 64, 1 ; GFX10-NEXT: s_sub_i32 s4, 64, 1
; GFX10-NEXT: v_lshrrev_b64 v[8:9], 1, v[4:5] ; GFX10-NEXT: v_lshrrev_b64 v[8:9], 1, v[4:5]
; GFX10-NEXT: v_lshlrev_b64 v[10:11], s4, v[6:7] ; GFX10-NEXT: v_lshlrev_b64 v[10:11], s4, v[6:7]
; GFX10-NEXT: s_sub_i32 s4, 1, 64 ; GFX10-NEXT: s_sub_i32 s4, 1, 64
; GFX10-NEXT: s_cmp_lt_u32 1, 64 ; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_lshrrev_b64 v[15:16], s4, v[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[12:13], s4, v[6:7]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0 ; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_or_b32_e32 v8, v8, v10 ; GFX10-NEXT: v_or_b32_e32 v8, v8, v10
; GFX10-NEXT: v_or_b32_e32 v9, v9, v11 ; GFX10-NEXT: v_or_b32_e32 v9, v9, v11
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: s_and_b32 s5, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s5, 1, vcc_lo
; GFX10-NEXT: s_and_b32 s4, 1, s4 ; GFX10-NEXT: s_and_b32 s4, 1, s4
; GFX10-NEXT: v_sub_nc_u32_e32 v14, 64, v18 ; GFX10-NEXT: v_sub_nc_u32_e32 v14, 64, v18
; GFX10-NEXT: v_lshrrev_b64 v[6:7], 1, v[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[6:7], 1, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e32 v3, v15, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v12, v12, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v15, v16, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v13, v13, v9, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s4 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5
; GFX10-NEXT: v_lshrrev_b64 v[8:9], v14, v[0:1] ; GFX10-NEXT: v_lshrrev_b64 v[8:9], v14, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[10:11], v18, v[27:28] ; GFX10-NEXT: v_lshlrev_b64 v[10:11], v18, v[2:3]
; GFX10-NEXT: v_sub_nc_u32_e32 v16, 64, v19 ; GFX10-NEXT: v_sub_nc_u32_e32 v16, 64, v19
; GFX10-NEXT: v_cndmask_b32_e32 v4, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, v12, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v5, v15, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v13, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v21, 0, v6, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, 0, v6, s4
; GFX10-NEXT: v_cndmask_b32_e64 v22, 0, v7, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v7, 0, v7, s4
; GFX10-NEXT: v_subrev_nc_u32_e32 v23, 64, v18 ; GFX10-NEXT: v_subrev_nc_u32_e32 v20, 64, v18
; GFX10-NEXT: v_or_b32_e32 v10, v8, v10 ; GFX10-NEXT: v_or_b32_e32 v10, v8, v10
; GFX10-NEXT: v_subrev_nc_u32_e32 v8, 64, v19 ; GFX10-NEXT: v_subrev_nc_u32_e32 v8, 64, v19
; GFX10-NEXT: v_lshrrev_b64 v[14:15], v19, v[4:5] ; GFX10-NEXT: v_lshrrev_b64 v[14:15], v19, v[4:5]
; GFX10-NEXT: v_lshlrev_b64 v[16:17], v16, v[21:22] ; GFX10-NEXT: v_lshlrev_b64 v[16:17], v16, v[6:7]
; GFX10-NEXT: v_lshlrev_b64 v[12:13], v18, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[12:13], v18, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[0:1], v23, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], v20, v[0:1]
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v18 ; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v18
; GFX10-NEXT: v_or_b32_e32 v11, v9, v11 ; GFX10-NEXT: v_or_b32_e32 v11, v9, v11
; GFX10-NEXT: v_lshrrev_b64 v[8:9], v8, v[21:22] ; GFX10-NEXT: v_lshrrev_b64 v[8:9], v8, v[6:7]
; GFX10-NEXT: v_or_b32_e32 v14, v14, v16 ; GFX10-NEXT: v_or_b32_e32 v14, v14, v16
; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v19 ; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v19
; GFX10-NEXT: v_or_b32_e32 v15, v15, v17 ; GFX10-NEXT: v_or_b32_e32 v15, v15, v17
; GFX10-NEXT: v_cndmask_b32_e32 v10, v0, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, v0, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v11, v1, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v11, v1, v11, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[0:1], v19, v[21:22] ; GFX10-NEXT: v_lshrrev_b64 v[0:1], v19, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v23, v8, v14, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v19
; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v18 ; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v18
; GFX10-NEXT: v_cndmask_b32_e64 v8, v8, v14, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v19
; GFX10-NEXT: v_cndmask_b32_e64 v6, v9, v15, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, v9, v15, s4
; GFX10-NEXT: v_cndmask_b32_e32 v31, 0, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v12, 0, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, 0, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v7, 0, v13, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v4, v23, v4, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v10, v2, s6
; GFX10-NEXT: v_cndmask_b32_e64 v15, v10, v27, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v8, v4, s5
; GFX10-NEXT: v_cndmask_b32_e64 v5, v6, v5, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v6, v5, s5
; GFX10-NEXT: v_cndmask_b32_e64 v6, 0, v0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, 0, v0, s4
; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, v1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, v1, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, v11, v28, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v11, v3, s6
; GFX10-NEXT: v_or_b32_e32 v0, v31, v4 ; GFX10-NEXT: v_or_b32_e32 v0, v12, v4
; GFX10-NEXT: v_or_b32_e32 v1, v7, v5 ; GFX10-NEXT: v_or_b32_e32 v1, v7, v5
; GFX10-NEXT: v_or_b32_e32 v2, v15, v6 ; GFX10-NEXT: v_or_b32_e32 v2, v2, v6
; GFX10-NEXT: v_or_b32_e32 v3, v3, v8 ; GFX10-NEXT: v_or_b32_e32 v3, v3, v8
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i128 @llvm.fshl.i128(i128 %lhs, i128 %rhs, i128 %amt) %result = call i128 @llvm.fshl.i128(i128 %lhs, i128 %rhs, i128 %amt)
ret i128 %result ret i128 %result
} }
define amdgpu_ps <4 x float> @v_fshl_i128_ssv(i128 inreg %lhs, i128 inreg %rhs, i128 %amt) { define amdgpu_ps <4 x float> @v_fshl_i128_ssv(i128 inreg %lhs, i128 inreg %rhs, i128 %amt) {
; GFX6-LABEL: v_fshl_i128_ssv: ; GFX6-LABEL: v_fshl_i128_ssv:
▲ Show 20 Lines • Show All 218 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_sub_nc_u32_e32 v0, 64, v13 ; GFX10-NEXT: v_sub_nc_u32_e32 v0, 64, v13
; GFX10-NEXT: s_cselect_b64 s[6:7], s[4:5], s[6:7] ; GFX10-NEXT: s_cselect_b64 s[6:7], s[4:5], s[6:7]
; GFX10-NEXT: s_cmp_lg_u32 s15, 0 ; GFX10-NEXT: s_cmp_lg_u32 s15, 0
; GFX10-NEXT: v_subrev_nc_u32_e32 v10, 64, v12 ; GFX10-NEXT: v_subrev_nc_u32_e32 v10, 64, v12
; GFX10-NEXT: s_cselect_b64 s[8:9], s[12:13], 0 ; GFX10-NEXT: s_cselect_b64 s[8:9], s[12:13], 0
; GFX10-NEXT: v_subrev_nc_u32_e32 v14, 64, v13 ; GFX10-NEXT: v_subrev_nc_u32_e32 v14, 64, v13
; GFX10-NEXT: v_lshrrev_b64 v[6:7], v13, s[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[6:7], v13, s[6:7]
; GFX10-NEXT: v_lshlrev_b64 v[8:9], v0, s[8:9] ; GFX10-NEXT: v_lshlrev_b64 v[8:9], v0, s[8:9]
; GFX10-NEXT: v_lshlrev_b64 v[15:16], v10, s[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[10:11], v10, s[0:1]
; GFX10-NEXT: v_or_b32_e32 v3, v1, v3 ; GFX10-NEXT: v_or_b32_e32 v3, v1, v3
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v12 ; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v12
; GFX10-NEXT: v_lshrrev_b64 v[0:1], v14, s[8:9] ; GFX10-NEXT: v_lshrrev_b64 v[0:1], v14, s[8:9]
; GFX10-NEXT: v_lshlrev_b64 v[4:5], v12, s[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[4:5], v12, s[0:1]
; GFX10-NEXT: v_or_b32_e32 v6, v6, v8 ; GFX10-NEXT: v_or_b32_e32 v6, v6, v8
; GFX10-NEXT: v_cmp_gt_u32_e64 s0, 64, v13 ; GFX10-NEXT: v_cmp_gt_u32_e64 s0, 64, v13
; GFX10-NEXT: v_or_b32_e32 v7, v7, v9 ; GFX10-NEXT: v_or_b32_e32 v7, v7, v9
; GFX10-NEXT: v_cndmask_b32_e32 v8, v15, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v8, v10, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v10, v16, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, v11, v3, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v13, s[8:9] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], v13, s[8:9]
; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 0, v12
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v6, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v6, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 0, v13 ; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 0, v13
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v7, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v7, s0
; GFX10-NEXT: v_cndmask_b32_e32 v11, 0, v4, vcc_lo ; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 0, v12
; GFX10-NEXT: v_cndmask_b32_e32 v15, 0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, 0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v19, v8, s2, s4 ; GFX10-NEXT: v_cndmask_b32_e32 v5, 0, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, s6, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, s6, s1
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s7, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s7, s1
; GFX10-NEXT: v_cndmask_b32_e64 v6, v8, s2, s4
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v6, v10, s3, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v7, v10, s3, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0
; GFX10-NEXT: v_or_b32_e32 v0, v11, v0 ; GFX10-NEXT: v_or_b32_e32 v0, v4, v0
; GFX10-NEXT: v_or_b32_e32 v1, v15, v1 ; GFX10-NEXT: v_or_b32_e32 v1, v5, v1
; GFX10-NEXT: v_or_b32_e32 v2, v19, v2 ; GFX10-NEXT: v_or_b32_e32 v2, v6, v2
; GFX10-NEXT: v_or_b32_e32 v3, v6, v3 ; GFX10-NEXT: v_or_b32_e32 v3, v7, v3
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.fshl.i128(i128 %lhs, i128 %rhs, i128 %amt) %result = call i128 @llvm.fshl.i128(i128 %lhs, i128 %rhs, i128 %amt)
%cast.result = bitcast i128 %result to <4 x float> %cast.result = bitcast i128 %result to <4 x float>
ret <4 x float> %cast.result ret <4 x float> %cast.result
} }
define amdgpu_ps <4 x float> @v_fshl_i128_svs(i128 inreg %lhs, i128 %rhs, i128 inreg %amt) { define amdgpu_ps <4 x float> @v_fshl_i128_svs(i128 inreg %lhs, i128 %rhs, i128 inreg %amt) {
; GFX6-LABEL: v_fshl_i128_svs: ; GFX6-LABEL: v_fshl_i128_svs:
▲ Show 20 Lines • Show All 240 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_cselect_b64 s[8:9], s[8:9], 0 ; GFX10-NEXT: s_cselect_b64 s[8:9], s[8:9], 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[6:7], s[0:1] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[6:7], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s13, 0 ; GFX10-NEXT: s_cmp_lg_u32 s13, 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[0:1] ; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[0:1]
; GFX10-NEXT: s_sub_i32 s0, 64, 1 ; GFX10-NEXT: s_sub_i32 s0, 64, 1
; GFX10-NEXT: v_lshlrev_b64 v[6:7], s0, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[6:7], s0, v[2:3]
; GFX10-NEXT: s_sub_i32 s0, 1, 64 ; GFX10-NEXT: s_sub_i32 s0, 1, 64
; GFX10-NEXT: s_cmp_lt_u32 1, 64 ; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_lshrrev_b64 v[11:12], s0, v[2:3] ; GFX10-NEXT: v_lshrrev_b64 v[8:9], s0, v[2:3]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0 ; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_or_b32_e32 v4, v4, v6 ; GFX10-NEXT: v_or_b32_e32 v4, v4, v6
; GFX10-NEXT: v_or_b32_e32 v5, v5, v7 ; GFX10-NEXT: v_or_b32_e32 v5, v5, v7
; GFX10-NEXT: s_cselect_b32 s0, 1, 0 ; GFX10-NEXT: s_cselect_b32 s0, 1, 0
; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo
; GFX10-NEXT: s_and_b32 s0, 1, s0 ; GFX10-NEXT: s_and_b32 s0, 1, s0
; GFX10-NEXT: v_lshrrev_b64 v[2:3], 1, v[2:3] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], 1, v[2:3]
; GFX10-NEXT: v_cndmask_b32_e32 v7, v11, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, v8, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v11, v12, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v9, v5, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1 ; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1
; GFX10-NEXT: v_cndmask_b32_e32 v0, v7, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v11, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0
; GFX10-NEXT: s_sub_i32 s0, 64, s4 ; GFX10-NEXT: s_sub_i32 s0, 64, s4
; GFX10-NEXT: v_lshrrev_b64 v[4:5], s4, v[0:1] ; GFX10-NEXT: v_lshrrev_b64 v[4:5], s4, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[6:7], s0, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[6:7], s0, v[2:3]
; GFX10-NEXT: s_sub_i32 s0, s4, 64 ; GFX10-NEXT: s_sub_i32 s0, s4, 64
; GFX10-NEXT: s_cmp_lt_u32 s4, 64 ; GFX10-NEXT: s_cmp_lt_u32 s4, 64
; GFX10-NEXT: v_lshrrev_b64 v[11:12], s0, v[2:3] ; GFX10-NEXT: v_lshrrev_b64 v[8:9], s0, v[2:3]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s4, 0 ; GFX10-NEXT: s_cmp_eq_u32 s4, 0
; GFX10-NEXT: v_or_b32_e32 v4, v4, v6 ; GFX10-NEXT: v_or_b32_e32 v4, v4, v6
; GFX10-NEXT: v_or_b32_e32 v5, v5, v7 ; GFX10-NEXT: v_or_b32_e32 v5, v5, v7
; GFX10-NEXT: s_cselect_b32 s0, 1, 0 ; GFX10-NEXT: s_cselect_b32 s0, 1, 0
; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo
; GFX10-NEXT: s_and_b32 s0, 1, s0 ; GFX10-NEXT: s_and_b32 s0, 1, s0
; GFX10-NEXT: v_lshrrev_b64 v[2:3], s4, v[2:3] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], s4, v[2:3]
; GFX10-NEXT: v_cndmask_b32_e32 v7, v11, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, v8, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v11, v12, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v9, v5, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1 ; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1
; GFX10-NEXT: v_cndmask_b32_e32 v0, v7, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v11, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0
; GFX10-NEXT: v_or_b32_e32 v0, s8, v0 ; GFX10-NEXT: v_or_b32_e32 v0, s8, v0
; GFX10-NEXT: v_or_b32_e32 v1, s9, v1 ; GFX10-NEXT: v_or_b32_e32 v1, s9, v1
; GFX10-NEXT: v_or_b32_e32 v2, s2, v2 ; GFX10-NEXT: v_or_b32_e32 v2, s2, v2
; GFX10-NEXT: v_or_b32_e32 v3, s3, v3 ; GFX10-NEXT: v_or_b32_e32 v3, s3, v3
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.fshl.i128(i128 %lhs, i128 %rhs, i128 %amt) %result = call i128 @llvm.fshl.i128(i128 %lhs, i128 %rhs, i128 %amt)
▲ Show 20 Lines • Show All 216 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_sub_i32 s5, s8, 64 ; GFX10-NEXT: s_sub_i32 s5, s8, 64
; GFX10-NEXT: s_cmp_lt_u32 s8, 64 ; GFX10-NEXT: s_cmp_lt_u32 s8, 64
; GFX10-NEXT: v_lshrrev_b64 v[4:5], s9, v[0:1] ; GFX10-NEXT: v_lshrrev_b64 v[4:5], s9, v[0:1]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s8, 0 ; GFX10-NEXT: s_cmp_eq_u32 s8, 0
; GFX10-NEXT: v_lshlrev_b64 v[6:7], s8, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[6:7], s8, v[2:3]
; GFX10-NEXT: v_lshlrev_b64 v[8:9], s8, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[8:9], s8, v[0:1]
; GFX10-NEXT: s_cselect_b32 s6, 1, 0 ; GFX10-NEXT: s_cselect_b32 s6, 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[11:12], s5, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], s5, v[0:1]
; GFX10-NEXT: s_and_b32 s5, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s5, 1, vcc_lo
; GFX10-NEXT: s_and_b32 s12, 1, s6 ; GFX10-NEXT: s_and_b32 s12, 1, s6
; GFX10-NEXT: s_sub_i32 s13, 1, 64 ; GFX10-NEXT: s_sub_i32 s13, 1, 64
; GFX10-NEXT: s_sub_i32 s8, 64, 1 ; GFX10-NEXT: s_sub_i32 s8, 64, 1
; GFX10-NEXT: s_cmp_lt_u32 1, 64 ; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_or_b32_e32 v4, v4, v6 ; GFX10-NEXT: v_or_b32_e32 v4, v4, v6
; GFX10-NEXT: v_or_b32_e32 v5, v5, v7 ; GFX10-NEXT: v_or_b32_e32 v5, v5, v7
; GFX10-NEXT: v_cndmask_b32_e32 v6, 0, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v6, 0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, 0, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v7, 0, v9, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s5 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s5
; GFX10-NEXT: s_cselect_b32 s5, 1, 0 ; GFX10-NEXT: s_cselect_b32 s5, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0 ; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: s_cselect_b32 s14, 1, 0 ; GFX10-NEXT: s_cselect_b32 s14, 1, 0
; GFX10-NEXT: s_lshr_b64 s[6:7], s[0:1], 1 ; GFX10-NEXT: s_lshr_b64 s[6:7], s[0:1], 1
; GFX10-NEXT: s_lshl_b64 s[8:9], s[2:3], s8 ; GFX10-NEXT: s_lshl_b64 s[8:9], s[2:3], s8
; GFX10-NEXT: s_lshr_b64 s[10:11], s[2:3], 1 ; GFX10-NEXT: s_lshr_b64 s[10:11], s[2:3], 1
; GFX10-NEXT: s_or_b64 s[6:7], s[6:7], s[8:9] ; GFX10-NEXT: s_or_b64 s[6:7], s[6:7], s[8:9]
; GFX10-NEXT: s_lshr_b64 s[2:3], s[2:3], s13 ; GFX10-NEXT: s_lshr_b64 s[2:3], s[2:3], s13
; GFX10-NEXT: s_cmp_lg_u32 s5, 0 ; GFX10-NEXT: s_cmp_lg_u32 s5, 0
; GFX10-NEXT: v_cndmask_b32_e32 v0, v11, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: s_cselect_b64 s[2:3], s[6:7], s[2:3] ; GFX10-NEXT: s_cselect_b64 s[2:3], s[6:7], s[2:3]
; GFX10-NEXT: s_cmp_lg_u32 s14, 0 ; GFX10-NEXT: s_cmp_lg_u32 s14, 0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v12, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[2:3] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[2:3]
; GFX10-NEXT: s_cmp_lg_u32 s5, 0 ; GFX10-NEXT: s_cmp_lg_u32 s5, 0
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s12 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s12
; GFX10-NEXT: s_cselect_b64 s[2:3], s[10:11], 0 ; GFX10-NEXT: s_cselect_b64 s[2:3], s[10:11], 0
; GFX10-NEXT: s_sub_i32 s10, s4, 64 ; GFX10-NEXT: s_sub_i32 s10, s4, 64
; GFX10-NEXT: s_sub_i32 s5, 64, s4 ; GFX10-NEXT: s_sub_i32 s5, 64, s4
; GFX10-NEXT: s_cmp_lt_u32 s4, 64 ; GFX10-NEXT: s_cmp_lt_u32 s4, 64
; GFX10-NEXT: v_cndmask_b32_e32 v2, v0, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v0, v2, vcc_lo
▲ Show 20 Lines • Show All 354 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_or_b32_e32 v2, v2, v4 ; GFX9-NEXT: v_or_b32_e32 v2, v2, v4
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_fshl_i128_65: ; GFX10-LABEL: v_fshl_i128_65:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: s_movk_i32 s4, 0x41 ; GFX10-NEXT: s_movk_i32 s4, 0x41
; GFX10-NEXT: v_lshrrev_b32_e32 v19, 31, v5 ; GFX10-NEXT: v_lshrrev_b32_e32 v16, 31, v5
; GFX10-NEXT: s_sub_i32 s5, 64, s4 ; GFX10-NEXT: s_sub_i32 s5, 64, s4
; GFX10-NEXT: v_lshlrev_b64 v[10:11], s4, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[10:11], s4, v[2:3]
; GFX10-NEXT: v_lshrrev_b64 v[8:9], s5, v[0:1] ; GFX10-NEXT: v_lshrrev_b64 v[8:9], s5, v[0:1]
; GFX10-NEXT: s_sub_i32 s5, s4, 64 ; GFX10-NEXT: s_sub_i32 s5, s4, 64
; GFX10-NEXT: s_cmp_lt_u32 s4, 64 ; GFX10-NEXT: s_cmp_lt_u32 s4, 64
; GFX10-NEXT: v_lshlrev_b64 v[12:13], s4, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[12:13], s4, v[0:1]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s4, 0 ; GFX10-NEXT: s_cmp_eq_u32 s4, 0
; GFX10-NEXT: v_lshlrev_b64 v[0:1], s5, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], s5, v[0:1]
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: s_sub_i32 s5, 64, 63 ; GFX10-NEXT: s_sub_i32 s5, 64, 63
; GFX10-NEXT: v_or_b32_e32 v15, v9, v11
; GFX10-NEXT: v_or_b32_e32 v14, v8, v10 ; GFX10-NEXT: v_or_b32_e32 v14, v8, v10
; GFX10-NEXT: v_or_b32_e32 v15, v9, v11
; GFX10-NEXT: v_lshlrev_b64 v[8:9], s5, v[6:7] ; GFX10-NEXT: v_lshlrev_b64 v[8:9], s5, v[6:7]
; GFX10-NEXT: s_and_b32 s6, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s6, 1, vcc_lo
; GFX10-NEXT: s_and_b32 s7, 1, s4 ; GFX10-NEXT: s_and_b32 s7, 1, s4
; GFX10-NEXT: s_sub_i32 s4, 63, 64 ; GFX10-NEXT: s_sub_i32 s4, 63, 64
; GFX10-NEXT: v_cndmask_b32_e32 v11, 0, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v12, 0, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v27, 0, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v13, 0, v13, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s6 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s6
; GFX10-NEXT: v_lshrrev_b64 v[23:24], s4, v[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[10:11], s4, v[6:7]
; GFX10-NEXT: s_cmp_lt_u32 63, 64 ; GFX10-NEXT: s_cmp_lt_u32 63, 64
; GFX10-NEXT: v_or_b32_e32 v6, v19, v8 ; GFX10-NEXT: v_or_b32_e32 v6, v16, v8
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 63, 0 ; GFX10-NEXT: s_cmp_eq_u32 63, 0
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 s6, 0, s7 ; GFX10-NEXT: v_cmp_ne_u32_e64 s6, 0, s7
; GFX10-NEXT: s_cselect_b32 s5, 1, 0 ; GFX10-NEXT: s_cselect_b32 s5, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v6, v23, v6, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, v10, v6, s4
; GFX10-NEXT: s_and_b32 s5, 1, s5 ; GFX10-NEXT: s_and_b32 s5, 1, s5
; GFX10-NEXT: v_cndmask_b32_e64 v8, v24, v9, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v8, v11, v9, s4
; GFX10-NEXT: s_and_b32 s4, 1, s4 ; GFX10-NEXT: s_and_b32 s4, 1, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, s5 ; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, s5
; GFX10-NEXT: v_cndmask_b32_e64 v19, v0, v2, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v0, v2, s6
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 31, v7 ; GFX10-NEXT: v_lshrrev_b32_e32 v0, 31, v7
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4
; GFX10-NEXT: v_cndmask_b32_e32 v10, v1, v15, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v7, v1, v15, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v4, v6, v4, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v6, v4, s5
; GFX10-NEXT: v_cndmask_b32_e64 v5, v8, v5, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v8, v5, s5
; GFX10-NEXT: v_cndmask_b32_e64 v6, 0, v0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, 0, v0, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, v10, v3, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v7, v3, s6
; GFX10-NEXT: v_or_b32_e32 v0, v11, v4 ; GFX10-NEXT: v_or_b32_e32 v0, v12, v4
; GFX10-NEXT: v_or_b32_e32 v1, v27, v5 ; GFX10-NEXT: v_or_b32_e32 v1, v13, v5
; GFX10-NEXT: v_or_b32_e32 v2, v19, v6 ; GFX10-NEXT: v_or_b32_e32 v2, v2, v6
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i128 @llvm.fshl.i128(i128 %lhs, i128 %rhs, i128 65) %result = call i128 @llvm.fshl.i128(i128 %lhs, i128 %rhs, i128 65)
ret i128 %result ret i128 %result
} }
define amdgpu_ps <2 x i128> @s_fshl_v2i128(<2 x i128> inreg %lhs, <2 x i128> inreg %rhs, <2 x i128> inreg %amt) { define amdgpu_ps <2 x i128> @s_fshl_v2i128(<2 x i128> inreg %lhs, <2 x i128> inreg %rhs, <2 x i128> inreg %amt) {
; GFX6-LABEL: s_fshl_v2i128: ; GFX6-LABEL: s_fshl_v2i128:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 330 Lines • ▼ Show 20 Lines
; GFX9-NEXT: s_or_b64 s[4:5], s[16:17], s[4:5] ; GFX9-NEXT: s_or_b64 s[4:5], s[16:17], s[4:5]
; GFX9-NEXT: s_or_b64 s[6:7], s[6:7], s[8:9] ; GFX9-NEXT: s_or_b64 s[6:7], s[6:7], s[8:9]
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: s_fshl_v2i128: ; GFX10-LABEL: s_fshl_v2i128:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_movk_i32 s18, 0x7f ; GFX10-NEXT: s_movk_i32 s18, 0x7f
; GFX10-NEXT: s_mov_b32 s19, 0 ; GFX10-NEXT: s_mov_b32 s19, 0
; GFX10-NEXT: s_mov_b32 s30, s0
; GFX10-NEXT: s_and_b64 s[22:23], s[16:17], s[18:19] ; GFX10-NEXT: s_and_b64 s[22:23], s[16:17], s[18:19]
; GFX10-NEXT: s_andn2_b64 s[16:17], s[18:19], s[16:17] ; GFX10-NEXT: s_andn2_b64 s[16:17], s[18:19], s[16:17]
; GFX10-NEXT: s_sub_i32 s17, s22, 64 ; GFX10-NEXT: s_sub_i32 s17, s22, 64
; GFX10-NEXT: s_sub_i32 s23, 64, s22 ; GFX10-NEXT: s_sub_i32 s23, 64, s22
; GFX10-NEXT: s_cmp_lt_u32 s22, 64 ; GFX10-NEXT: s_cmp_lt_u32 s22, 64
; GFX10-NEXT: s_mov_b32 s31, s1
; GFX10-NEXT: s_cselect_b32 s28, 1, 0 ; GFX10-NEXT: s_cselect_b32 s28, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s22, 0 ; GFX10-NEXT: s_cmp_eq_u32 s22, 0
; GFX10-NEXT: s_cselect_b32 s29, 1, 0 ; GFX10-NEXT: s_cselect_b32 s29, 1, 0
; GFX10-NEXT: s_lshr_b64 s[24:25], s[30:31], s23 ; GFX10-NEXT: s_lshr_b64 s[24:25], s[0:1], s23
; GFX10-NEXT: s_lshl_b64 s[26:27], s[2:3], s22 ; GFX10-NEXT: s_lshl_b64 s[26:27], s[2:3], s22
; GFX10-NEXT: s_lshl_b64 s[22:23], s[30:31], s22 ; GFX10-NEXT: s_lshl_b64 s[22:23], s[0:1], s22
; GFX10-NEXT: s_or_b64 s[24:25], s[24:25], s[26:27] ; GFX10-NEXT: s_or_b64 s[24:25], s[24:25], s[26:27]
; GFX10-NEXT: s_lshl_b64 s[0:1], s[30:31], s17 ; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], s17
; GFX10-NEXT: s_cmp_lg_u32 s28, 0 ; GFX10-NEXT: s_cmp_lg_u32 s28, 0
; GFX10-NEXT: s_cselect_b64 s[22:23], s[22:23], 0 ; GFX10-NEXT: s_cselect_b64 s[22:23], s[22:23], 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[24:25], s[0:1] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[24:25], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s29, 0 ; GFX10-NEXT: s_cmp_lg_u32 s29, 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[0:1] ; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[0:1]
; GFX10-NEXT: s_sub_i32 s28, 1, 64 ; GFX10-NEXT: s_sub_i32 s28, 1, 64
; GFX10-NEXT: s_sub_i32 s29, 64, 1 ; GFX10-NEXT: s_sub_i32 s29, 64, 1
; GFX10-NEXT: s_cmp_lt_u32 1, 64 ; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: s_cselect_b32 s17, 1, 0 ; GFX10-NEXT: s_cselect_b32 s17, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0 ; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: s_cselect_b32 s30, 1, 0 ; GFX10-NEXT: s_cselect_b32 s30, 1, 0
; GFX10-NEXT: s_lshr_b64 s[0:1], s[8:9], 1 ; GFX10-NEXT: s_lshr_b64 s[0:1], s[8:9], 1
; GFX10-NEXT: s_lshl_b64 s[24:25], s[10:11], s29 ; GFX10-NEXT: s_lshl_b64 s[24:25], s[10:11], s29
; GFX10-NEXT: s_lshr_b64 s[26:27], s[10:11], 1 ; GFX10-NEXT: s_lshr_b64 s[26:27], s[10:11], 1
; GFX10-NEXT: s_or_b64 s[0:1], s[0:1], s[24:25] ; GFX10-NEXT: s_or_b64 s[0:1], s[0:1], s[24:25]
; GFX10-NEXT: s_lshr_b64 s[10:11], s[10:11], s28 ; GFX10-NEXT: s_lshr_b64 s[10:11], s[10:11], s28
; GFX10-NEXT: s_cmp_lg_u32 s17, 0 ; GFX10-NEXT: s_cmp_lg_u32 s17, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[10:11] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[10:11]
; GFX10-NEXT: s_cmp_lg_u32 s30, 0 ; GFX10-NEXT: s_cmp_lg_u32 s30, 0
; GFX10-NEXT: s_cselect_b64 s[46:47], s[8:9], s[0:1] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[8:9], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s17, 0 ; GFX10-NEXT: s_cmp_lg_u32 s17, 0
; GFX10-NEXT: s_cselect_b64 s[8:9], s[26:27], 0 ; GFX10-NEXT: s_cselect_b64 s[8:9], s[26:27], 0
; GFX10-NEXT: s_sub_i32 s26, s16, 64 ; GFX10-NEXT: s_sub_i32 s26, s16, 64
; GFX10-NEXT: s_sub_i32 s17, 64, s16 ; GFX10-NEXT: s_sub_i32 s17, 64, s16
; GFX10-NEXT: s_cmp_lt_u32 s16, 64 ; GFX10-NEXT: s_cmp_lt_u32 s16, 64
; GFX10-NEXT: s_cselect_b32 s27, 1, 0 ; GFX10-NEXT: s_cselect_b32 s27, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s16, 0 ; GFX10-NEXT: s_cmp_eq_u32 s16, 0
; GFX10-NEXT: s_cselect_b32 s30, 1, 0 ; GFX10-NEXT: s_cselect_b32 s30, 1, 0
; GFX10-NEXT: s_lshr_b64 s[10:11], s[46:47], s16 ; GFX10-NEXT: s_lshr_b64 s[10:11], s[0:1], s16
; GFX10-NEXT: s_lshl_b64 s[24:25], s[8:9], s17 ; GFX10-NEXT: s_lshl_b64 s[24:25], s[8:9], s17
; GFX10-NEXT: s_lshr_b64 s[16:17], s[8:9], s16 ; GFX10-NEXT: s_lshr_b64 s[16:17], s[8:9], s16
; GFX10-NEXT: s_or_b64 s[10:11], s[10:11], s[24:25] ; GFX10-NEXT: s_or_b64 s[10:11], s[10:11], s[24:25]
; GFX10-NEXT: s_lshr_b64 s[8:9], s[8:9], s26 ; GFX10-NEXT: s_lshr_b64 s[8:9], s[8:9], s26
; GFX10-NEXT: s_cmp_lg_u32 s27, 0 ; GFX10-NEXT: s_cmp_lg_u32 s27, 0
; GFX10-NEXT: s_cselect_b64 s[8:9], s[10:11], s[8:9] ; GFX10-NEXT: s_cselect_b64 s[8:9], s[10:11], s[8:9]
; GFX10-NEXT: s_cmp_lg_u32 s30, 0 ; GFX10-NEXT: s_cmp_lg_u32 s30, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[46:47], s[8:9] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[8:9]
; GFX10-NEXT: s_cmp_lg_u32 s27, 0 ; GFX10-NEXT: s_cmp_lg_u32 s27, 0
; GFX10-NEXT: s_cselect_b64 s[8:9], s[16:17], 0 ; GFX10-NEXT: s_cselect_b64 s[8:9], s[16:17], 0
; GFX10-NEXT: s_andn2_b64 s[10:11], s[18:19], s[20:21] ; GFX10-NEXT: s_andn2_b64 s[10:11], s[18:19], s[20:21]
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[8:9] ; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[8:9]
; GFX10-NEXT: s_and_b64 s[8:9], s[20:21], s[18:19] ; GFX10-NEXT: s_and_b64 s[8:9], s[20:21], s[18:19]
; GFX10-NEXT: s_or_b64 s[0:1], s[22:23], s[0:1] ; GFX10-NEXT: s_or_b64 s[0:1], s[22:23], s[0:1]
; GFX10-NEXT: s_sub_i32 s11, s8, 64 ; GFX10-NEXT: s_sub_i32 s11, s8, 64
; GFX10-NEXT: s_sub_i32 s9, 64, s8 ; GFX10-NEXT: s_sub_i32 s9, 64, s8
▲ Show 20 Lines • Show All 451 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshrrev_b64 v[21:22], s9, v[10:11] ; GFX10-NEXT: v_lshrrev_b64 v[21:22], s9, v[10:11]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0 ; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_or_b32_e32 v16, v16, v18 ; GFX10-NEXT: v_or_b32_e32 v16, v16, v18
; GFX10-NEXT: v_or_b32_e32 v17, v17, v19 ; GFX10-NEXT: v_or_b32_e32 v17, v17, v19
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: s_and_b32 s5, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s5, 1, vcc_lo
; GFX10-NEXT: s_and_b32 s4, 1, s4 ; GFX10-NEXT: s_and_b32 s4, 1, s4
; GFX10-NEXT: v_mov_b32_e32 v29, v2
; GFX10-NEXT: v_mov_b32_e32 v30, v3
; GFX10-NEXT: v_sub_nc_u32_e32 v23, 64, v27 ; GFX10-NEXT: v_sub_nc_u32_e32 v23, 64, v27
; GFX10-NEXT: v_lshrrev_b64 v[10:11], 1, v[10:11] ; GFX10-NEXT: v_lshrrev_b64 v[10:11], 1, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e32 v21, v21, v16, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v21, v21, v16, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v22, v22, v17, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v22, v22, v17, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s4 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5
; GFX10-NEXT: v_lshrrev_b64 v[16:17], v23, v[0:1] ; GFX10-NEXT: v_lshrrev_b64 v[16:17], v23, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[18:19], v27, v[29:30] ; GFX10-NEXT: v_lshlrev_b64 v[18:19], v27, v[2:3]
; GFX10-NEXT: v_sub_nc_u32_e32 v25, 64, v28 ; GFX10-NEXT: v_sub_nc_u32_e32 v25, 64, v28
; GFX10-NEXT: v_cndmask_b32_e32 v34, v21, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v8, v21, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v35, v22, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v9, v22, v9, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v10, 0, v10, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v10, 0, v10, s4
; GFX10-NEXT: v_cndmask_b32_e64 v11, 0, v11, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v11, 0, v11, s4
; GFX10-NEXT: v_or_b32_e32 v18, v16, v18 ; GFX10-NEXT: v_or_b32_e32 v18, v16, v18
; GFX10-NEXT: v_subrev_nc_u32_e32 v31, 64, v27 ; GFX10-NEXT: v_subrev_nc_u32_e32 v29, 64, v27
; GFX10-NEXT: v_subrev_nc_u32_e32 v16, 64, v28 ; GFX10-NEXT: v_subrev_nc_u32_e32 v16, 64, v28
; GFX10-NEXT: v_lshrrev_b64 v[23:24], v28, v[34:35] ; GFX10-NEXT: v_lshrrev_b64 v[23:24], v28, v[8:9]
; GFX10-NEXT: v_lshlrev_b64 v[25:26], v25, v[10:11] ; GFX10-NEXT: v_lshlrev_b64 v[25:26], v25, v[10:11]
; GFX10-NEXT: v_lshlrev_b64 v[21:22], v27, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[21:22], v27, v[0:1]
; GFX10-NEXT: v_or_b32_e32 v19, v17, v19 ; GFX10-NEXT: v_or_b32_e32 v19, v17, v19
; GFX10-NEXT: v_lshlrev_b64 v[0:1], v31, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], v29, v[0:1]
; GFX10-NEXT: v_lshrrev_b64 v[16:17], v16, v[10:11] ; GFX10-NEXT: v_lshrrev_b64 v[16:17], v16, v[10:11]
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v27 ; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v27
; GFX10-NEXT: v_or_b32_e32 v23, v23, v25 ; GFX10-NEXT: v_or_b32_e32 v23, v23, v25
; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v28 ; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v28
; GFX10-NEXT: v_or_b32_e32 v24, v24, v26 ; GFX10-NEXT: v_or_b32_e32 v24, v24, v26
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v28 ; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v28
; GFX10-NEXT: v_cndmask_b32_e32 v18, v0, v18, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v18, v0, v18, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v19, v1, v19, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v19, v1, v19, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[0:1], v28, v[10:11] ; GFX10-NEXT: v_lshrrev_b64 v[0:1], v28, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e64 v16, v16, v23, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v16, v16, v23, s4
; GFX10-NEXT: v_cndmask_b32_e64 v10, v17, v24, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v10, v17, v24, s4
; GFX10-NEXT: v_cndmask_b32_e32 v21, 0, v21, vcc_lo ; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v27
; GFX10-NEXT: v_cndmask_b32_e32 v11, 0, v22, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v11, 0, v22, vcc_lo
; GFX10-NEXT: s_cmp_lt_u32 1, 64 ; GFX10-NEXT: v_cndmask_b32_e32 v21, 0, v21, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v8, v16, v34, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v8, v16, v8, s5
; GFX10-NEXT: v_cndmask_b32_e64 v9, v10, v35, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v9, v10, v9, s5
; GFX10-NEXT: v_cndmask_b32_e64 v22, v19, v3, s6
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v0, s4
; GFX10-NEXT: v_cndmask_b32_e64 v23, 0, v1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v23, 0, v1, s4
; GFX10-NEXT: v_xor_b32_e32 v16, -1, v20 ; GFX10-NEXT: v_xor_b32_e32 v16, -1, v20
; GFX10-NEXT: v_or_b32_e32 v0, v21, v8
; GFX10-NEXT: v_or_b32_e32 v1, v11, v9 ; GFX10-NEXT: v_or_b32_e32 v1, v11, v9
; GFX10-NEXT: v_or_b32_e32 v0, v21, v8
; GFX10-NEXT: v_lshrrev_b64 v[8:9], 1, v[12:13] ; GFX10-NEXT: v_lshrrev_b64 v[8:9], 1, v[12:13]
; GFX10-NEXT: v_lshlrev_b64 v[10:11], s8, v[14:15] ; GFX10-NEXT: v_lshlrev_b64 v[10:11], s8, v[14:15]
; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v27 ; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_and_b32_e32 v27, s7, v16 ; GFX10-NEXT: v_and_b32_e32 v25, s7, v16
; GFX10-NEXT: v_and_b32_e32 v24, s7, v20
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0 ; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_lshrrev_b64 v[16:17], s9, v[14:15] ; GFX10-NEXT: v_lshrrev_b64 v[16:17], s9, v[14:15]
; GFX10-NEXT: v_or_b32_e32 v11, v9, v11
; GFX10-NEXT: v_or_b32_e32 v10, v8, v10 ; GFX10-NEXT: v_or_b32_e32 v10, v8, v10
; GFX10-NEXT: v_or_b32_e32 v11, v9, v11
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: s_and_b32 s5, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s5, 1, vcc_lo
; GFX10-NEXT: s_and_b32 s4, 1, s4 ; GFX10-NEXT: s_and_b32 s4, 1, s4
; GFX10-NEXT: v_and_b32_e32 v24, s7, v20 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v18, v2, s6
; GFX10-NEXT: v_lshrrev_b64 v[8:9], 1, v[14:15] ; GFX10-NEXT: v_lshrrev_b64 v[8:9], 1, v[14:15]
; GFX10-NEXT: v_cndmask_b32_e64 v22, v19, v30, s6 ; GFX10-NEXT: v_sub_nc_u32_e32 v18, 64, v24
; GFX10-NEXT: v_cndmask_b32_e32 v19, v16, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v16, v16, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v31, v17, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v17, v17, v11, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s4 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5
; GFX10-NEXT: v_cndmask_b32_e64 v2, v18, v29, s6 ; GFX10-NEXT: v_lshrrev_b64 v[10:11], v18, v[4:5]
; GFX10-NEXT: v_sub_nc_u32_e32 v18, 64, v24
; GFX10-NEXT: v_lshlrev_b64 v[14:15], v24, v[6:7] ; GFX10-NEXT: v_lshlrev_b64 v[14:15], v24, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e32 v13, v31, v13, vcc_lo ; GFX10-NEXT: v_sub_nc_u32_e32 v20, 64, v25
; GFX10-NEXT: v_cndmask_b32_e32 v12, v19, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v12, v16, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v13, v17, v13, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, v8, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, v8, s4
; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, v9, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, v9, s4
; GFX10-NEXT: v_sub_nc_u32_e32 v31, 64, v27
; GFX10-NEXT: v_lshrrev_b64 v[35:36], v18, v[4:5]
; GFX10-NEXT: v_or_b32_e32 v2, v2, v3 ; GFX10-NEXT: v_or_b32_e32 v2, v2, v3
; GFX10-NEXT: v_subrev_nc_u32_e32 v3, 64, v24 ; GFX10-NEXT: v_subrev_nc_u32_e32 v3, 64, v24
; GFX10-NEXT: v_subrev_nc_u32_e32 v10, 64, v27
; GFX10-NEXT: v_lshlrev_b64 v[16:17], v24, v[4:5] ; GFX10-NEXT: v_lshlrev_b64 v[16:17], v24, v[4:5]
; GFX10-NEXT: v_lshrrev_b64 v[18:19], v27, v[12:13] ; GFX10-NEXT: v_or_b32_e32 v14, v10, v14
; GFX10-NEXT: v_lshlrev_b64 v[20:21], v31, v[8:9] ; GFX10-NEXT: v_subrev_nc_u32_e32 v10, 64, v25
; GFX10-NEXT: v_lshrrev_b64 v[18:19], v25, v[12:13]
; GFX10-NEXT: v_lshlrev_b64 v[20:21], v20, v[8:9]
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v24 ; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v24
; GFX10-NEXT: v_lshlrev_b64 v[3:4], v3, v[4:5] ; GFX10-NEXT: v_lshlrev_b64 v[3:4], v3, v[4:5]
; GFX10-NEXT: v_or_b32_e32 v5, v36, v15 ; GFX10-NEXT: v_or_b32_e32 v5, v11, v15
; GFX10-NEXT: v_or_b32_e32 v14, v35, v14
; GFX10-NEXT: v_lshrrev_b64 v[10:11], v10, v[8:9] ; GFX10-NEXT: v_lshrrev_b64 v[10:11], v10, v[8:9]
; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v25
; GFX10-NEXT: v_cndmask_b32_e32 v15, 0, v16, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v15, 0, v16, vcc_lo
; GFX10-NEXT: v_or_b32_e32 v16, v18, v20 ; GFX10-NEXT: v_or_b32_e32 v16, v18, v20
; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v27
; GFX10-NEXT: v_or_b32_e32 v18, v19, v21 ; GFX10-NEXT: v_or_b32_e32 v18, v19, v21
; GFX10-NEXT: v_cndmask_b32_e32 v31, v3, v14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v14, v3, v14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v5, v4, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v4, v5, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[3:4], v27, v[8:9] ; GFX10-NEXT: v_lshrrev_b64 v[3:4], v25, v[8:9]
; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v24 ; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v24
; GFX10-NEXT: v_cndmask_b32_e64 v8, v11, v18, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v8, v11, v18, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v27 ; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v25
; GFX10-NEXT: v_cndmask_b32_e64 v10, v10, v16, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v10, v10, v16, s4
; GFX10-NEXT: v_cndmask_b32_e32 v9, 0, v17, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v9, 0, v17, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v14, v5, v7, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v7, v5, v7, s6
; GFX10-NEXT: v_cndmask_b32_e64 v11, 0, v4, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v11, 0, v4, s4
; GFX10-NEXT: v_cndmask_b32_e64 v19, v31, v6, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v6, v14, v6, s6
; GFX10-NEXT: v_cndmask_b32_e64 v5, v10, v12, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v10, v12, s5
; GFX10-NEXT: v_cndmask_b32_e64 v10, 0, v3, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v10, 0, v3, s4
; GFX10-NEXT: v_cndmask_b32_e64 v8, v8, v13, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v8, v8, v13, s5
; GFX10-NEXT: v_or_b32_e32 v3, v22, v23 ; GFX10-NEXT: v_or_b32_e32 v3, v22, v23
; GFX10-NEXT: v_or_b32_e32 v7, v14, v11 ; GFX10-NEXT: v_or_b32_e32 v7, v7, v11
; GFX10-NEXT: v_or_b32_e32 v4, v15, v5 ; GFX10-NEXT: v_or_b32_e32 v4, v15, v5
; GFX10-NEXT: v_or_b32_e32 v6, v19, v10 ; GFX10-NEXT: v_or_b32_e32 v6, v6, v10
; GFX10-NEXT: v_or_b32_e32 v5, v9, v8 ; GFX10-NEXT: v_or_b32_e32 v5, v9, v8
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i128> @llvm.fshl.v2i128(<2 x i128> %lhs, <2 x i128> %rhs, <2 x i128> %amt) %result = call <2 x i128> @llvm.fshl.v2i128(<2 x i128> %lhs, <2 x i128> %rhs, <2 x i128> %amt)
ret <2 x i128> %result ret <2 x i128> %result
} }
declare i7 @llvm.fshl.i7(i7, i7, i7) #0 declare i7 @llvm.fshl.i7(i7, i7, i7) #0
declare i8 @llvm.fshl.i8(i8, i8, i8) #0 declare i8 @llvm.fshl.i8(i8, i8, i8) #0
Show All 30 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/fshr.ll

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 254 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 7, v2 ; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 7, v2
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo
; GFX10-NEXT: v_subrev_nc_u32_e32 v3, 7, v2 ; GFX10-NEXT: v_subrev_nc_u32_e32 v3, 7, v2
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 7, v2 ; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 7, v2
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo
; GFX10-NEXT: v_mov_b32_e32 v3, 0x7f ; GFX10-NEXT: v_mov_b32_e32 v3, 0x7f
; GFX10-NEXT: v_sub_nc_u16 v4, 6, v2 ; GFX10-NEXT: v_sub_nc_u16 v4, 6, v2
; GFX10-NEXT: v_and_b32_e32 v2, v2, v3 ; GFX10-NEXT: v_and_b32_e32 v2, v2, v3
; GFX10-NEXT: v_and_b32_e32 v7, v4, v3 ; GFX10-NEXT: v_and_b32_e32 v4, v4, v3
; GFX10-NEXT: v_lshrrev_b16 v1, v2, v1 ; GFX10-NEXT: v_lshrrev_b16 v1, v2, v1
; GFX10-NEXT: v_lshlrev_b16 v0, v7, v0 ; GFX10-NEXT: v_lshlrev_b16 v0, v4, v0
; GFX10-NEXT: v_or_b32_e32 v0, v0, v1 ; GFX10-NEXT: v_or_b32_e32 v0, v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i7 @llvm.fshr.i7(i7 %lhs, i7 %rhs, i7 %amt) %result = call i7 @llvm.fshr.i7(i7 %lhs, i7 %rhs, i7 %amt)
ret i7 %result ret i7 %result
} }
define amdgpu_ps i8 @s_fshr_i8(i8 inreg %lhs, i8 inreg %rhs, i8 inreg %amt) { define amdgpu_ps i8 @s_fshr_i8(i8 inreg %lhs, i8 inreg %rhs, i8 inreg %amt) {
; GFX6-LABEL: s_fshr_i8: ; GFX6-LABEL: s_fshr_i8:
▲ Show 20 Lines • Show All 879 Lines • ▼ Show 20 Lines
; ;
; GFX10-LABEL: v_fshr_v4i8: ; GFX10-LABEL: v_fshr_v4i8:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 8, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 8, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX10-NEXT: v_xor_b32_e32 v8, -1, v2 ; GFX10-NEXT: v_xor_b32_e32 v8, -1, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 16, v2
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v6
; GFX10-NEXT: v_lshlrev_b16 v3, 1, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 16, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 16, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v0
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v6 ; GFX10-NEXT: v_xor_b32_e32 v14, -1, v12
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 16, v2 ; GFX10-NEXT: v_and_b32_e32 v11, 7, v11
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v1
; GFX10-NEXT: v_and_b32_e32 v15, 7, v8 ; GFX10-NEXT: v_and_b32_e32 v8, 7, v8
; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0 ; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0
; GFX10-NEXT: v_and_b32_e32 v14, 7, v11 ; GFX10-NEXT: v_mov_b32_e32 v13, 0xff
; GFX10-NEXT: v_lshlrev_b16 v3, 1, v3 ; GFX10-NEXT: v_lshlrev_b16 v3, v11, v3
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v10 ; GFX10-NEXT: v_xor_b32_e32 v11, -1, v10
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v1
; GFX10-NEXT: v_lshlrev_b16 v0, v15, v0
; GFX10-NEXT: v_mov_b32_e32 v15, 0xff
; GFX10-NEXT: v_lshlrev_b16 v3, v14, v3
; GFX10-NEXT: v_xor_b32_e32 v14, -1, v12
; GFX10-NEXT: s_movk_i32 s4, 0xff ; GFX10-NEXT: s_movk_i32 s4, 0xff
; GFX10-NEXT: v_lshlrev_b16 v0, v8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v9, 24, v1
; GFX10-NEXT: v_and_b32_e32 v8, s4, v1 ; GFX10-NEXT: v_and_b32_e32 v8, s4, v1
; GFX10-NEXT: v_and_b32_sdwa v1, v1, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v1, v1, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_e32 v10, 7, v10 ; GFX10-NEXT: v_and_b32_e32 v10, 7, v10
; GFX10-NEXT: v_and_b32_e32 v6, 7, v6 ; GFX10-NEXT: v_and_b32_e32 v6, 7, v6
; GFX10-NEXT: v_and_b32_e32 v7, s4, v7 ; GFX10-NEXT: v_and_b32_e32 v7, s4, v7
; GFX10-NEXT: v_and_b32_e32 v11, 7, v11 ; GFX10-NEXT: v_and_b32_e32 v11, 7, v11
; GFX10-NEXT: v_lshlrev_b16 v4, 1, v4 ; GFX10-NEXT: v_lshlrev_b16 v4, 1, v4
; GFX10-NEXT: v_and_b32_e32 v15, 7, v14 ; GFX10-NEXT: v_and_b32_e32 v13, 7, v14
; GFX10-NEXT: v_lshlrev_b16 v5, 1, v5 ; GFX10-NEXT: v_lshlrev_b16 v5, 1, v5
; GFX10-NEXT: v_and_b32_e32 v12, 7, v12 ; GFX10-NEXT: v_and_b32_e32 v12, 7, v12
; GFX10-NEXT: v_and_b32_e32 v2, 7, v2 ; GFX10-NEXT: v_and_b32_e32 v2, 7, v2
; GFX10-NEXT: v_lshrrev_b16 v6, v6, v7 ; GFX10-NEXT: v_lshrrev_b16 v6, v6, v7
; GFX10-NEXT: v_lshlrev_b16 v4, v11, v4 ; GFX10-NEXT: v_lshlrev_b16 v4, v11, v4
; GFX10-NEXT: v_lshrrev_b16 v1, v10, v1 ; GFX10-NEXT: v_lshrrev_b16 v1, v10, v1
; GFX10-NEXT: v_lshlrev_b16 v5, v15, v5 ; GFX10-NEXT: v_lshlrev_b16 v5, v13, v5
; GFX10-NEXT: v_lshrrev_b16 v7, v12, v9 ; GFX10-NEXT: v_lshrrev_b16 v7, v12, v9
; GFX10-NEXT: v_lshrrev_b16 v2, v2, v8 ; GFX10-NEXT: v_lshrrev_b16 v2, v2, v8
; GFX10-NEXT: v_or_b32_e32 v3, v3, v6 ; GFX10-NEXT: v_or_b32_e32 v3, v3, v6
; GFX10-NEXT: v_or_b32_e32 v1, v4, v1 ; GFX10-NEXT: v_or_b32_e32 v1, v4, v1
; GFX10-NEXT: v_mov_b32_e32 v6, 8 ; GFX10-NEXT: v_mov_b32_e32 v6, 8
; GFX10-NEXT: v_or_b32_e32 v4, v5, v7 ; GFX10-NEXT: v_or_b32_e32 v4, v5, v7
; GFX10-NEXT: v_or_b32_e32 v0, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v0, v0, v2
; GFX10-NEXT: v_and_b32_e32 v1, s4, v1 ; GFX10-NEXT: v_and_b32_e32 v1, s4, v1
▲ Show 20 Lines • Show All 984 Lines • ▼ Show 20 Lines
; ;
; GFX10-LABEL: v_fshr_v2i24: ; GFX10-LABEL: v_fshr_v2i24:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v6, 24 ; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v6, 24
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v7, 24 ; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v7, 24
; GFX10-NEXT: s_sub_i32 s4, 0, 24 ; GFX10-NEXT: s_sub_i32 s4, 0, 24
; GFX10-NEXT: v_mov_b32_e32 v12, 0xffffff ; GFX10-NEXT: v_mov_b32_e32 v10, 0xffffff
; GFX10-NEXT: v_and_b32_e32 v4, 0xffffff, v4 ; GFX10-NEXT: v_and_b32_e32 v4, 0xffffff, v4
; GFX10-NEXT: v_rcp_iflag_f32_e32 v6, v6 ; GFX10-NEXT: v_rcp_iflag_f32_e32 v6, v6
; GFX10-NEXT: v_rcp_iflag_f32_e32 v7, v7 ; GFX10-NEXT: v_rcp_iflag_f32_e32 v7, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 1, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX10-NEXT: v_and_b32_e32 v5, v5, v12 ; GFX10-NEXT: v_and_b32_e32 v5, v5, v10
; GFX10-NEXT: v_and_b32_e32 v2, v2, v12 ; GFX10-NEXT: v_and_b32_e32 v2, v2, v10
; GFX10-NEXT: v_and_b32_e32 v3, v3, v12 ; GFX10-NEXT: v_and_b32_e32 v3, v3, v10
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 1, v1 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 1, v1
; GFX10-NEXT: v_mul_f32_e32 v6, 0x4f7ffffe, v6 ; GFX10-NEXT: v_mul_f32_e32 v6, 0x4f7ffffe, v6
; GFX10-NEXT: v_mul_f32_e32 v7, 0x4f7ffffe, v7 ; GFX10-NEXT: v_mul_f32_e32 v7, 0x4f7ffffe, v7
; GFX10-NEXT: v_cvt_u32_f32_e32 v6, v6 ; GFX10-NEXT: v_cvt_u32_f32_e32 v6, v6
; GFX10-NEXT: v_cvt_u32_f32_e32 v7, v7 ; GFX10-NEXT: v_cvt_u32_f32_e32 v7, v7
; GFX10-NEXT: v_mul_lo_u32 v8, s4, v6 ; GFX10-NEXT: v_mul_lo_u32 v8, s4, v6
; GFX10-NEXT: v_mul_lo_u32 v9, s4, v7 ; GFX10-NEXT: v_mul_lo_u32 v9, s4, v7
; GFX10-NEXT: v_mul_hi_u32 v8, v6, v8 ; GFX10-NEXT: v_mul_hi_u32 v8, v6, v8
Show All 10 Lines
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v4 ; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v4
; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 24, v5 ; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 24, v5
; GFX10-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v5 ; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v5
; GFX10-NEXT: v_subrev_nc_u32_e32 v6, 24, v4 ; GFX10-NEXT: v_subrev_nc_u32_e32 v6, 24, v4
; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v4 ; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v4
; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 24, v5 ; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 24, v5
; GFX10-NEXT: v_cndmask_b32_e32 v11, v4, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v5 ; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v5
; GFX10-NEXT: v_and_b32_e32 v4, v11, v12 ; GFX10-NEXT: v_sub_nc_u32_e32 v6, 23, v4
; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc_lo
; GFX10-NEXT: v_sub_nc_u32_e32 v6, 23, v11 ; GFX10-NEXT: v_and_b32_e32 v4, v4, v10
; GFX10-NEXT: v_lshrrev_b32_e32 v2, v4, v2 ; GFX10-NEXT: v_and_b32_e32 v6, v6, v10
; GFX10-NEXT: v_sub_nc_u32_e32 v7, 23, v5 ; GFX10-NEXT: v_sub_nc_u32_e32 v7, 23, v5
; GFX10-NEXT: v_and_b32_e32 v5, v5, v12 ; GFX10-NEXT: v_and_b32_e32 v5, v5, v10
; GFX10-NEXT: v_and_b32_e32 v11, v6, v12 ; GFX10-NEXT: v_lshrrev_b32_e32 v2, v4, v2
; GFX10-NEXT: v_and_b32_e32 v4, v7, v12 ; GFX10-NEXT: v_and_b32_e32 v4, v7, v10
; GFX10-NEXT: v_lshrrev_b32_e32 v3, v5, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, v5, v3
; GFX10-NEXT: v_lshl_or_b32 v0, v0, v11, v2 ; GFX10-NEXT: v_lshl_or_b32 v0, v0, v6, v2
; GFX10-NEXT: v_lshl_or_b32 v1, v1, v4, v3 ; GFX10-NEXT: v_lshl_or_b32 v1, v1, v4, v3
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i24> @llvm.fshr.v2i24(<2 x i24> %lhs, <2 x i24> %rhs, <2 x i24> %amt) %result = call <2 x i24> @llvm.fshr.v2i24(<2 x i24> %lhs, <2 x i24> %rhs, <2 x i24> %amt)
ret <2 x i24> %result ret <2 x i24> %result
} }
define amdgpu_ps i32 @s_fshr_i32(i32 inreg %lhs, i32 inreg %rhs, i32 inreg %amt) { define amdgpu_ps i32 @s_fshr_i32(i32 inreg %lhs, i32 inreg %rhs, i32 inreg %amt) {
; GFX6-LABEL: s_fshr_i32: ; GFX6-LABEL: s_fshr_i32:
▲ Show 20 Lines • Show All 2,172 Lines • ▼ Show 20 Lines
; ;
; GFX10-LABEL: v_fshr_i64: ; GFX10-LABEL: v_fshr_i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v5, -1, v4 ; GFX10-NEXT: v_xor_b32_e32 v5, -1, v4
; GFX10-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX10-NEXT: v_and_b32_e32 v4, 63, v4 ; GFX10-NEXT: v_and_b32_e32 v4, 63, v4
; GFX10-NEXT: v_and_b32_e32 v7, 63, v5 ; GFX10-NEXT: v_and_b32_e32 v5, 63, v5
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v4, v[2:3] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], v4, v[2:3]
; GFX10-NEXT: v_lshlrev_b64 v[0:1], v7, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], v5, v[0:1]
; GFX10-NEXT: v_or_b32_e32 v0, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v0, v0, v2
; GFX10-NEXT: v_or_b32_e32 v1, v1, v3 ; GFX10-NEXT: v_or_b32_e32 v1, v1, v3
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 %amt) %result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 %amt)
ret i64 %result ret i64 %result
} }
define i64 @v_fshr_i64_5(i64 %lhs, i64 %rhs) { define i64 @v_fshr_i64_5(i64 %lhs, i64 %rhs) {
▲ Show 20 Lines • Show All 390 Lines • ▼ Show 20 Lines
; GFX10-LABEL: v_fshr_v2i64: ; GFX10-LABEL: v_fshr_v2i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v9, -1, v8 ; GFX10-NEXT: v_xor_b32_e32 v9, -1, v8
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v10 ; GFX10-NEXT: v_xor_b32_e32 v11, -1, v10
; GFX10-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], 1, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[2:3], 1, v[2:3]
; GFX10-NEXT: v_and_b32_e32 v19, 63, v8 ; GFX10-NEXT: v_and_b32_e32 v8, 63, v8
; GFX10-NEXT: v_and_b32_e32 v15, 63, v9 ; GFX10-NEXT: v_and_b32_e32 v9, 63, v9
; GFX10-NEXT: v_and_b32_e32 v9, 63, v11 ; GFX10-NEXT: v_and_b32_e32 v11, 63, v11
; GFX10-NEXT: v_and_b32_e32 v13, 63, v10 ; GFX10-NEXT: v_and_b32_e32 v10, 63, v10
; GFX10-NEXT: v_lshrrev_b64 v[4:5], v19, v[4:5] ; GFX10-NEXT: v_lshrrev_b64 v[4:5], v8, v[4:5]
; GFX10-NEXT: v_lshlrev_b64 v[11:12], v15, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], v9, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[15:16], v9, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[2:3], v11, v[2:3]
; GFX10-NEXT: v_lshrrev_b64 v[6:7], v13, v[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[6:7], v10, v[6:7]
; GFX10-NEXT: v_or_b32_e32 v0, v11, v4 ; GFX10-NEXT: v_or_b32_e32 v0, v0, v4
; GFX10-NEXT: v_or_b32_e32 v1, v12, v5 ; GFX10-NEXT: v_or_b32_e32 v1, v1, v5
; GFX10-NEXT: v_or_b32_e32 v2, v15, v6 ; GFX10-NEXT: v_or_b32_e32 v2, v2, v6
; GFX10-NEXT: v_or_b32_e32 v3, v16, v7 ; GFX10-NEXT: v_or_b32_e32 v3, v3, v7
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %lhs, <2 x i64> %rhs, <2 x i64> %amt) %result = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %lhs, <2 x i64> %rhs, <2 x i64> %amt)
ret <2 x i64> %result ret <2 x i64> %result
} }
define amdgpu_ps i128 @s_fshr_i128(i128 inreg %lhs, i128 inreg %rhs, i128 inreg %amt) { define amdgpu_ps i128 @s_fshr_i128(i128 inreg %lhs, i128 inreg %rhs, i128 inreg %amt) {
; GFX6-LABEL: s_fshr_i128: ; GFX6-LABEL: s_fshr_i128:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 456 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_or_b32_e32 v10, v10, v12 ; GFX10-NEXT: v_or_b32_e32 v10, v10, v12
; GFX10-NEXT: v_xor_b32_e32 v15, -1, v8 ; GFX10-NEXT: v_xor_b32_e32 v15, -1, v8
; GFX10-NEXT: s_movk_i32 s5, 0x7f ; GFX10-NEXT: s_movk_i32 s5, 0x7f
; GFX10-NEXT: s_and_b32 s6, 1, s7 ; GFX10-NEXT: s_and_b32 s6, 1, s7
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v11, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v11, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v10, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v10, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s6 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s6
; GFX10-NEXT: v_and_b32_e32 v19, s5, v15 ; GFX10-NEXT: v_and_b32_e32 v19, s5, v15
; GFX10-NEXT: v_and_b32_e32 v20, s5, v8
; GFX10-NEXT: v_cndmask_b32_e32 v9, 0, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v9, 0, v13, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v10, 0, v14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, 0, v14, vcc_lo
; GFX10-NEXT: v_and_b32_e32 v20, s5, v8
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v3, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v3, s4
; GFX10-NEXT: v_sub_nc_u32_e32 v11, 64, v19 ; GFX10-NEXT: v_sub_nc_u32_e32 v2, 64, v19
; GFX10-NEXT: v_sub_nc_u32_e32 v17, 64, v20
; GFX10-NEXT: v_mov_b32_e32 v25, v4
; GFX10-NEXT: v_mov_b32_e32 v26, v5
; GFX10-NEXT: v_subrev_nc_u32_e32 v8, 64, v19 ; GFX10-NEXT: v_subrev_nc_u32_e32 v8, 64, v19
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v11, v[9:10] ; GFX10-NEXT: v_sub_nc_u32_e32 v17, 64, v20
; GFX10-NEXT: v_lshlrev_b64 v[11:12], v19, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[13:14], v19, v[9:10] ; GFX10-NEXT: v_lshlrev_b64 v[13:14], v19, v[9:10]
; GFX10-NEXT: v_subrev_nc_u32_e32 v21, 64, v20 ; GFX10-NEXT: v_lshlrev_b64 v[11:12], v19, v[0:1]
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v2, v[9:10]
; GFX10-NEXT: v_lshlrev_b64 v[8:9], v8, v[9:10] ; GFX10-NEXT: v_lshlrev_b64 v[8:9], v8, v[9:10]
; GFX10-NEXT: v_lshrrev_b64 v[15:16], v20, v[25:26] ; GFX10-NEXT: v_lshrrev_b64 v[15:16], v20, v[4:5]
; GFX10-NEXT: v_lshlrev_b64 v[17:18], v17, v[6:7] ; GFX10-NEXT: v_lshlrev_b64 v[17:18], v17, v[6:7]
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v19 ; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v19
; GFX10-NEXT: v_or_b32_e32 v10, v3, v12
; GFX10-NEXT: v_or_b32_e32 v11, v2, v11
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v21, v[6:7]
; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v20 ; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v20
; GFX10-NEXT: v_cndmask_b32_e32 v23, 0, v13, vcc_lo ; GFX10-NEXT: v_or_b32_e32 v11, v2, v11
; GFX10-NEXT: v_subrev_nc_u32_e32 v2, 64, v20
; GFX10-NEXT: v_or_b32_e32 v10, v3, v12
; GFX10-NEXT: v_cndmask_b32_e32 v12, 0, v13, vcc_lo
; GFX10-NEXT: v_or_b32_e32 v13, v15, v17 ; GFX10-NEXT: v_or_b32_e32 v13, v15, v17
; GFX10-NEXT: v_cndmask_b32_e32 v15, v9, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v8, v8, v11, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v2, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e32 v9, v9, v10, vcc_lo
; GFX10-NEXT: v_or_b32_e32 v10, v16, v18 ; GFX10-NEXT: v_or_b32_e32 v10, v16, v18
; GFX10-NEXT: v_lshrrev_b64 v[6:7], v20, v[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[6:7], v20, v[6:7]
; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v19 ; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v19
; GFX10-NEXT: v_cndmask_b32_e32 v11, v8, v11, vcc_lo ; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v20
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v13, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v13, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v10, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v10, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v20
; GFX10-NEXT: v_cndmask_b32_e64 v9, v15, v1, s6
; GFX10-NEXT: v_cndmask_b32_e64 v8, v11, v0, s6
; GFX10-NEXT: v_cndmask_b32_e32 v10, 0, v14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, 0, v14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v2, v25, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v8, v8, v0, s6
; GFX10-NEXT: v_cndmask_b32_e64 v1, v3, v26, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v9, v9, v1, s6
; GFX10-NEXT: v_cndmask_b32_e64 v0, v2, v4, s5
; GFX10-NEXT: v_cndmask_b32_e64 v1, v3, v5, s5
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v6, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v6, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v7, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v7, s4
; GFX10-NEXT: v_or_b32_e32 v0, v23, v0 ; GFX10-NEXT: v_or_b32_e32 v0, v12, v0
; GFX10-NEXT: v_or_b32_e32 v1, v10, v1 ; GFX10-NEXT: v_or_b32_e32 v1, v10, v1
; GFX10-NEXT: v_or_b32_e32 v2, v8, v2 ; GFX10-NEXT: v_or_b32_e32 v2, v8, v2
; GFX10-NEXT: v_or_b32_e32 v3, v9, v3 ; GFX10-NEXT: v_or_b32_e32 v3, v9, v3
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 %amt) %result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 %amt)
ret i128 %result ret i128 %result
} }
▲ Show 20 Lines • Show All 218 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshlrev_b64 v[2:3], v13, s[8:9] ; GFX10-NEXT: v_lshlrev_b64 v[2:3], v13, s[8:9]
; GFX10-NEXT: v_subrev_nc_u32_e32 v10, 64, v13 ; GFX10-NEXT: v_subrev_nc_u32_e32 v10, 64, v13
; GFX10-NEXT: v_lshrrev_b64 v[6:7], v12, s[4:5] ; GFX10-NEXT: v_lshrrev_b64 v[6:7], v12, s[4:5]
; GFX10-NEXT: v_lshlrev_b64 v[8:9], v8, s[6:7] ; GFX10-NEXT: v_lshlrev_b64 v[8:9], v8, s[6:7]
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v13 ; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v13
; GFX10-NEXT: v_cmp_gt_u32_e64 s0, 64, v12 ; GFX10-NEXT: v_cmp_gt_u32_e64 s0, 64, v12
; GFX10-NEXT: v_or_b32_e32 v2, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v2, v0, v2
; GFX10-NEXT: v_subrev_nc_u32_e32 v0, 64, v12 ; GFX10-NEXT: v_subrev_nc_u32_e32 v0, 64, v12
; GFX10-NEXT: v_lshlrev_b64 v[15:16], v10, s[10:11] ; GFX10-NEXT: v_lshlrev_b64 v[10:11], v10, s[10:11]
; GFX10-NEXT: v_or_b32_e32 v3, v1, v3 ; GFX10-NEXT: v_or_b32_e32 v3, v1, v3
; GFX10-NEXT: v_or_b32_e32 v6, v6, v8 ; GFX10-NEXT: v_or_b32_e32 v6, v6, v8
; GFX10-NEXT: v_or_b32_e32 v7, v7, v9 ; GFX10-NEXT: v_or_b32_e32 v7, v7, v9
; GFX10-NEXT: v_lshrrev_b64 v[0:1], v0, s[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[0:1], v0, s[6:7]
; GFX10-NEXT: v_lshlrev_b64 v[4:5], v13, s[10:11] ; GFX10-NEXT: v_lshlrev_b64 v[4:5], v13, s[10:11]
; GFX10-NEXT: v_cndmask_b32_e32 v8, v15, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v8, v10, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v10, v16, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, v11, v3, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v12, s[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], v12, s[6:7]
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v13
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v6, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 0, v12 ; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 0, v12
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v6, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v7, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v7, s0
; GFX10-NEXT: v_cndmask_b32_e32 v11, 0, v4, vcc_lo ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v13
; GFX10-NEXT: v_cndmask_b32_e32 v15, 0, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, 0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v19, v8, s8, s2 ; GFX10-NEXT: v_cndmask_b32_e32 v5, 0, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, s4, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, s4, s1
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s5, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s5, s1
; GFX10-NEXT: v_cndmask_b32_e64 v6, v8, s8, s2
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v6, v10, s9, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v7, v10, s9, s2
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0
; GFX10-NEXT: v_or_b32_e32 v0, v11, v0 ; GFX10-NEXT: v_or_b32_e32 v0, v4, v0
; GFX10-NEXT: v_or_b32_e32 v1, v15, v1 ; GFX10-NEXT: v_or_b32_e32 v1, v5, v1
; GFX10-NEXT: v_or_b32_e32 v2, v19, v2 ; GFX10-NEXT: v_or_b32_e32 v2, v6, v2
; GFX10-NEXT: v_or_b32_e32 v3, v6, v3 ; GFX10-NEXT: v_or_b32_e32 v3, v7, v3
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 %amt) %result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 %amt)
%cast.result = bitcast i128 %result to <4 x float> %cast.result = bitcast i128 %result to <4 x float>
ret <4 x float> %cast.result ret <4 x float> %cast.result
} }
define amdgpu_ps <4 x float> @v_fshr_i128_svs(i128 inreg %lhs, i128 %rhs, i128 inreg %amt) { define amdgpu_ps <4 x float> @v_fshr_i128_svs(i128 inreg %lhs, i128 %rhs, i128 inreg %amt) {
; GFX6-LABEL: v_fshr_i128_svs: ; GFX6-LABEL: v_fshr_i128_svs:
▲ Show 20 Lines • Show All 238 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_cselect_b64 s[4:5], s[4:5], 0 ; GFX10-NEXT: s_cselect_b64 s[4:5], s[4:5], 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[6:7] ; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[6:7]
; GFX10-NEXT: s_cmp_lg_u32 s13, 0 ; GFX10-NEXT: s_cmp_lg_u32 s13, 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[0:1], s[2:3] ; GFX10-NEXT: s_cselect_b64 s[2:3], s[0:1], s[2:3]
; GFX10-NEXT: s_sub_i32 s0, 64, s8 ; GFX10-NEXT: s_sub_i32 s0, 64, s8
; GFX10-NEXT: v_lshlrev_b64 v[6:7], s0, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[6:7], s0, v[2:3]
; GFX10-NEXT: s_sub_i32 s0, s8, 64 ; GFX10-NEXT: s_sub_i32 s0, s8, 64
; GFX10-NEXT: s_cmp_lt_u32 s8, 64 ; GFX10-NEXT: s_cmp_lt_u32 s8, 64
; GFX10-NEXT: v_lshrrev_b64 v[11:12], s0, v[2:3] ; GFX10-NEXT: v_lshrrev_b64 v[8:9], s0, v[2:3]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s8, 0 ; GFX10-NEXT: s_cmp_eq_u32 s8, 0
; GFX10-NEXT: v_or_b32_e32 v4, v4, v6 ; GFX10-NEXT: v_or_b32_e32 v4, v4, v6
; GFX10-NEXT: v_or_b32_e32 v5, v5, v7 ; GFX10-NEXT: v_or_b32_e32 v5, v5, v7
; GFX10-NEXT: s_cselect_b32 s0, 1, 0 ; GFX10-NEXT: s_cselect_b32 s0, 1, 0
; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo
; GFX10-NEXT: s_and_b32 s0, 1, s0 ; GFX10-NEXT: s_and_b32 s0, 1, s0
; GFX10-NEXT: v_lshrrev_b64 v[2:3], s8, v[2:3] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], s8, v[2:3]
; GFX10-NEXT: v_cndmask_b32_e32 v7, v11, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, v8, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v11, v12, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v9, v5, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1 ; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1
; GFX10-NEXT: v_cndmask_b32_e32 v0, v7, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v11, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0
; GFX10-NEXT: v_or_b32_e32 v0, s4, v0 ; GFX10-NEXT: v_or_b32_e32 v0, s4, v0
; GFX10-NEXT: v_or_b32_e32 v1, s5, v1 ; GFX10-NEXT: v_or_b32_e32 v1, s5, v1
; GFX10-NEXT: v_or_b32_e32 v2, s2, v2 ; GFX10-NEXT: v_or_b32_e32 v2, s2, v2
; GFX10-NEXT: v_or_b32_e32 v3, s3, v3 ; GFX10-NEXT: v_or_b32_e32 v3, s3, v3
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 %amt) %result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 %amt)
▲ Show 20 Lines • Show All 227 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_mov_b32 s7, 0 ; GFX10-NEXT: s_mov_b32 s7, 0
; GFX10-NEXT: v_lshlrev_b64 v[6:7], 1, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[6:7], 1, v[2:3]
; GFX10-NEXT: s_and_b64 s[8:9], s[4:5], s[6:7] ; GFX10-NEXT: s_and_b64 s[8:9], s[4:5], s[6:7]
; GFX10-NEXT: s_andn2_b64 s[6:7], s[6:7], s[4:5] ; GFX10-NEXT: s_andn2_b64 s[6:7], s[6:7], s[4:5]
; GFX10-NEXT: s_sub_i32 s4, 64, 1 ; GFX10-NEXT: s_sub_i32 s4, 64, 1
; GFX10-NEXT: s_sub_i32 s5, 1, 64 ; GFX10-NEXT: s_sub_i32 s5, 1, 64
; GFX10-NEXT: v_lshrrev_b64 v[4:5], s4, v[0:1] ; GFX10-NEXT: v_lshrrev_b64 v[4:5], s4, v[0:1]
; GFX10-NEXT: s_cmp_lt_u32 1, 64 ; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_lshlrev_b64 v[13:14], s5, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[8:9], 1, v[0:1]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0 ; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[8:9], 1, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], s5, v[0:1]
; GFX10-NEXT: s_cselect_b32 s7, 1, 0 ; GFX10-NEXT: s_cselect_b32 s7, 1, 0
; GFX10-NEXT: s_and_b32 s4, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s4, 1, vcc_lo
; GFX10-NEXT: v_or_b32_e32 v6, v4, v6 ; GFX10-NEXT: v_or_b32_e32 v6, v4, v6
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4
; GFX10-NEXT: v_or_b32_e32 v5, v5, v7 ; GFX10-NEXT: v_or_b32_e32 v5, v5, v7
; GFX10-NEXT: s_and_b32 s5, 1, s7 ; GFX10-NEXT: s_and_b32 s5, 1, s7
; GFX10-NEXT: v_cndmask_b32_e32 v4, 0, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, 0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v13, v6, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v6, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v14, v5, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v5, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5
; GFX10-NEXT: v_cndmask_b32_e32 v5, 0, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, 0, v9, vcc_lo
; GFX10-NEXT: s_sub_i32 s5, s6, 64 ; GFX10-NEXT: s_sub_i32 s5, s6, 64
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v3, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v3, s4
; GFX10-NEXT: s_sub_i32 s4, 64, s6 ; GFX10-NEXT: s_sub_i32 s4, 64, s6
; GFX10-NEXT: s_cmp_lt_u32 s6, 64 ; GFX10-NEXT: s_cmp_lt_u32 s6, 64
; GFX10-NEXT: v_lshrrev_b64 v[11:12], s4, v[4:5] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], s4, v[4:5]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[6:7], s6, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[6:7], s6, v[0:1]
; GFX10-NEXT: s_cmp_eq_u32 s6, 0 ; GFX10-NEXT: s_cmp_eq_u32 s6, 0
; GFX10-NEXT: v_lshlrev_b64 v[8:9], s6, v[4:5] ; GFX10-NEXT: v_lshlrev_b64 v[8:9], s6, v[4:5]
; GFX10-NEXT: s_cselect_b32 s6, 1, 0 ; GFX10-NEXT: s_cselect_b32 s6, 1, 0
; GFX10-NEXT: s_and_b32 s4, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s4, 1, vcc_lo
; GFX10-NEXT: v_lshlrev_b64 v[4:5], s5, v[4:5] ; GFX10-NEXT: v_lshlrev_b64 v[4:5], s5, v[4:5]
; GFX10-NEXT: v_or_b32_e32 v2, v11, v6 ; GFX10-NEXT: v_or_b32_e32 v2, v2, v6
; GFX10-NEXT: v_or_b32_e32 v3, v12, v7 ; GFX10-NEXT: v_or_b32_e32 v3, v3, v7
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4
; GFX10-NEXT: s_sub_i32 s10, s8, 64 ; GFX10-NEXT: s_sub_i32 s10, s8, 64
; GFX10-NEXT: v_cndmask_b32_e32 v6, 0, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v6, 0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, 0, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v7, 0, v9, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v4, v2, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v4, v2, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, v5, v3, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v5, v3, s4
; GFX10-NEXT: s_and_b32 s4, 1, s6 ; GFX10-NEXT: s_and_b32 s4, 1, s6
; GFX10-NEXT: s_sub_i32 s6, 64, s8 ; GFX10-NEXT: s_sub_i32 s6, 64, s8
▲ Show 20 Lines • Show All 364 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_sub_i32 s4, 64, 63 ; GFX10-NEXT: s_sub_i32 s4, 64, 63
; GFX10-NEXT: s_sub_i32 s5, 63, 64 ; GFX10-NEXT: s_sub_i32 s5, 63, 64
; GFX10-NEXT: s_cmp_lt_u32 63, 64 ; GFX10-NEXT: s_cmp_lt_u32 63, 64
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 31, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v11, 31, v0
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 63, 0 ; GFX10-NEXT: s_cmp_eq_u32 63, 0
; GFX10-NEXT: v_lshrrev_b64 v[8:9], s4, v[0:1] ; GFX10-NEXT: v_lshrrev_b64 v[8:9], s4, v[0:1]
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[14:15], s5, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], s5, v[0:1]
; GFX10-NEXT: s_and_b32 s5, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s5, 1, vcc_lo
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 31, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v10, 31, v2
; GFX10-NEXT: v_cndmask_b32_e32 v23, 0, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v13, 0, v11, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s5 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s5
; GFX10-NEXT: s_movk_i32 s6, 0x41 ; GFX10-NEXT: s_movk_i32 s6, 0x41
; GFX10-NEXT: s_and_b32 s4, 1, s4 ; GFX10-NEXT: s_and_b32 s4, 1, s4
; GFX10-NEXT: s_sub_i32 s5, 64, s6 ; GFX10-NEXT: s_sub_i32 s5, 64, s6
; GFX10-NEXT: v_or_b32_e32 v12, v9, v10 ; GFX10-NEXT: v_or_b32_e32 v12, v9, v10
; GFX10-NEXT: v_cndmask_b32_e32 v19, v14, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v14, v0, v8, vcc_lo
; GFX10-NEXT: v_lshlrev_b64 v[10:11], s5, v[6:7] ; GFX10-NEXT: v_lshlrev_b64 v[10:11], s5, v[6:7]
; GFX10-NEXT: v_lshrrev_b64 v[8:9], s6, v[4:5] ; GFX10-NEXT: v_lshrrev_b64 v[8:9], s6, v[4:5]
; GFX10-NEXT: s_sub_i32 s5, s6, 64 ; GFX10-NEXT: s_sub_i32 s5, s6, 64
; GFX10-NEXT: s_cmp_lt_u32 s6, 64 ; GFX10-NEXT: s_cmp_lt_u32 s6, 64
; GFX10-NEXT: v_cndmask_b32_e32 v12, v15, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v12, v1, v12, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[15:16], s5, v[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[0:1], s5, v[6:7]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s6, 0 ; GFX10-NEXT: s_cmp_eq_u32 s6, 0
; GFX10-NEXT: v_or_b32_e32 v8, v8, v10 ; GFX10-NEXT: v_or_b32_e32 v8, v8, v10
; GFX10-NEXT: v_or_b32_e32 v9, v9, v11 ; GFX10-NEXT: v_or_b32_e32 v9, v9, v11
; GFX10-NEXT: s_cselect_b32 s5, 1, 0 ; GFX10-NEXT: s_cselect_b32 s5, 1, 0
; GFX10-NEXT: v_lshrrev_b64 v[6:7], s6, v[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[6:7], s6, v[6:7]
; GFX10-NEXT: s_and_b32 s5, 1, s5 ; GFX10-NEXT: s_and_b32 s5, 1, s5
; GFX10-NEXT: s_and_b32 s6, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s6, 1, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4
; GFX10-NEXT: v_cndmask_b32_e32 v11, v15, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v15, v16, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, s5 ; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, s5
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s6 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s6
; GFX10-NEXT: v_cndmask_b32_e64 v2, v19, v2, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v14, v2, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, v12, v3, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v12, v3, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v15, v5, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v5, s5
; GFX10-NEXT: v_cndmask_b32_e64 v0, v11, v4, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v4, s5
; GFX10-NEXT: v_cndmask_b32_e32 v4, 0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v4, 0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v5, 0, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, 0, v7, vcc_lo
; GFX10-NEXT: v_or_b32_e32 v1, v23, v1 ; GFX10-NEXT: v_or_b32_e32 v1, v13, v1
; GFX10-NEXT: v_or_b32_e32 v2, v2, v4 ; GFX10-NEXT: v_or_b32_e32 v2, v2, v4
; GFX10-NEXT: v_or_b32_e32 v3, v3, v5 ; GFX10-NEXT: v_or_b32_e32 v3, v3, v5
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 65) %result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 65)
ret i128 %result ret i128 %result
} }
define amdgpu_ps <2 x i128> @s_fshr_v2i128(<2 x i128> inreg %lhs, <2 x i128> inreg %rhs, <2 x i128> inreg %amt) { define amdgpu_ps <2 x i128> @s_fshr_v2i128(<2 x i128> inreg %lhs, <2 x i128> inreg %rhs, <2 x i128> inreg %amt) {
▲ Show 20 Lines • Show All 331 Lines • ▼ Show 20 Lines
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_movk_i32 s18, 0x7f ; GFX10-NEXT: s_movk_i32 s18, 0x7f
; GFX10-NEXT: s_mov_b32 s19, 0 ; GFX10-NEXT: s_mov_b32 s19, 0
; GFX10-NEXT: s_sub_i32 s30, 1, 64 ; GFX10-NEXT: s_sub_i32 s30, 1, 64
; GFX10-NEXT: s_and_b64 s[22:23], s[16:17], s[18:19] ; GFX10-NEXT: s_and_b64 s[22:23], s[16:17], s[18:19]
; GFX10-NEXT: s_andn2_b64 s[16:17], s[18:19], s[16:17] ; GFX10-NEXT: s_andn2_b64 s[16:17], s[18:19], s[16:17]
; GFX10-NEXT: s_sub_i32 s31, 64, 1 ; GFX10-NEXT: s_sub_i32 s31, 64, 1
; GFX10-NEXT: s_cmp_lt_u32 1, 64 ; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: s_mov_b32 s62, s10
; GFX10-NEXT: s_cselect_b32 s17, 1, 0 ; GFX10-NEXT: s_cselect_b32 s17, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0 ; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: s_mov_b32 s63, s11
; GFX10-NEXT: s_cselect_b32 s23, 1, 0 ; GFX10-NEXT: s_cselect_b32 s23, 1, 0
; GFX10-NEXT: s_lshr_b64 s[24:25], s[0:1], s31 ; GFX10-NEXT: s_lshr_b64 s[24:25], s[0:1], s31
; GFX10-NEXT: s_lshl_b64 s[26:27], s[2:3], 1 ; GFX10-NEXT: s_lshl_b64 s[26:27], s[2:3], 1
; GFX10-NEXT: s_lshl_b64 s[28:29], s[0:1], 1 ; GFX10-NEXT: s_lshl_b64 s[28:29], s[0:1], 1
; GFX10-NEXT: s_or_b64 s[24:25], s[24:25], s[26:27] ; GFX10-NEXT: s_or_b64 s[24:25], s[24:25], s[26:27]
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], s30 ; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], s30
; GFX10-NEXT: s_cmp_lg_u32 s17, 0 ; GFX10-NEXT: s_cmp_lg_u32 s17, 0
; GFX10-NEXT: s_cselect_b64 s[26:27], s[28:29], 0 ; GFX10-NEXT: s_cselect_b64 s[26:27], s[28:29], 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[24:25], s[0:1] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[24:25], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s23, 0 ; GFX10-NEXT: s_cmp_lg_u32 s23, 0
; GFX10-NEXT: s_cselect_b64 s[46:47], s[2:3], s[0:1] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX10-NEXT: s_sub_i32 s23, s16, 64 ; GFX10-NEXT: s_sub_i32 s23, s16, 64
; GFX10-NEXT: s_sub_i32 s2, 64, s16 ; GFX10-NEXT: s_sub_i32 s2, 64, s16
; GFX10-NEXT: s_cmp_lt_u32 s16, 64 ; GFX10-NEXT: s_cmp_lt_u32 s16, 64
; GFX10-NEXT: s_cselect_b32 s28, 1, 0 ; GFX10-NEXT: s_cselect_b32 s28, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s16, 0 ; GFX10-NEXT: s_cmp_eq_u32 s16, 0
; GFX10-NEXT: s_cselect_b32 s29, 1, 0 ; GFX10-NEXT: s_cselect_b32 s29, 1, 0
; GFX10-NEXT: s_lshl_b64 s[24:25], s[46:47], s16 ; GFX10-NEXT: s_lshl_b64 s[24:25], s[0:1], s16
; GFX10-NEXT: s_lshr_b64 s[2:3], s[26:27], s2 ; GFX10-NEXT: s_lshr_b64 s[2:3], s[26:27], s2
; GFX10-NEXT: s_lshl_b64 s[16:17], s[26:27], s16 ; GFX10-NEXT: s_lshl_b64 s[16:17], s[26:27], s16
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[24:25] ; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[24:25]
; GFX10-NEXT: s_lshl_b64 s[24:25], s[26:27], s23 ; GFX10-NEXT: s_lshl_b64 s[24:25], s[26:27], s23
; GFX10-NEXT: s_cmp_lg_u32 s28, 0 ; GFX10-NEXT: s_cmp_lg_u32 s28, 0
; GFX10-NEXT: s_cselect_b64 s[78:79], s[16:17], 0 ; GFX10-NEXT: s_cselect_b64 s[16:17], s[16:17], 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[24:25] ; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[24:25]
; GFX10-NEXT: s_cmp_lg_u32 s29, 0 ; GFX10-NEXT: s_cmp_lg_u32 s29, 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[46:47], s[2:3] ; GFX10-NEXT: s_cselect_b64 s[2:3], s[0:1], s[2:3]
; GFX10-NEXT: s_sub_i32 s26, s22, 64 ; GFX10-NEXT: s_sub_i32 s26, s22, 64
; GFX10-NEXT: s_sub_i32 s23, 64, s22 ; GFX10-NEXT: s_sub_i32 s23, 64, s22
; GFX10-NEXT: s_cmp_lt_u32 s22, 64 ; GFX10-NEXT: s_cmp_lt_u32 s22, 64
; GFX10-NEXT: s_cselect_b32 s27, 1, 0 ; GFX10-NEXT: s_cselect_b32 s27, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s22, 0 ; GFX10-NEXT: s_cmp_eq_u32 s22, 0
; GFX10-NEXT: s_cselect_b32 s28, 1, 0 ; GFX10-NEXT: s_cselect_b32 s28, 1, 0
; GFX10-NEXT: s_lshr_b64 s[0:1], s[8:9], s22 ; GFX10-NEXT: s_lshr_b64 s[0:1], s[8:9], s22
; GFX10-NEXT: s_lshl_b64 s[24:25], s[62:63], s23 ; GFX10-NEXT: s_lshl_b64 s[24:25], s[10:11], s23
; GFX10-NEXT: s_lshr_b64 s[22:23], s[62:63], s22 ; GFX10-NEXT: s_lshr_b64 s[22:23], s[10:11], s22
; GFX10-NEXT: s_or_b64 s[0:1], s[0:1], s[24:25] ; GFX10-NEXT: s_or_b64 s[0:1], s[0:1], s[24:25]
; GFX10-NEXT: s_lshr_b64 s[10:11], s[62:63], s26 ; GFX10-NEXT: s_lshr_b64 s[10:11], s[10:11], s26
; GFX10-NEXT: s_cmp_lg_u32 s27, 0 ; GFX10-NEXT: s_cmp_lg_u32 s27, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[10:11] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[10:11]
; GFX10-NEXT: s_cmp_lg_u32 s28, 0 ; GFX10-NEXT: s_cmp_lg_u32 s28, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[8:9], s[0:1] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[8:9], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s27, 0 ; GFX10-NEXT: s_cmp_lg_u32 s27, 0
; GFX10-NEXT: s_cselect_b64 s[8:9], s[22:23], 0 ; GFX10-NEXT: s_cselect_b64 s[8:9], s[22:23], 0
; GFX10-NEXT: s_or_b64 s[0:1], s[78:79], s[0:1] ; GFX10-NEXT: s_or_b64 s[0:1], s[16:17], s[0:1]
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[8:9] ; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[8:9]
; GFX10-NEXT: s_and_b64 s[8:9], s[20:21], s[18:19] ; GFX10-NEXT: s_and_b64 s[8:9], s[20:21], s[18:19]
; GFX10-NEXT: s_andn2_b64 s[10:11], s[18:19], s[20:21] ; GFX10-NEXT: s_andn2_b64 s[10:11], s[18:19], s[20:21]
; GFX10-NEXT: s_cmp_lt_u32 1, 64 ; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: s_cselect_b32 s9, 1, 0 ; GFX10-NEXT: s_cselect_b32 s9, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0 ; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: s_cselect_b32 s11, 1, 0 ; GFX10-NEXT: s_cselect_b32 s11, 1, 0
; GFX10-NEXT: s_lshr_b64 s[16:17], s[4:5], s31 ; GFX10-NEXT: s_lshr_b64 s[16:17], s[4:5], s31
▲ Show 20 Lines • Show All 427 Lines • ▼ Show 20 Lines
; ;
; GFX10-LABEL: v_fshr_v2i128: ; GFX10-LABEL: v_fshr_v2i128:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: s_sub_i32 s5, 64, 1 ; GFX10-NEXT: s_sub_i32 s5, 64, 1
; GFX10-NEXT: s_sub_i32 s6, 1, 64 ; GFX10-NEXT: s_sub_i32 s6, 1, 64
; GFX10-NEXT: s_cmp_lt_u32 1, 64 ; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_lshrrev_b64 v[27:28], s5, v[0:1] ; GFX10-NEXT: v_lshrrev_b64 v[17:18], s5, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[21:22], 1, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[21:22], 1, v[2:3]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0 ; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[23:24], 1, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[23:24], 1, v[0:1]
; GFX10-NEXT: s_cselect_b32 s8, 1, 0 ; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: s_and_b32 s4, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s4, 1, vcc_lo
; GFX10-NEXT: v_lshlrev_b64 v[0:1], s6, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], s6, v[0:1]
; GFX10-NEXT: v_xor_b32_e32 v19, -1, v16 ; GFX10-NEXT: v_or_b32_e32 v21, v17, v21
; GFX10-NEXT: v_or_b32_e32 v21, v27, v21
; GFX10-NEXT: v_or_b32_e32 v18, v28, v22
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4
; GFX10-NEXT: v_or_b32_e32 v18, v18, v22
; GFX10-NEXT: v_xor_b32_e32 v19, -1, v16
; GFX10-NEXT: s_movk_i32 s7, 0x7f ; GFX10-NEXT: s_movk_i32 s7, 0x7f
; GFX10-NEXT: s_and_b32 s8, 1, s8 ; GFX10-NEXT: s_and_b32 s8, 1, s8
; GFX10-NEXT: v_and_b32_e32 v31, s7, v19
; GFX10-NEXT: v_cndmask_b32_e32 v17, 0, v23, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v18, s4
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v21, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v21, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v18, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s8 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s8
; GFX10-NEXT: v_and_b32_e32 v25, s7, v19
; GFX10-NEXT: v_cndmask_b32_e32 v17, 0, v23, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v18, 0, v24, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v18, 0, v24, vcc_lo
; GFX10-NEXT: v_sub_nc_u32_e32 v19, 64, v31
; GFX10-NEXT: v_and_b32_e32 v26, s7, v16 ; GFX10-NEXT: v_and_b32_e32 v26, s7, v16
; GFX10-NEXT: v_subrev_nc_u32_e32 v16, 64, v31
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v3, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v3, s4
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v19, v[17:18] ; GFX10-NEXT: v_sub_nc_u32_e32 v2, 64, v25
; GFX10-NEXT: v_mov_b32_e32 v35, v10 ; GFX10-NEXT: v_subrev_nc_u32_e32 v16, 64, v25
; GFX10-NEXT: v_mov_b32_e32 v36, v11 ; GFX10-NEXT: v_sub_nc_u32_e32 v19, 64, v26
; GFX10-NEXT: v_sub_nc_u32_e32 v25, 64, v26 ; GFX10-NEXT: v_lshlrev_b64 v[23:24], v25, v[17:18]
; GFX10-NEXT: v_lshlrev_b64 v[21:22], v31, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[21:22], v25, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[23:24], v31, v[17:18] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], v2, v[17:18]
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v31 ; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v25
; GFX10-NEXT: v_subrev_nc_u32_e32 v29, 64, v26 ; GFX10-NEXT: v_subrev_nc_u32_e32 v27, 64, v26
; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v26 ; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v26
; GFX10-NEXT: v_lshrrev_b64 v[27:28], s5, v[4:5] ; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_or_b32_e32 v21, v2, v21 ; GFX10-NEXT: v_or_b32_e32 v21, v2, v21
; GFX10-NEXT: v_or_b32_e32 v22, v3, v22 ; GFX10-NEXT: v_or_b32_e32 v22, v3, v22
; GFX10-NEXT: v_lshlrev_b64 v[2:3], v16, v[17:18] ; GFX10-NEXT: v_lshlrev_b64 v[2:3], v16, v[17:18]
; GFX10-NEXT: v_lshlrev_b64 v[18:19], v25, v[35:36]
; GFX10-NEXT: v_lshrrev_b64 v[16:17], v26, v[8:9] ; GFX10-NEXT: v_lshrrev_b64 v[16:17], v26, v[8:9]
; GFX10-NEXT: v_lshlrev_b64 v[18:19], v19, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e32 v23, 0, v23, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v23, 0, v23, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v39, 0, v24, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v24, 0, v24, vcc_lo
; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_cndmask_b32_e32 v21, v2, v21, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v21, v2, v21, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v22, v3, v22, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v22, v3, v22, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v29, v[35:36] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], v27, v[10:11]
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v31 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v25
; GFX10-NEXT: v_or_b32_e32 v16, v16, v18 ; GFX10-NEXT: v_or_b32_e32 v16, v16, v18
; GFX10-NEXT: v_or_b32_e32 v17, v17, v19 ; GFX10-NEXT: v_or_b32_e32 v17, v17, v19
; GFX10-NEXT: v_lshlrev_b64 v[10:11], 1, v[6:7] ; GFX10-NEXT: v_xor_b32_e32 v25, -1, v20
; GFX10-NEXT: s_cselect_b32 s5, 1, 0
; GFX10-NEXT: v_cndmask_b32_e32 v18, v21, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v18, v21, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v31, v22, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v22, v22, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v16, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v16, s4
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v26 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v26
; GFX10-NEXT: s_cmp_eq_u32 1, 0 ; GFX10-NEXT: v_lshrrev_b64 v[0:1], v26, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e64 v19, v3, v17, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v19, v3, v17, s4
; GFX10-NEXT: s_cselect_b32 s8, 1, 0 ; GFX10-NEXT: v_lshlrev_b64 v[10:11], 1, v[6:7]
; GFX10-NEXT: v_lshlrev_b64 v[16:17], 1, v[4:5] ; GFX10-NEXT: v_lshlrev_b64 v[16:17], 1, v[4:5]
; GFX10-NEXT: v_cndmask_b32_e32 v21, v2, v8, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[2:3], s5, v[4:5]
; GFX10-NEXT: s_cselect_b32 s5, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[4:5], s6, v[4:5] ; GFX10-NEXT: v_lshlrev_b64 v[4:5], s6, v[4:5]
; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: s_and_b32 s6, 1, s5 ; GFX10-NEXT: s_and_b32 s6, 1, s5
; GFX10-NEXT: v_cndmask_b32_e32 v21, v2, v8, vcc_lo ; GFX10-NEXT: v_or_b32_e32 v2, v2, v10
; GFX10-NEXT: v_xor_b32_e32 v25, -1, v20
; GFX10-NEXT: v_or_b32_e32 v2, v27, v10
; GFX10-NEXT: v_or_b32_e32 v3, v28, v11
; GFX10-NEXT: v_cmp_ne_u32_e64 s6, 0, s6 ; GFX10-NEXT: v_cmp_ne_u32_e64 s6, 0, s6
; GFX10-NEXT: v_or_b32_e32 v3, v3, v11
; GFX10-NEXT: s_and_b32 s8, 1, s8 ; GFX10-NEXT: s_and_b32 s8, 1, s8
; GFX10-NEXT: v_lshrrev_b64 v[0:1], v26, v[35:36]
; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, v16, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, v16, s5
; GFX10-NEXT: v_cndmask_b32_e32 v16, v19, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v16, v19, v9, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v11, v4, v2, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v4, v2, s6
; GFX10-NEXT: v_and_b32_e32 v30, s7, v25 ; GFX10-NEXT: v_and_b32_e32 v25, s7, v25
; GFX10-NEXT: v_cndmask_b32_e64 v4, v5, v3, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v5, v3, s6
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s8 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s8
; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, v17, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, v17, s5
; GFX10-NEXT: v_cndmask_b32_e64 v19, 0, v0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v19, 0, v0, s4
; GFX10-NEXT: v_sub_nc_u32_e32 v2, 64, v30
; GFX10-NEXT: v_or_b32_e32 v0, v23, v21 ; GFX10-NEXT: v_or_b32_e32 v0, v23, v21
; GFX10-NEXT: v_cndmask_b32_e32 v3, v11, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v4, v4, v7, vcc_lo
; GFX10-NEXT: v_and_b32_e32 v23, s7, v20 ; GFX10-NEXT: v_and_b32_e32 v23, s7, v20
; GFX10-NEXT: v_lshrrev_b64 v[5:6], v2, v[8:9] ; GFX10-NEXT: v_cndmask_b32_e32 v3, v2, v6, vcc_lo
; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 64, v30 ; GFX10-NEXT: v_cndmask_b32_e32 v4, v4, v7, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, v1, s4 ; GFX10-NEXT: v_sub_nc_u32_e32 v2, 64, v25
; GFX10-NEXT: v_lshlrev_b64 v[10:11], v30, v[3:4] ; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 64, v25
; GFX10-NEXT: v_sub_nc_u32_e32 v20, 64, v23 ; GFX10-NEXT: v_sub_nc_u32_e32 v20, 64, v23
; GFX10-NEXT: v_or_b32_e32 v1, v39, v16 ; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, v1, s4
; GFX10-NEXT: v_lshlrev_b64 v[10:11], v25, v[3:4]
; GFX10-NEXT: v_lshrrev_b64 v[5:6], v2, v[8:9]
; GFX10-NEXT: v_or_b32_e32 v1, v24, v16
; GFX10-NEXT: v_or_b32_e32 v2, v18, v19 ; GFX10-NEXT: v_or_b32_e32 v2, v18, v19
; GFX10-NEXT: v_lshlrev_b64 v[16:17], v30, v[8:9] ; GFX10-NEXT: v_lshlrev_b64 v[16:17], v25, v[8:9]
; GFX10-NEXT: v_lshrrev_b64 v[18:19], v23, v[12:13] ; GFX10-NEXT: v_lshrrev_b64 v[18:19], v23, v[12:13]
; GFX10-NEXT: v_lshlrev_b64 v[20:21], v20, v[14:15]
; GFX10-NEXT: v_or_b32_e32 v10, v5, v10 ; GFX10-NEXT: v_or_b32_e32 v10, v5, v10
; GFX10-NEXT: v_subrev_nc_u32_e32 v5, 64, v23 ; GFX10-NEXT: v_subrev_nc_u32_e32 v5, 64, v23
; GFX10-NEXT: v_lshlrev_b64 v[20:21], v20, v[14:15] ; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v25
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v30
; GFX10-NEXT: v_lshlrev_b64 v[7:8], v7, v[8:9] ; GFX10-NEXT: v_lshlrev_b64 v[7:8], v7, v[8:9]
; GFX10-NEXT: v_or_b32_e32 v9, v6, v11 ; GFX10-NEXT: v_or_b32_e32 v9, v6, v11
; GFX10-NEXT: v_lshrrev_b64 v[34:35], v5, v[14:15]
; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v23 ; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v23
; GFX10-NEXT: v_lshrrev_b64 v[5:6], v5, v[14:15]
; GFX10-NEXT: v_cndmask_b32_e32 v11, 0, v16, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v11, 0, v16, vcc_lo
; GFX10-NEXT: v_or_b32_e32 v16, v18, v20 ; GFX10-NEXT: v_or_b32_e32 v16, v18, v20
; GFX10-NEXT: v_or_b32_e32 v18, v19, v21 ; GFX10-NEXT: v_or_b32_e32 v18, v19, v21
; GFX10-NEXT: v_cndmask_b32_e32 v10, v7, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, v7, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v9, v8, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v9, v8, v9, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[7:8], v23, v[14:15] ; GFX10-NEXT: v_lshrrev_b64 v[7:8], v23, v[14:15]
; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v30 ; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v25
; GFX10-NEXT: v_cndmask_b32_e64 v5, v34, v16, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v16, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v23 ; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v23
; GFX10-NEXT: v_cndmask_b32_e64 v6, v35, v18, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v18, s4
; GFX10-NEXT: v_cndmask_b32_e32 v14, 0, v17, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v14, 0, v17, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v15, v10, v3, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v10, v10, v3, s6
; GFX10-NEXT: v_cndmask_b32_e64 v9, v9, v4, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v9, v9, v4, s6
; GFX10-NEXT: v_cndmask_b32_e64 v4, v5, v12, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v5, v12, s5
; GFX10-NEXT: v_cndmask_b32_e64 v5, v6, v13, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v6, v13, s5
; GFX10-NEXT: v_cndmask_b32_e64 v6, 0, v7, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, 0, v7, s4
; GFX10-NEXT: v_cndmask_b32_e64 v7, 0, v8, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v7, 0, v8, s4
; GFX10-NEXT: v_or_b32_e32 v3, v31, v26 ; GFX10-NEXT: v_or_b32_e32 v3, v22, v26
; GFX10-NEXT: v_or_b32_e32 v4, v11, v4 ; GFX10-NEXT: v_or_b32_e32 v4, v11, v4
; GFX10-NEXT: v_or_b32_e32 v5, v14, v5 ; GFX10-NEXT: v_or_b32_e32 v5, v14, v5
; GFX10-NEXT: v_or_b32_e32 v6, v15, v6 ; GFX10-NEXT: v_or_b32_e32 v6, v10, v6
; GFX10-NEXT: v_or_b32_e32 v7, v9, v7 ; GFX10-NEXT: v_or_b32_e32 v7, v9, v7
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i128> @llvm.fshr.v2i128(<2 x i128> %lhs, <2 x i128> %rhs, <2 x i128> %amt) %result = call <2 x i128> @llvm.fshr.v2i128(<2 x i128> %lhs, <2 x i128> %rhs, <2 x i128> %amt)
ret <2 x i128> %result ret <2 x i128> %result
} }
declare i7 @llvm.fshr.i7(i7, i7, i7) #0 declare i7 @llvm.fshr.i7(i7, i7, i7) #0
declare i8 @llvm.fshr.i8(i8, i8, i8) #0 declare i8 @llvm.fshr.i8(i8, i8, i8) #0
Show All 30 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.i16.ll

Show First 20 Lines • Show All 2,229 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v1 ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v1
; GFX10-NEXT: v_lshlrev_b32_e64 v7, v0, s0 ; GFX10-NEXT: v_lshlrev_b32_e64 v7, v0, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v1 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v1
; GFX10-NEXT: v_lshlrev_b32_e64 v0, v0, s1 ; GFX10-NEXT: v_lshlrev_b32_e64 v0, v0, s1
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v1 ; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v1
; GFX10-NEXT: v_xor_b32_e32 v7, -1, v7 ; GFX10-NEXT: v_xor_b32_e32 v7, -1, v7
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v11, v2, v5, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v5, s0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v11, v6, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v6, s1
; GFX10-NEXT: v_and_or_b32 v7, v2, v7, v0 ; GFX10-NEXT: v_and_or_b32 v7, v2, v7, v0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v4, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v4, v7, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v5, v7, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v5, v7, s0
; GFX10-NEXT: v_mov_b32_e32 v4, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v3, v7, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v3, v7, s2
; GFX10-NEXT: v_cndmask_b32_e64 v3, v6, v7, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v6, v7, s1
; GFX10-NEXT: v_mov_b32_e32 v5, 0 ; GFX10-NEXT: v_mov_b32_e32 v5, 0
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off ; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
▲ Show 20 Lines • Show All 229 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v1 ; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v1
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v1 ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v1
; GFX10-NEXT: v_lshlrev_b32_e64 v8, v0, s0 ; GFX10-NEXT: v_lshlrev_b32_e64 v8, v0, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v1 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v1
; GFX10-NEXT: v_lshlrev_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT: v_xor_b32_e32 v2, -1, v8 ; GFX10-NEXT: v_xor_b32_e32 v2, -1, v8
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v10, v3, v6, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v6, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v10, v7, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v7, s1
; GFX10-NEXT: v_and_or_b32 v3, v3, v2, v0 ; GFX10-NEXT: v_and_or_b32 v3, v3, v2, v0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, v3, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, v3, s2
; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v3, vcc_lo
; GFX10-NEXT: v_mov_b32_e32 v4, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v3, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v7, v3, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v7, v3, s1
; GFX10-NEXT: v_mov_b32_e32 v5, 0 ; GFX10-NEXT: v_mov_b32_e32 v5, 0
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off ; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
▲ Show 20 Lines • Show All 402 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, s7, 2 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, s7, 2
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, s7, 3 ; GFX10-NEXT: v_cmp_eq_u32_e64 s1, s7, 3
; GFX10-NEXT: v_cmp_eq_u32_e64 s4, s7, 4 ; GFX10-NEXT: v_cmp_eq_u32_e64 s4, s7, 4
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, s7, 5 ; GFX10-NEXT: v_cmp_eq_u32_e64 s5, s7, 5
; GFX10-NEXT: v_cmp_eq_u32_e64 s6, s7, 6 ; GFX10-NEXT: v_cmp_eq_u32_e64 s6, s7, 6
; GFX10-NEXT: s_and_b32 s9, s2, s8 ; GFX10-NEXT: s_and_b32 s9, s2, s8
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, s7, 7 ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, s7, 7
; GFX10-NEXT: s_and_b32 s3, s3, 1 ; GFX10-NEXT: s_and_b32 s3, s3, 1
; GFX10-NEXT: v_mov_b32_e32 v13, 0 ; GFX10-NEXT: v_mov_b32_e32 v10, 0
; GFX10-NEXT: s_lshl_b32 s3, s3, 4 ; GFX10-NEXT: s_lshl_b32 s3, s3, 4
; GFX10-NEXT: v_mov_b32_e32 v11, 0
; GFX10-NEXT: s_lshl_b32 s8, s8, s3 ; GFX10-NEXT: s_lshl_b32 s8, s8, s3
; GFX10-NEXT: s_lshl_b32 s3, s9, s3 ; GFX10-NEXT: s_lshl_b32 s3, s9, s3
; GFX10-NEXT: s_not_b32 s8, s8 ; GFX10-NEXT: s_not_b32 s8, s8
; GFX10-NEXT: v_mov_b32_e32 v13, 0
; GFX10-NEXT: s_waitcnt vmcnt(1) ; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_cndmask_b32_e32 v11, v2, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v11, v4, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v4, s0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v5, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v5, s1
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v6, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v6, s4
; GFX10-NEXT: v_cndmask_b32_e64 v11, v0, v7, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v7, s5
; GFX10-NEXT: v_cndmask_b32_e64 v0, v11, v8, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v8, s6
; GFX10-NEXT: v_mov_b32_e32 v10, 0
; GFX10-NEXT: v_mov_b32_e32 v11, 0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v9, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v9, s2
; GFX10-NEXT: v_and_or_b32 v12, v0, s8, s3 ; GFX10-NEXT: v_and_or_b32 v12, v0, s8, s3
; GFX10-NEXT: v_cmp_eq_u32_e64 s3, s7, 0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s3, s7, 0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v2, v12, s3 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v2, v12, s3
; GFX10-NEXT: v_cndmask_b32_e64 v2, v4, v12, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v4, v12, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v5, v12, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v5, v12, s1
; GFX10-NEXT: v_cndmask_b32_e64 v4, v6, v12, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v6, v12, s4
▲ Show 20 Lines • Show All 889 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_and_b32 s6, s2, s5 ; GFX10-NEXT: s_and_b32 s6, s2, s5
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v0
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v0
; GFX10-NEXT: v_cmp_eq_u32_e64 s3, 4, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s3, 4, v0
; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 5, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 5, v0
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 4, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 4, v2
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 6, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 6, v0
; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_cndmask_b32_e32 v11, v3, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v11, v5, s0
; GFX10-NEXT: v_lshlrev_b32_e64 v11, v2, s5 ; GFX10-NEXT: v_lshlrev_b32_e64 v11, v2, s5
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 7, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 7, v0
; GFX10-NEXT: v_lshlrev_b32_e64 v2, v2, s6 ; GFX10-NEXT: v_lshlrev_b32_e64 v2, v2, s6
; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v6, s1
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v11 ; GFX10-NEXT: v_xor_b32_e32 v11, -1, v11
; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v5, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v6, s1
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v7, s3 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v7, s3
; GFX10-NEXT: v_cndmask_b32_e64 v15, v1, v8, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v8, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v15, v9, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v9, s2
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v10, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v10, s5
; GFX10-NEXT: v_and_or_b32 v13, v1, v11, v2 ; GFX10-NEXT: v_and_or_b32 v13, v1, v11, v2
; GFX10-NEXT: v_mov_b32_e32 v11, 0 ; GFX10-NEXT: v_mov_b32_e32 v11, 0
; GFX10-NEXT: v_mov_b32_e32 v12, 0 ; GFX10-NEXT: v_mov_b32_e32 v12, 0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v3, v13, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v3, v13, s6
; GFX10-NEXT: v_cndmask_b32_e32 v1, v4, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v4, v13, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v5, v13, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v5, v13, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v6, v13, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v6, v13, s1
▲ Show 20 Lines • Show All 169 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, s6, 6 ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, s6, 6
; GFX10-NEXT: s_lshl_b32 s7, s5, 4 ; GFX10-NEXT: s_lshl_b32 s7, s5, 4
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, s6, 7 ; GFX10-NEXT: v_cmp_eq_u32_e64 s5, s6, 7
; GFX10-NEXT: s_mov_b32 s8, 0xffff ; GFX10-NEXT: s_mov_b32 s8, 0xffff
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s7, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s7, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT: s_lshl_b32 s7, s8, s7 ; GFX10-NEXT: s_lshl_b32 s7, s8, s7
; GFX10-NEXT: v_cmp_eq_u32_e64 s6, s6, 0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s6, s6, 0
; GFX10-NEXT: s_not_b32 s7, s7 ; GFX10-NEXT: s_not_b32 s7, s7
; GFX10-NEXT: v_mov_b32_e32 v11, 0
; GFX10-NEXT: v_mov_b32_e32 v12, 0
; GFX10-NEXT: v_mov_b32_e32 v14, 0 ; GFX10-NEXT: v_mov_b32_e32 v14, 0
; GFX10-NEXT: s_waitcnt vmcnt(1) ; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v3, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v5, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v5, s0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v6, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v6, s1
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e64 v11, v0, v7, s3 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v7, s3
; GFX10-NEXT: v_cndmask_b32_e64 v0, v11, v8, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v8, s4
; GFX10-NEXT: v_mov_b32_e32 v11, 0
; GFX10-NEXT: v_mov_b32_e32 v12, 0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v9, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v9, s2
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v10, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v10, s5
; GFX10-NEXT: v_and_or_b32 v13, v0, s7, v1 ; GFX10-NEXT: v_and_or_b32 v13, v0, s7, v1
; GFX10-NEXT: v_cndmask_b32_e64 v0, v3, v13, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v3, v13, s6
; GFX10-NEXT: v_cndmask_b32_e32 v1, v4, v13, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v4, v13, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v5, v13, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v5, v13, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v6, v13, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v6, v13, s1
; GFX10-NEXT: v_cndmask_b32_e64 v4, v7, v13, s3 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v7, v13, s3
▲ Show 20 Lines • Show All 155 Lines • ▼ Show 20 Lines
; GFX10-LABEL: insertelement_v_v16i16_v_v: ; GFX10-LABEL: insertelement_v_v16i16_v_v:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx4 v[4:7], v[0:1], off ; GFX10-NEXT: global_load_dwordx4 v[4:7], v[0:1], off
; GFX10-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 ; GFX10-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 1, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v0, 1, v3
; GFX10-NEXT: v_and_b32_e32 v3, 1, v3 ; GFX10-NEXT: v_and_b32_e32 v3, 1, v3
; GFX10-NEXT: s_mov_b32 s4, 0xffff ; GFX10-NEXT: s_mov_b32 s4, 0xffff
; GFX10-NEXT: v_mov_b32_e32 v15, 0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v0
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v0
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 4, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 4, v0
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 4, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 4, v3
; GFX10-NEXT: v_cmp_eq_u32_e64 s3, 5, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s3, 5, v0
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 7, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 7, v0
; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v0
; GFX10-NEXT: v_lshlrev_b32_e64 v12, v3, s4 ; GFX10-NEXT: v_lshlrev_b32_e64 v12, v3, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 6, v0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s4, 6, v0
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT: v_xor_b32_e32 v3, -1, v12 ; GFX10-NEXT: v_xor_b32_e32 v3, -1, v12
; GFX10-NEXT: v_mov_b32_e32 v12, 0 ; GFX10-NEXT: v_mov_b32_e32 v12, 0
; GFX10-NEXT: v_mov_b32_e32 v13, 0 ; GFX10-NEXT: v_mov_b32_e32 v13, 0
; GFX10-NEXT: s_waitcnt vmcnt(1) ; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_cndmask_b32_e32 v1, v4, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v4, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v6, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v6, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v7, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v7, s1
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cndmask_b32_e64 v15, v1, v8, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v8, s2
; GFX10-NEXT: v_cndmask_b32_e64 v1, v15, v9, s3 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v9, s3
; GFX10-NEXT: v_mov_b32_e32 v15, 0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v10, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v10, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v11, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v11, s5
; GFX10-NEXT: v_and_or_b32 v14, v1, v3, v2 ; GFX10-NEXT: v_and_or_b32 v14, v1, v3, v2
; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, v14, s6 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, v14, s6
; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v14, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v14, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v14, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v7, v14, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v7, v14, s1
; GFX10-NEXT: v_cndmask_b32_e64 v4, v8, v14, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v8, v14, s2
Show All 12 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.i8.ll

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 1,632 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_lshl_b32 s1, s1, 3 ; GFX10-NEXT: s_lshl_b32 s1, s1, 3
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 24, v0
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_movk_i32 s0, 0xff ; GFX10-NEXT: s_movk_i32 s0, 0xff
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; GFX10-NEXT: v_and_b32_sdwa v4, v0, s0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v4, v0, s0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_or_b32 v6, v0, s0, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, s0, v1
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_lshl_b32 s1, s0, s1 ; GFX10-NEXT: s_lshl_b32 s1, s0, s1
; GFX10-NEXT: s_not_b32 s1, s1 ; GFX10-NEXT: s_not_b32 s1, s1
; GFX10-NEXT: v_or3_b32 v0, v6, v4, v3 ; GFX10-NEXT: v_or3_b32 v0, v0, v4, v3
; GFX10-NEXT: v_and_or_b32 v0, v0, s1, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, s1, v1
; GFX10-NEXT: v_mov_b32_e32 v1, 8 ; GFX10-NEXT: v_mov_b32_e32 v1, 8
; GFX10-NEXT: v_lshrrev_b32_e32 v2, 8, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v2, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 24, v0
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v2, v0, s0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v2, v0, s0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; GFX10-NEXT: v_and_or_b32 v0, v0, s0, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, s0, v1
▲ Show 20 Lines • Show All 135 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_xor_b32_e32 v2, -1, v5 ; GFX10-NEXT: v_xor_b32_e32 v2, -1, v5
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0
; GFX10-NEXT: v_and_b32_sdwa v6, v0, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v6, v0, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; GFX10-NEXT: v_and_or_b32 v3, v0, s1, v3 ; GFX10-NEXT: v_and_or_b32 v0, v0, s1, v3
; GFX10-NEXT: v_or3_b32 v0, v3, v6, v4
; GFX10-NEXT: v_mov_b32_e32 v3, 0xff ; GFX10-NEXT: v_mov_b32_e32 v3, 0xff
; GFX10-NEXT: v_or3_b32 v0, v0, v6, v4
; GFX10-NEXT: v_and_or_b32 v0, v0, v2, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, v2, v1
; GFX10-NEXT: v_mov_b32_e32 v1, 8 ; GFX10-NEXT: v_mov_b32_e32 v1, 8
; GFX10-NEXT: v_lshrrev_b32_e32 v2, 8, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v2, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v2, v0, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v2, v0, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; GFX10-NEXT: v_and_or_b32 v3, v0, v3, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, v3, v1
; GFX10-NEXT: v_or3_b32 v2, v0, v2, v4
; GFX10-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-NEXT: v_mov_b32_e32 v0, 0
; GFX10-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-NEXT: v_mov_b32_e32 v1, 0
; GFX10-NEXT: v_or3_b32 v2, v3, v2, v4
; GFX10-NEXT: global_store_dword v[0:1], v2, off ; GFX10-NEXT: global_store_dword v[0:1], v2, off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <4 x i8>, <4 x i8> addrspace(1)* %ptr %vec = load <4 x i8>, <4 x i8> addrspace(1)* %ptr
%insert = insertelement <4 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <4 x i8> %vec, i8 %val, i32 %idx
store <4 x i8> %insert, <4 x i8> addrspace(1)* null store <4 x i8> %insert, <4 x i8> addrspace(1)* null
ret void ret void
} }
▲ Show 20 Lines • Show All 500 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v1
; GFX10-NEXT: v_and_b32_sdwa v6, v0, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v6, v0, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v7, v1, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v7, v1, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5
; GFX10-NEXT: v_and_or_b32 v11, v0, s1, v2 ; GFX10-NEXT: v_and_or_b32 v0, v0, s1, v2
; GFX10-NEXT: v_and_or_b32 v2, v1, s1, v3 ; GFX10-NEXT: v_and_or_b32 v1, v1, s1, v3
; GFX10-NEXT: s_lshr_b32 s0, s3, 2 ; GFX10-NEXT: s_lshr_b32 s0, s3, 2
; GFX10-NEXT: s_and_b32 s3, s3, 3 ; GFX10-NEXT: s_and_b32 s3, s3, 3
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 1 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s0, 1
; GFX10-NEXT: v_or3_b32 v0, v11, v6, v4 ; GFX10-NEXT: v_or3_b32 v0, v0, v6, v4
; GFX10-NEXT: v_or3_b32 v1, v2, v7, v5 ; GFX10-NEXT: v_or3_b32 v1, v1, v7, v5
; GFX10-NEXT: s_lshl_b32 s3, s3, 3 ; GFX10-NEXT: s_lshl_b32 s3, s3, 3
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, s0, 0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, s0, 0
; GFX10-NEXT: s_lshl_b32 s4, s1, s3 ; GFX10-NEXT: s_lshl_b32 s4, s1, s3
; GFX10-NEXT: s_lshl_b32 s2, s2, s3 ; GFX10-NEXT: s_lshl_b32 s2, s2, s3
; GFX10-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v0, v1, vcc_lo
; GFX10-NEXT: s_not_b32 s3, s4 ; GFX10-NEXT: s_not_b32 s3, s4
; GFX10-NEXT: v_and_or_b32 v2, v2, s3, s2 ; GFX10-NEXT: v_and_or_b32 v2, v2, s3, s2
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s0
▲ Show 20 Lines • Show All 282 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_and_b32_sdwa v6, v0, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v6, v0, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v1
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v7, v1, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v7, v1, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5
; GFX10-NEXT: v_and_or_b32 v11, v0, s2, v2 ; GFX10-NEXT: v_and_or_b32 v0, v0, s2, v2
; GFX10-NEXT: v_and_or_b32 v2, v1, s2, v3 ; GFX10-NEXT: v_and_or_b32 v1, v1, s2, v3
; GFX10-NEXT: v_or3_b32 v0, v11, v6, v4
; GFX10-NEXT: v_or3_b32 v1, v2, v7, v5
; GFX10-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-NEXT: v_mov_b32_e32 v2, 0
; GFX10-NEXT: v_or3_b32 v0, v0, v6, v4
; GFX10-NEXT: v_mov_b32_e32 v3, 0 ; GFX10-NEXT: v_mov_b32_e32 v3, 0
; GFX10-NEXT: v_or3_b32 v1, v1, v7, v5
; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off ; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <8 x i8>, <8 x i8> addrspace(4)* %ptr %vec = load <8 x i8>, <8 x i8> addrspace(4)* %ptr
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
store <8 x i8> %insert, <8 x i8> addrspace(1)* null store <8 x i8> %insert, <8 x i8> addrspace(1)* null
ret void ret void
} }
▲ Show 20 Lines • Show All 254 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_mov_b32_e32 v1, s1 ; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v2 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v2
; GFX10-NEXT: v_and_or_b32 v3, v5, v4, v3 ; GFX10-NEXT: v_and_or_b32 v3, v5, v4, v3
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v3, s0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: s_mov_b32 s0, 8 ; GFX10-NEXT: s_mov_b32 s0, 8
; GFX10-NEXT: v_lshrrev_b32_e32 v2, 8, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v2, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0
; GFX10-NEXT: v_and_b32_sdwa v7, v1, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v1
; GFX10-NEXT: v_and_b32_sdwa v6, v0, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5 ; GFX10-NEXT: v_and_b32_sdwa v7, v1, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v6, v0, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; GFX10-NEXT: v_and_or_b32 v11, v0, s2, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5
; GFX10-NEXT: v_and_or_b32 v2, v1, s2, v3 ; GFX10-NEXT: v_and_or_b32 v0, v0, s2, v2
; GFX10-NEXT: v_or3_b32 v0, v11, v6, v4 ; GFX10-NEXT: v_and_or_b32 v1, v1, s2, v3
; GFX10-NEXT: v_or3_b32 v1, v2, v7, v5
; GFX10-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-NEXT: v_mov_b32_e32 v2, 0
; GFX10-NEXT: v_mov_b32_e32 v3, 0 ; GFX10-NEXT: v_mov_b32_e32 v3, 0
; GFX10-NEXT: v_or3_b32 v0, v0, v6, v4
; GFX10-NEXT: v_or3_b32 v1, v1, v7, v5
; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off ; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <8 x i8>, <8 x i8> addrspace(4)* %ptr %vec = load <8 x i8>, <8 x i8> addrspace(4)* %ptr
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
store <8 x i8> %insert, <8 x i8> addrspace(1)* null store <8 x i8> %insert, <8 x i8> addrspace(1)* null
ret void ret void
} }
▲ Show 20 Lines • Show All 251 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_mov_b32_e32 v1, s1 ; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v3 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v3
; GFX10-NEXT: v_and_or_b32 v2, v5, v2, v4 ; GFX10-NEXT: v_and_or_b32 v2, v5, v2, v4
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
; GFX10-NEXT: s_mov_b32 s0, 8 ; GFX10-NEXT: s_mov_b32 s0, 8
; GFX10-NEXT: v_lshrrev_b32_e32 v2, 8, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v2, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v0
; GFX10-NEXT: v_and_b32_sdwa v7, v1, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v1
; GFX10-NEXT: v_and_b32_sdwa v6, v0, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5 ; GFX10-NEXT: v_and_b32_sdwa v7, v1, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v6, v0, s2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; GFX10-NEXT: v_and_or_b32 v11, v0, s2, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5
; GFX10-NEXT: v_and_or_b32 v2, v1, s2, v3 ; GFX10-NEXT: v_and_or_b32 v0, v0, s2, v2
; GFX10-NEXT: v_or3_b32 v0, v11, v6, v4 ; GFX10-NEXT: v_and_or_b32 v1, v1, s2, v3
; GFX10-NEXT: v_or3_b32 v1, v2, v7, v5
; GFX10-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-NEXT: v_mov_b32_e32 v2, 0
; GFX10-NEXT: v_mov_b32_e32 v3, 0 ; GFX10-NEXT: v_mov_b32_e32 v3, 0
; GFX10-NEXT: v_or3_b32 v0, v0, v6, v4
; GFX10-NEXT: v_or3_b32 v1, v1, v7, v5
; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off ; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <8 x i8>, <8 x i8> addrspace(4)* %ptr %vec = load <8 x i8>, <8 x i8> addrspace(4)* %ptr
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
store <8 x i8> %insert, <8 x i8> addrspace(1)* null store <8 x i8> %insert, <8 x i8> addrspace(1)* null
ret void ret void
} }
▲ Show 20 Lines • Show All 181 Lines • ▼ Show 20 Lines
; GFX7-NEXT: v_or_b32_e32 v1, v1, v6 ; GFX7-NEXT: v_or_b32_e32 v1, v1, v6
; GFX7-NEXT: v_or_b32_e32 v0, v0, v4 ; GFX7-NEXT: v_or_b32_e32 v0, v0, v4
; GFX7-NEXT: v_or_b32_e32 v1, v1, v7 ; GFX7-NEXT: v_or_b32_e32 v1, v1, v7
; GFX7-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 ; GFX7-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
; GFX7-NEXT: s_endpgm ; GFX7-NEXT: s_endpgm
; ;
; GFX10-LABEL: insertelement_v_v8i8_s_v: ; GFX10-LABEL: insertelement_v_v8i8_s_v:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: global_load_dwordx2 v[11:12], v[0:1], off ; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_mov_b32 s0, 8 ; GFX10-NEXT: s_mov_b32 s0, 8
; GFX10-NEXT: v_and_b32_e32 v3, 3, v2 ; GFX10-NEXT: v_and_b32_e32 v3, 3, v2
; GFX10-NEXT: s_movk_i32 s1, 0xff ; GFX10-NEXT: s_movk_i32 s1, 0xff
; GFX10-NEXT: v_lshrrev_b32_e32 v2, 2, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v2, 2, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 3, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 3, v3
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v11 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v12 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v11 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v12 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1
; GFX10-NEXT: v_and_b32_sdwa v8, v11, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v8, v0, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v5, s0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v5, s0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v9, v12, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v9, v1, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6 ; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7 ; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7
; GFX10-NEXT: v_and_or_b32 v11, v11, s1, v4 ; GFX10-NEXT: v_and_or_b32 v0, v0, s1, v4
; GFX10-NEXT: v_and_or_b32 v10, v12, s1, v5 ; GFX10-NEXT: v_and_or_b32 v1, v1, s1, v5
; GFX10-NEXT: v_lshlrev_b32_e64 v4, v3, s1 ; GFX10-NEXT: v_lshlrev_b32_e64 v4, v3, s1
; GFX10-NEXT: s_and_b32 s0, s2, s1 ; GFX10-NEXT: s_and_b32 s0, s2, s1
; GFX10-NEXT: v_or3_b32 v0, v11, v8, v6 ; GFX10-NEXT: v_or3_b32 v0, v0, v8, v6
; GFX10-NEXT: v_or3_b32 v1, v10, v9, v7 ; GFX10-NEXT: v_or3_b32 v1, v1, v9, v7
; GFX10-NEXT: v_lshlrev_b32_e64 v3, v3, s0 ; GFX10-NEXT: v_lshlrev_b32_e64 v3, v3, s0
; GFX10-NEXT: v_xor_b32_e32 v4, -1, v4 ; GFX10-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v2 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v2
; GFX10-NEXT: v_mov_b32_e32 v2, 8 ; GFX10-NEXT: v_mov_b32_e32 v2, 8
; GFX10-NEXT: v_cndmask_b32_e32 v5, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v0, v1, vcc_lo
; GFX10-NEXT: v_and_or_b32 v3, v5, v4, v3 ; GFX10-NEXT: v_and_or_b32 v3, v5, v4, v3
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v3, s0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
▲ Show 20 Lines • Show All 469 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v8, 24, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v9, 24, v1
; GFX10-NEXT: v_and_b32_sdwa v10, v0, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v10, v0, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_sdwa v6, s0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v6, s0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v7, s0, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, s0, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v11, v1, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v11, v1, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8 ; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9 ; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9
; GFX10-NEXT: v_and_or_b32 v15, v0, s1, v6 ; GFX10-NEXT: v_and_or_b32 v0, v0, s1, v6
; GFX10-NEXT: v_and_or_b32 v14, v1, s1, v7 ; GFX10-NEXT: v_and_or_b32 v1, v1, s1, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v6, v4, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v6, v4, v5
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v3 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v3
; GFX10-NEXT: v_or3_b32 v0, v15, v10, v8 ; GFX10-NEXT: v_or3_b32 v0, v0, v10, v8
; GFX10-NEXT: v_or3_b32 v1, v14, v11, v9 ; GFX10-NEXT: v_or3_b32 v1, v1, v11, v9
; GFX10-NEXT: v_xor_b32_e32 v4, -1, v6 ; GFX10-NEXT: v_xor_b32_e32 v4, -1, v6
; GFX10-NEXT: v_cndmask_b32_e32 v7, v0, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v6, v0, v1, vcc_lo
; GFX10-NEXT: v_and_or_b32 v2, v7, v4, v2 ; GFX10-NEXT: v_and_or_b32 v2, v6, v4, v2
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s0
; GFX10-NEXT: v_cndmask_b32_e32 v11, v1, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
; GFX10-NEXT: v_mov_b32_e32 v2, 8 ; GFX10-NEXT: v_mov_b32_e32 v2, 8
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v11 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v11 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1
; GFX10-NEXT: v_and_b32_sdwa v8, v0, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v8, v0, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_sdwa v3, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v3, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, v2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v2, v2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v4, v1, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6 ; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: v_and_b32_sdwa v4, v11, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7 ; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7
; GFX10-NEXT: v_and_or_b32 v3, v0, v5, v3 ; GFX10-NEXT: v_and_or_b32 v0, v0, v5, v3
; GFX10-NEXT: v_and_or_b32 v1, v11, v5, v2 ; GFX10-NEXT: v_and_or_b32 v1, v1, v5, v2
; GFX10-NEXT: v_or3_b32 v0, v3, v8, v6
; GFX10-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-NEXT: v_mov_b32_e32 v2, 0
; GFX10-NEXT: v_or3_b32 v1, v1, v4, v7
; GFX10-NEXT: v_mov_b32_e32 v3, 0 ; GFX10-NEXT: v_mov_b32_e32 v3, 0
; GFX10-NEXT: v_or3_b32 v0, v0, v8, v6
; GFX10-NEXT: v_or3_b32 v1, v1, v4, v7
; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off ; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <8 x i8>, <8 x i8> addrspace(1)* %ptr %vec = load <8 x i8>, <8 x i8> addrspace(1)* %ptr
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
store <8 x i8> %insert, <8 x i8> addrspace(1)* null store <8 x i8> %insert, <8 x i8> addrspace(1)* null
ret void ret void
} }
▲ Show 20 Lines • Show All 870 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshlrev_b32_sdwa v5, s0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v5, s0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v4, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v4, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v13, v0, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v13, v0, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v14, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v14, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v2
; GFX10-NEXT: v_and_or_b32 v0, v0, s4, v5 ; GFX10-NEXT: v_and_or_b32 v0, v0, s4, v5
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6 ; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v3
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8
; GFX10-NEXT: v_and_or_b32 v1, v1, s4, v7 ; GFX10-NEXT: v_and_or_b32 v1, v1, s4, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8
; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v4, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v4, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v15, v2, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v15, v2, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v10
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v3
; GFX10-NEXT: v_or3_b32 v0, v0, v13, v6 ; GFX10-NEXT: v_or3_b32 v0, v0, v13, v6
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v10
; GFX10-NEXT: v_and_or_b32 v2, v2, s4, v9 ; GFX10-NEXT: v_and_or_b32 v2, v2, s4, v9
; GFX10-NEXT: v_lshlrev_b32_sdwa v10, v4, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v4, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_or3_b32 v7, v1, v14, v8 ; GFX10-NEXT: v_or3_b32 v1, v1, v14, v8
; GFX10-NEXT: v_and_b32_sdwa v16, v3, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v16, v3, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v12 ; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v12
; GFX10-NEXT: v_or3_b32 v2, v2, v15, v5 ; GFX10-NEXT: v_or3_b32 v2, v2, v15, v5
; GFX10-NEXT: v_and_or_b32 v3, v3, s4, v10 ; GFX10-NEXT: v_and_or_b32 v3, v3, s4, v7
; GFX10-NEXT: v_cndmask_b32_e32 v5, v0, v7, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v0, v1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, s5, 2 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, s5, 2
; GFX10-NEXT: v_or3_b32 v3, v3, v16, v6 ; GFX10-NEXT: v_or3_b32 v3, v3, v16, v6
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v3, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v3, s1
; GFX10-NEXT: v_and_or_b32 v5, v5, s3, s2 ; GFX10-NEXT: v_and_or_b32 v5, v5, s3, s2
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, s5, 0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, s5, 0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v5, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v5, s0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v5, s2
; GFX10-NEXT: v_cndmask_b32_e32 v1, v7, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v5, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v5, s1
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 8, v2 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v5, s2
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 8, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v0
; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v4, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v5, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v8, 24, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v3
; GFX10-NEXT: v_lshlrev_b32_sdwa v5, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v4, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v4, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v4, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v4, v4, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v4, v4, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v3
; GFX10-NEXT: v_and_b32_sdwa v13, v0, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v13, v0, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_or_b32 v0, v0, s4, v5
; GFX10-NEXT: v_and_b32_sdwa v15, v2, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10
; GFX10-NEXT: v_and_or_b32 v5, v2, s4, v9
; GFX10-NEXT: v_and_b32_sdwa v14, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v14, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6 ; GFX10-NEXT: v_and_b32_sdwa v15, v2, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v16, v3, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v16, v3, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: v_and_or_b32 v0, v0, s4, v5
; GFX10-NEXT: v_and_or_b32 v3, v3, s4, v4
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8 ; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8
; GFX10-NEXT: v_and_or_b32 v1, v1, s4, v7 ; GFX10-NEXT: v_and_or_b32 v1, v1, s4, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12 ; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10
; GFX10-NEXT: v_and_or_b32 v18, v3, s4, v4 ; GFX10-NEXT: v_and_or_b32 v2, v2, s4, v9
; GFX10-NEXT: v_or3_b32 v2, v5, v15, v10
; GFX10-NEXT: v_mov_b32_e32 v4, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: v_or3_b32 v0, v0, v13, v6 ; GFX10-NEXT: v_or3_b32 v0, v0, v13, v6
; GFX10-NEXT: v_or3_b32 v1, v1, v14, v8 ; GFX10-NEXT: v_or3_b32 v1, v1, v14, v8
; GFX10-NEXT: v_or3_b32 v3, v18, v16, v11 ; GFX10-NEXT: v_or3_b32 v3, v3, v16, v11
; GFX10-NEXT: v_or3_b32 v2, v2, v15, v10
; GFX10-NEXT: v_mov_b32_e32 v5, 0 ; GFX10-NEXT: v_mov_b32_e32 v5, 0
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off ; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <16 x i8>, <16 x i8> addrspace(1 )* %ptr %vec = load <16 x i8>, <16 x i8> addrspace(1 )* %ptr
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
store <16 x i8> %insert, <16 x i8> addrspace(1)* null store <16 x i8> %insert, <16 x i8> addrspace(1)* null
ret void ret void
} }
▲ Show 20 Lines • Show All 433 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_lshl_b32 s8, s5, s4 ; GFX10-NEXT: s_lshl_b32 s8, s5, s4
; GFX10-NEXT: s_andn2_b32 s6, s6, s8 ; GFX10-NEXT: s_andn2_b32 s6, s6, s8
; GFX10-NEXT: v_lshl_or_b32 v4, v0, s4, s6 ; GFX10-NEXT: v_lshl_or_b32 v4, v0, s4, s6
; GFX10-NEXT: v_mov_b32_e32 v0, s0 ; GFX10-NEXT: v_mov_b32_e32 v0, s0
; GFX10-NEXT: v_mov_b32_e32 v1, s1 ; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: v_mov_b32_e32 v2, s2 ; GFX10-NEXT: v_mov_b32_e32 v2, s2
; GFX10-NEXT: v_mov_b32_e32 v3, s3 ; GFX10-NEXT: v_mov_b32_e32 v3, s3
; GFX10-NEXT: s_mov_b32 s0, 8 ; GFX10-NEXT: s_mov_b32 s0, 8
; GFX10-NEXT: v_cndmask_b32_e32 v15, v0, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s7, 1 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s7, 1
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v15 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s7, 2 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s7, 2
; GFX10-NEXT: v_and_b32_sdwa v0, v15, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 8, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 8, v1
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s7, 3 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s7, 3
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1
; GFX10-NEXT: v_lshlrev_b32_sdwa v6, v9, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v6, v9, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v2
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v4, vcc_lo
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v15 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v2
; GFX10-NEXT: v_and_or_b32 v6, v1, s5, v6 ; GFX10-NEXT: v_and_or_b32 v6, v1, s5, v6
; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v9, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v9, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v3
; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v3
; GFX10-NEXT: v_and_b32_sdwa v1, v1, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v1, v1, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7 ; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7
; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v9, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v9, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_or_b32 v19, v15, s5, v4 ; GFX10-NEXT: v_and_or_b32 v4, v0, s5, v4
; GFX10-NEXT: v_and_b32_sdwa v0, v0, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_or_b32 v8, v2, s5, v8 ; GFX10-NEXT: v_and_or_b32 v8, v2, s5, v8
; GFX10-NEXT: v_and_b32_sdwa v15, v2, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v2, v2, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10 ; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10
; GFX10-NEXT: v_and_or_b32 v9, v3, s5, v9 ; GFX10-NEXT: v_and_or_b32 v9, v3, s5, v9
; GFX10-NEXT: v_and_b32_sdwa v14, v3, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v3, v3, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12 ; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12
; GFX10-NEXT: v_or3_b32 v0, v19, v0, v5 ; GFX10-NEXT: v_or3_b32 v0, v4, v0, v5
; GFX10-NEXT: v_mov_b32_e32 v4, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: v_or3_b32 v1, v6, v1, v7 ; GFX10-NEXT: v_or3_b32 v1, v6, v1, v7
; GFX10-NEXT: v_or3_b32 v2, v8, v15, v10 ; GFX10-NEXT: v_or3_b32 v2, v8, v2, v10
; GFX10-NEXT: v_or3_b32 v3, v9, v14, v11 ; GFX10-NEXT: v_or3_b32 v3, v9, v3, v11
; GFX10-NEXT: v_mov_b32_e32 v5, 0 ; GFX10-NEXT: v_mov_b32_e32 v5, 0
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off ; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <16 x i8>, <16 x i8> addrspace(4)* %ptr %vec = load <16 x i8>, <16 x i8> addrspace(4)* %ptr
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
store <16 x i8> %insert, <16 x i8> addrspace(1)* null store <16 x i8> %insert, <16 x i8> addrspace(1)* null
ret void ret void
} }
▲ Show 20 Lines • Show All 439 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshlrev_b32_e64 v0, v0, s2 ; GFX10-NEXT: v_lshlrev_b32_e64 v0, v0, s2
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v4 ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s11, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s11, s1
; GFX10-NEXT: v_and_or_b32 v5, v1, v2, v0 ; GFX10-NEXT: v_and_or_b32 v5, v1, v2, v0
; GFX10-NEXT: v_mov_b32_e32 v0, s8 ; GFX10-NEXT: v_mov_b32_e32 v0, s8
; GFX10-NEXT: v_mov_b32_e32 v1, s9 ; GFX10-NEXT: v_mov_b32_e32 v1, s9
; GFX10-NEXT: v_mov_b32_e32 v2, s10 ; GFX10-NEXT: v_mov_b32_e32 v2, s10
; GFX10-NEXT: v_mov_b32_e32 v3, s11 ; GFX10-NEXT: v_mov_b32_e32 v3, s11
; GFX10-NEXT: v_cndmask_b32_e64 v15, v0, v5, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v5, s2
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v5, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v5, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v5, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v5, s1
; GFX10-NEXT: s_mov_b32 s2, 8 ; GFX10-NEXT: s_mov_b32 s2, 8
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v15 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 8, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 8, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v15 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v0
; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v2
; GFX10-NEXT: v_lshlrev_b32_sdwa v6, v9, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v6, v9, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v9, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v9, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v9, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v9, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v3
; GFX10-NEXT: v_and_or_b32 v19, v15, s5, v4 ; GFX10-NEXT: v_and_or_b32 v4, v0, s5, v4
; GFX10-NEXT: v_and_b32_sdwa v0, v15, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v0, v0, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5
; GFX10-NEXT: v_and_or_b32 v6, v1, s5, v6 ; GFX10-NEXT: v_and_or_b32 v6, v1, s5, v6
; GFX10-NEXT: v_and_or_b32 v8, v2, s5, v8
; GFX10-NEXT: v_and_or_b32 v9, v3, s5, v9
; GFX10-NEXT: v_and_b32_sdwa v1, v1, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v1, v1, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7 ; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7
; GFX10-NEXT: v_and_or_b32 v8, v2, s5, v8 ; GFX10-NEXT: v_and_b32_sdwa v2, v2, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v15, v2, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10 ; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10
; GFX10-NEXT: v_and_or_b32 v9, v3, s5, v9 ; GFX10-NEXT: v_and_b32_sdwa v3, v3, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v14, v3, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12 ; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12
; GFX10-NEXT: v_or3_b32 v0, v19, v0, v5 ; GFX10-NEXT: v_or3_b32 v0, v4, v0, v5
; GFX10-NEXT: v_mov_b32_e32 v4, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: v_or3_b32 v1, v6, v1, v7 ; GFX10-NEXT: v_or3_b32 v1, v6, v1, v7
; GFX10-NEXT: v_or3_b32 v2, v8, v15, v10 ; GFX10-NEXT: v_or3_b32 v2, v8, v2, v10
; GFX10-NEXT: v_or3_b32 v3, v9, v14, v11 ; GFX10-NEXT: v_or3_b32 v3, v9, v3, v11
; GFX10-NEXT: v_mov_b32_e32 v5, 0 ; GFX10-NEXT: v_mov_b32_e32 v5, 0
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off ; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <16 x i8>, <16 x i8> addrspace(4)* %ptr %vec = load <16 x i8>, <16 x i8> addrspace(4)* %ptr
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
store <16 x i8> %insert, <16 x i8> addrspace(1)* null store <16 x i8> %insert, <16 x i8> addrspace(1)* null
ret void ret void
} }
▲ Show 20 Lines • Show All 436 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s6, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s6, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v4 ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v4
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s7, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s7, s1
; GFX10-NEXT: v_and_or_b32 v5, v2, v1, v0 ; GFX10-NEXT: v_and_or_b32 v5, v2, v1, v0
; GFX10-NEXT: v_mov_b32_e32 v0, s4 ; GFX10-NEXT: v_mov_b32_e32 v0, s4
; GFX10-NEXT: v_mov_b32_e32 v1, s5 ; GFX10-NEXT: v_mov_b32_e32 v1, s5
; GFX10-NEXT: v_mov_b32_e32 v2, s6 ; GFX10-NEXT: v_mov_b32_e32 v2, s6
; GFX10-NEXT: v_mov_b32_e32 v3, s7 ; GFX10-NEXT: v_mov_b32_e32 v3, s7
; GFX10-NEXT: v_cndmask_b32_e64 v15, v0, v5, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v5, s2
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v5, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v5, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v5, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v5, s1
; GFX10-NEXT: s_mov_b32 s2, 8 ; GFX10-NEXT: s_mov_b32 s2, 8
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v15 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 8, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 8, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v15 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v0
; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v2
; GFX10-NEXT: v_lshlrev_b32_sdwa v6, v9, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v6, v9, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v9, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v9, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v9, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v9, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v3
; GFX10-NEXT: v_and_or_b32 v19, v15, s8, v4 ; GFX10-NEXT: v_and_or_b32 v4, v0, s8, v4
; GFX10-NEXT: v_and_b32_sdwa v0, v15, s8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v0, v0, s8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v5
; GFX10-NEXT: v_and_or_b32 v6, v1, s8, v6 ; GFX10-NEXT: v_and_or_b32 v6, v1, s8, v6
; GFX10-NEXT: v_and_or_b32 v8, v2, s8, v8
; GFX10-NEXT: v_and_or_b32 v9, v3, s8, v9
; GFX10-NEXT: v_and_b32_sdwa v1, v1, s8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v1, v1, s8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7 ; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7
; GFX10-NEXT: v_and_or_b32 v8, v2, s8, v8 ; GFX10-NEXT: v_and_b32_sdwa v2, v2, s8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v15, v2, s8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10 ; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10
; GFX10-NEXT: v_and_or_b32 v9, v3, s8, v9 ; GFX10-NEXT: v_and_b32_sdwa v3, v3, s8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v14, v3, s8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12 ; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12
; GFX10-NEXT: v_or3_b32 v0, v19, v0, v5 ; GFX10-NEXT: v_or3_b32 v0, v4, v0, v5
; GFX10-NEXT: v_mov_b32_e32 v4, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: v_or3_b32 v1, v6, v1, v7 ; GFX10-NEXT: v_or3_b32 v1, v6, v1, v7
; GFX10-NEXT: v_or3_b32 v2, v8, v15, v10 ; GFX10-NEXT: v_or3_b32 v2, v8, v2, v10
; GFX10-NEXT: v_or3_b32 v3, v9, v14, v11 ; GFX10-NEXT: v_or3_b32 v3, v9, v3, v11
; GFX10-NEXT: v_mov_b32_e32 v5, 0 ; GFX10-NEXT: v_mov_b32_e32 v5, 0
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off ; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <16 x i8>, <16 x i8> addrspace(4)* %ptr %vec = load <16 x i8>, <16 x i8> addrspace(4)* %ptr
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
store <16 x i8> %insert, <16 x i8> addrspace(1)* null store <16 x i8> %insert, <16 x i8> addrspace(1)* null
ret void ret void
} }
▲ Show 20 Lines • Show All 309 Lines • ▼ Show 20 Lines
; GFX7-NEXT: v_or_b32_e32 v0, v0, v6 ; GFX7-NEXT: v_or_b32_e32 v0, v0, v6
; GFX7-NEXT: v_or_b32_e32 v3, v3, v4 ; GFX7-NEXT: v_or_b32_e32 v3, v3, v4
; GFX7-NEXT: buffer_store_dwordx4 v[0:3], off, s[8:11], 0 ; GFX7-NEXT: buffer_store_dwordx4 v[0:3], off, s[8:11], 0
; GFX7-NEXT: s_endpgm ; GFX7-NEXT: s_endpgm
; ;
; GFX10-LABEL: insertelement_v_v16i8_s_v: ; GFX10-LABEL: insertelement_v_v16i8_s_v:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: global_load_dwordx4 v[3:6], v[0:1], off ; GFX10-NEXT: global_load_dwordx4 v[3:6], v[0:1], off
; GFX10-NEXT: v_mov_b32_e32 v22, 8 ; GFX10-NEXT: v_mov_b32_e32 v1, 8
; GFX10-NEXT: s_mov_b32 s0, 8 ; GFX10-NEXT: s_mov_b32 s0, 8
; GFX10-NEXT: s_movk_i32 s3, 0xff ; GFX10-NEXT: s_movk_i32 s3, 0xff
; GFX10-NEXT: v_and_b32_e32 v0, 3, v2 ; GFX10-NEXT: v_and_b32_e32 v0, 3, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v2, 2, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v2, 2, v2
; GFX10-NEXT: s_and_b32 s1, s2, s3 ; GFX10-NEXT: s_and_b32 s1, s2, s3
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2 ; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v2 ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v2
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 8, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 8, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 24, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v8, 24, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v4 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v5 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v5
; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v22, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, s0, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v26, s0, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v1, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v5
; GFX10-NEXT: v_lshrrev_b32_e32 v13, 8, v6
; GFX10-NEXT: v_lshlrev_b32_sdwa v11, v22, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_or_b32 v19, v4, s3, v9
; GFX10-NEXT: v_and_b32_sdwa v15, v3, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v15, v3, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v16, v4, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v16, v4, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v5
; GFX10-NEXT: v_and_or_b32 v3, v3, s3, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8 ; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8
; GFX10-NEXT: v_and_or_b32 v26, v3, s3, v26 ; GFX10-NEXT: v_lshrrev_b32_e32 v13, 8, v6
; GFX10-NEXT: v_and_or_b32 v4, v4, s3, v9
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10 ; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10
; GFX10-NEXT: v_lshrrev_b32_e32 v14, 24, v6 ; GFX10-NEXT: v_lshlrev_b32_sdwa v11, v1, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v17, v5, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v17, v5, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshrrev_b32_e32 v14, 24, v6
; GFX10-NEXT: v_or3_b32 v3, v3, v15, v8
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v12 ; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v12
; GFX10-NEXT: v_and_or_b32 v30, v5, s3, v11 ; GFX10-NEXT: v_and_or_b32 v5, v5, s3, v11
; GFX10-NEXT: v_or3_b32 v3, v26, v15, v8 ; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v1, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v22, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_or3_b32 v4, v4, v16, v10
; GFX10-NEXT: v_or3_b32 v26, v19, v16, v10
; GFX10-NEXT: v_and_b32_sdwa v18, v6, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v18, v6, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_or3_b32 v5, v30, v17, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v14 ; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v14
; GFX10-NEXT: v_and_or_b32 v11, v6, s3, v9 ; GFX10-NEXT: v_or3_b32 v5, v5, v17, v7
; GFX10-NEXT: v_cndmask_b32_e32 v7, v3, v26, vcc_lo ; GFX10-NEXT: v_and_or_b32 v6, v6, s3, v9
; GFX10-NEXT: v_cndmask_b32_e32 v7, v3, v4, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v2 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v2
; GFX10-NEXT: v_lshlrev_b32_e64 v9, v0, s3 ; GFX10-NEXT: v_lshlrev_b32_e64 v9, v0, s3
; GFX10-NEXT: v_lshlrev_b32_e64 v0, v0, s1 ; GFX10-NEXT: v_lshlrev_b32_e64 v0, v0, s1
; GFX10-NEXT: v_or3_b32 v6, v11, v18, v8 ; GFX10-NEXT: v_or3_b32 v6, v6, v18, v8
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v2 ; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v2
; GFX10-NEXT: v_cndmask_b32_e64 v7, v7, v5, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v7, v7, v5, s0
; GFX10-NEXT: v_xor_b32_e32 v10, -1, v9 ; GFX10-NEXT: v_xor_b32_e32 v8, -1, v9
; GFX10-NEXT: v_cndmask_b32_e64 v7, v7, v6, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v7, v7, v6, s1
; GFX10-NEXT: v_and_or_b32 v0, v7, v10, v0 ; GFX10-NEXT: v_and_or_b32 v0, v7, v8, v0
; GFX10-NEXT: v_cndmask_b32_e32 v18, v26, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v4, v5, v0, s0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v3, v0, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v3, v0, s2
; GFX10-NEXT: v_cndmask_b32_e32 v3, v4, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v4, v5, v0, s0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v6, v0, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v6, v0, s1
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v18
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 8, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 8, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 24, v18 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 24, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v4 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v4
; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v22, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v5, v22, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v0
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, v22, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v5, v1, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v22, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v1, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v2 ; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v1, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, v1, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v13, v2, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v14, v3, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v15, v4, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v15, v4, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_or_b32 v19, v2, s3, v5 ; GFX10-NEXT: v_and_or_b32 v2, v2, s3, v5
; GFX10-NEXT: v_and_b32_sdwa v14, v18, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v16, v0, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v16, v0, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: v_and_or_b32 v5, v0, s3, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8 ; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8
; GFX10-NEXT: v_and_or_b32 v3, v3, s3, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10 ; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12 ; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12
; GFX10-NEXT: v_and_or_b32 v3, v18, s3, v7
; GFX10-NEXT: v_and_or_b32 v5, v0, s3, v1
; GFX10-NEXT: v_and_or_b32 v4, v4, s3, v9 ; GFX10-NEXT: v_and_or_b32 v4, v4, s3, v9
; GFX10-NEXT: v_and_b32_sdwa v13, v2, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_or3_b32 v0, v2, v13, v6
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: v_or3_b32 v1, v3, v14, v8 ; GFX10-NEXT: v_or3_b32 v1, v3, v14, v8
; GFX10-NEXT: v_or3_b32 v3, v5, v16, v11 ; GFX10-NEXT: v_or3_b32 v3, v5, v16, v11
; GFX10-NEXT: v_or3_b32 v2, v4, v15, v10 ; GFX10-NEXT: v_or3_b32 v2, v4, v15, v10
; GFX10-NEXT: v_mov_b32_e32 v4, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: v_or3_b32 v0, v19, v13, v6
; GFX10-NEXT: v_mov_b32_e32 v5, 0 ; GFX10-NEXT: v_mov_b32_e32 v5, 0
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off ; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <16 x i8>, <16 x i8> addrspace(1)* %ptr %vec = load <16 x i8>, <16 x i8> addrspace(1)* %ptr
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
store <16 x i8> %insert, <16 x i8> addrspace(1)* null store <16 x i8> %insert, <16 x i8> addrspace(1)* null
ret void ret void
} }
▲ Show 20 Lines • Show All 308 Lines • ▼ Show 20 Lines
; GFX7-NEXT: v_or_b32_e32 v0, v0, v6 ; GFX7-NEXT: v_or_b32_e32 v0, v0, v6
; GFX7-NEXT: v_or_b32_e32 v3, v3, v4 ; GFX7-NEXT: v_or_b32_e32 v3, v3, v4
; GFX7-NEXT: buffer_store_dwordx4 v[0:3], off, s[8:11], 0 ; GFX7-NEXT: buffer_store_dwordx4 v[0:3], off, s[8:11], 0
; GFX7-NEXT: s_endpgm ; GFX7-NEXT: s_endpgm
; ;
; GFX10-LABEL: insertelement_v_v16i8_v_s: ; GFX10-LABEL: insertelement_v_v16i8_v_s:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: global_load_dwordx4 v[3:6], v[0:1], off ; GFX10-NEXT: global_load_dwordx4 v[3:6], v[0:1], off
; GFX10-NEXT: v_mov_b32_e32 v18, 8 ; GFX10-NEXT: v_mov_b32_e32 v0, 8
; GFX10-NEXT: s_mov_b32 s0, 8 ; GFX10-NEXT: s_mov_b32 s0, 8
; GFX10-NEXT: s_movk_i32 s3, 0xff ; GFX10-NEXT: s_movk_i32 s3, 0xff
; GFX10-NEXT: s_lshr_b32 s4, s2, 2 ; GFX10-NEXT: s_lshr_b32 s4, s2, 2
; GFX10-NEXT: s_and_b32 s1, s2, 3 ; GFX10-NEXT: s_and_b32 s1, s2, 3
; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s4, 1 ; GFX10-NEXT: v_cmp_eq_u32_e64 vcc_lo, s4, 1
; GFX10-NEXT: s_lshl_b32 s2, s1, 3 ; GFX10-NEXT: s_lshl_b32 s2, s1, 3
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, s4, 3 ; GFX10-NEXT: v_cmp_eq_u32_e64 s1, s4, 3
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_lshl_b32 s2, s3, s2 ; GFX10-NEXT: s_lshl_b32 s2, s3, s2
; GFX10-NEXT: s_not_b32 s2, s2 ; GFX10-NEXT: s_not_b32 s2, s2
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v4 ; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 24, v4 ; GFX10-NEXT: v_lshrrev_b32_e32 v9, 24, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v27, 8, v5 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 8, v5
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v19, v18, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v0, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 8, v6
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 24, v5 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 24, v5
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 8, v6
; GFX10-NEXT: v_and_b32_sdwa v14, v3, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v14, v3, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v15, v4, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_or_b32 v22, v4, s3, v19
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7
; GFX10-NEXT: v_and_or_b32 v1, v3, s3, v1 ; GFX10-NEXT: v_and_or_b32 v1, v3, s3, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v7
; GFX10-NEXT: v_and_or_b32 v3, v4, s3, v8
; GFX10-NEXT: v_and_b32_sdwa v15, v4, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9 ; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9
; GFX10-NEXT: v_lshlrev_b32_sdwa v10, v18, v27 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v10, v0, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v23, v5, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v16, v5, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshrrev_b32_e32 v13, 24, v6 ; GFX10-NEXT: v_lshrrev_b32_e32 v13, 24, v6
; GFX10-NEXT: v_or3_b32 v1, v1, v14, v7 ; GFX10-NEXT: v_or3_b32 v1, v1, v14, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v11 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 24, v11
; GFX10-NEXT: v_and_or_b32 v5, v5, s3, v10 ; GFX10-NEXT: v_and_or_b32 v5, v5, s3, v10
; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v18, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v0, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_or3_b32 v3, v22, v15, v9 ; GFX10-NEXT: v_or3_b32 v3, v3, v15, v9
; GFX10-NEXT: v_and_b32_sdwa v17, v6, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v17, v6, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v13 ; GFX10-NEXT: v_lshlrev_b32_e32 v7, 24, v13
; GFX10-NEXT: v_or3_b32 v4, v5, v23, v4 ; GFX10-NEXT: v_or3_b32 v4, v5, v16, v4
; GFX10-NEXT: v_and_or_b32 v6, v6, s3, v8 ; GFX10-NEXT: v_and_or_b32 v6, v6, s3, v8
; GFX10-NEXT: v_cndmask_b32_e32 v5, v1, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v5, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, s4, 2 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, s4, 2
; GFX10-NEXT: v_or3_b32 v7, v6, v17, v7 ; GFX10-NEXT: v_or3_b32 v6, v6, v17, v7
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v4, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v4, s0
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v7, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v6, s1
; GFX10-NEXT: v_and_or_b32 v2, v5, s2, v2 ; GFX10-NEXT: v_and_or_b32 v2, v5, s2, v2
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, s4, 0 ; GFX10-NEXT: v_cmp_eq_u32_e64 s2, s4, 0
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v19, v1, v2, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v2, s2
; GFX10-NEXT: v_cndmask_b32_e32 v22, v3, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v2, s1
; GFX10-NEXT: v_cndmask_b32_e64 v2, v7, v2, s1 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 8, v4 ; GFX10-NEXT: v_lshrrev_b32_e32 v9, 8, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v19 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v22
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 8, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 24, v22 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 24, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v4 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 24, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v2
; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v18, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v5, v0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v0, v18, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v0, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v5, v18, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v0, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v9, v18, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v0, v0, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v19 ; GFX10-NEXT: v_and_b32_sdwa v13, v1, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v13, v19, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v14, v3, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v15, v4, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v15, v4, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_or_b32 v19, v19, s3, v5 ; GFX10-NEXT: v_and_or_b32 v1, v1, s3, v5
; GFX10-NEXT: v_and_b32_sdwa v14, v22, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v16, v2, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v16, v2, s3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: v_and_or_b32 v5, v2, s3, v0
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8 ; GFX10-NEXT: v_lshlrev_b32_e32 v8, 24, v8
; GFX10-NEXT: v_and_or_b32 v3, v22, s3, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10 ; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12
; GFX10-NEXT: v_and_or_b32 v4, v4, s3, v9 ; GFX10-NEXT: v_and_or_b32 v4, v4, s3, v9
; GFX10-NEXT: v_and_or_b32 v5, v2, s3, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v12
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6 ; GFX10-NEXT: v_and_or_b32 v3, v3, s3, v7
; GFX10-NEXT: v_or3_b32 v1, v3, v14, v8 ; GFX10-NEXT: v_or3_b32 v0, v1, v13, v6
; GFX10-NEXT: v_or3_b32 v2, v4, v15, v10 ; GFX10-NEXT: v_or3_b32 v2, v4, v15, v10
; GFX10-NEXT: v_or3_b32 v1, v3, v14, v8
; GFX10-NEXT: v_or3_b32 v3, v5, v16, v11 ; GFX10-NEXT: v_or3_b32 v3, v5, v16, v11
; GFX10-NEXT: v_mov_b32_e32 v4, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: v_or3_b32 v0, v19, v13, v6
; GFX10-NEXT: v_mov_b32_e32 v5, 0 ; GFX10-NEXT: v_mov_b32_e32 v5, 0
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off ; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <16 x i8>, <16 x i8> addrspace(1)* %ptr %vec = load <16 x i8>, <16 x i8> addrspace(1)* %ptr
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
store <16 x i8> %insert, <16 x i8> addrspace(1)* null store <16 x i8> %insert, <16 x i8> addrspace(1)* null
ret void ret void
} }
▲ Show 20 Lines • Show All 331 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v5 ; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v5
; GFX10-NEXT: v_lshrrev_b32_e32 v13, 8, v6 ; GFX10-NEXT: v_lshrrev_b32_e32 v13, 8, v6
; GFX10-NEXT: v_lshlrev_b32_sdwa v9, s0, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v9, s0, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v11, v8, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v11, v8, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v17, v4, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v17, v4, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_sdwa v18, v5, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v18, v5, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshrrev_b32_e32 v14, 24, v6 ; GFX10-NEXT: v_lshrrev_b32_e32 v14, 24, v6
; GFX10-NEXT: v_and_or_b32 v4, v4, s1, v9 ; GFX10-NEXT: v_and_or_b32 v4, v4, s1, v9
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10
; GFX10-NEXT: v_lshrrev_b32_e32 v15, 8, v7
; GFX10-NEXT: v_and_or_b32 v5, v5, s1, v11 ; GFX10-NEXT: v_and_or_b32 v5, v5, s1, v11
; GFX10-NEXT: v_lshlrev_b32_e32 v12, 24, v12 ; GFX10-NEXT: v_lshlrev_b32_e32 v12, 24, v12
; GFX10-NEXT: v_lshrrev_b32_e32 v15, 8, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v10
; GFX10-NEXT: v_lshlrev_b32_sdwa v13, v8, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v13, v8, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v19, v6, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v19, v6, s1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v14
; GFX10-NEXT: v_lshlrev_b32_sdwa v14, v8, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshrrev_b32_e32 v16, 24, v7 ; GFX10-NEXT: v_lshrrev_b32_e32 v16, 24, v7
; GFX10-NEXT: v_or3_b32 v4, v4, v17, v10
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v14
; GFX10-NEXT: v_and_or_b32 v6, v6, s1, v13 ; GFX10-NEXT: v_and_or_b32 v6, v6, s1, v13
; GFX10-NEXT: v_or3_b32 v15, v4, v17, v10 ; GFX10-NEXT: v_lshlrev_b32_sdwa v11, v8, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_or3_b32 v5, v5, v18, v12 ; GFX10-NEXT: v_or3_b32 v5, v5, v18, v12
; GFX10-NEXT: v_and_b32_sdwa v20, v7, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v20, v7, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v16 ; GFX10-NEXT: v_lshlrev_b32_e32 v10, 24, v16
; GFX10-NEXT: v_and_or_b32 v7, v7, v1, v14
; GFX10-NEXT: v_or3_b32 v6, v6, v19, v9 ; GFX10-NEXT: v_or3_b32 v6, v6, v19, v9
; GFX10-NEXT: v_cndmask_b32_e32 v9, v15, v5, vcc_lo ; GFX10-NEXT: v_and_or_b32 v7, v7, v1, v11
; GFX10-NEXT: v_cndmask_b32_e32 v9, v4, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v3 ; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v3
; GFX10-NEXT: v_lshlrev_b32_e32 v11, v0, v1 ; GFX10-NEXT: v_lshlrev_b32_e32 v11, v0, v1
; GFX10-NEXT: v_or3_b32 v7, v7, v20, v10
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v3 ; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v3
; GFX10-NEXT: v_or3_b32 v7, v7, v20, v10
; GFX10-NEXT: v_lshlrev_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_cndmask_b32_e64 v9, v9, v6, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v9, v9, v6, s0
; GFX10-NEXT: v_xor_b32_e32 v2, -1, v11 ; GFX10-NEXT: v_xor_b32_e32 v2, -1, v11
; GFX10-NEXT: v_cndmask_b32_e64 v9, v9, v7, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v9, v9, v7, s1
; GFX10-NEXT: v_and_or_b32 v0, v9, v2, v0 ; GFX10-NEXT: v_and_or_b32 v0, v9, v2, v0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v4, v0, s2
; GFX10-NEXT: v_cndmask_b32_e64 v4, v6, v0, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v6, v0, s0
; GFX10-NEXT: v_cndmask_b32_e64 v27, v7, v0, s1 ; GFX10-NEXT: v_cndmask_b32_e32 v3, v5, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v15, v0, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v7, v0, s1
; GFX10-NEXT: v_cndmask_b32_e32 v18, v5, v0, vcc_lo
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 8, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v19, 8, v27
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v18 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 8, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 24, v18 ; GFX10-NEXT: v_lshrrev_b32_e32 v9, 24, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 24, v4 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 24, v4
; GFX10-NEXT: v_lshlrev_b32_sdwa v10, v8, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshrrev_b32_e32 v13, 24, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v13, 24, v27 ; GFX10-NEXT: v_lshlrev_b32_sdwa v5, v8, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v23, v8, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v8, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v8, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v8, v19 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v10, v8, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v21, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_lshlrev_b32_sdwa v8, v8, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_sdwa v15, v18, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v14, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_or_b32 v19, v2, v1, v23 ; GFX10-NEXT: v_and_b32_sdwa v15, v3, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: v_and_b32_sdwa v16, v4, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v16, v4, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_or_b32 v10, v4, v1, v10 ; GFX10-NEXT: v_and_or_b32 v2, v2, v1, v5
; GFX10-NEXT: v_and_b32_sdwa v17, v27, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-NEXT: v_and_b32_sdwa v17, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: v_and_or_b32 v5, v0, v1, v8
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9 ; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v11 ; GFX10-NEXT: v_lshlrev_b32_e32 v11, 24, v11
; GFX10-NEXT: v_lshlrev_b32_e32 v12, 24, v13 ; GFX10-NEXT: v_lshlrev_b32_e32 v12, 24, v13
; GFX10-NEXT: v_and_or_b32 v3, v27, v1, v8 ; GFX10-NEXT: v_and_or_b32 v3, v3, v1, v7
; GFX10-NEXT: v_and_or_b32 v2, v18, v1, v7 ; GFX10-NEXT: v_and_or_b32 v4, v4, v1, v10
; GFX10-NEXT: v_or3_b32 v0, v2, v14, v6
; GFX10-NEXT: v_or3_b32 v1, v3, v15, v9
; GFX10-NEXT: v_or3_b32 v2, v4, v16, v11
; GFX10-NEXT: v_or3_b32 v3, v5, v17, v12
; GFX10-NEXT: v_mov_b32_e32 v4, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: v_or3_b32 v0, v19, v21, v6
; GFX10-NEXT: v_mov_b32_e32 v5, 0 ; GFX10-NEXT: v_mov_b32_e32 v5, 0
; GFX10-NEXT: v_or3_b32 v3, v3, v17, v12
; GFX10-NEXT: v_or3_b32 v1, v2, v15, v9
; GFX10-NEXT: v_or3_b32 v2, v10, v16, v11
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off ; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%vec = load <16 x i8>, <16 x i8> addrspace(1)* %ptr %vec = load <16 x i8>, <16 x i8> addrspace(1)* %ptr
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx %insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
store <16 x i8> %insert, <16 x i8> addrspace(1)* null store <16 x i8> %insert, <16 x i8> addrspace(1)* null
ret void ret void
} }

llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll

Show First 20 Lines • Show All 983 Lines • ▼ Show 20 Lines
; MOVREL-NEXT: s_mov_b32 s4, s6 ; MOVREL-NEXT: s_mov_b32 s4, s6
; MOVREL-NEXT: s_mov_b32 s6, s8 ; MOVREL-NEXT: s_mov_b32 s6, s8
; MOVREL-NEXT: s_mov_b32 s8, s10 ; MOVREL-NEXT: s_mov_b32 s8, s10
; MOVREL-NEXT: s_mov_b32 s10, s12 ; MOVREL-NEXT: s_mov_b32 s10, s12
; MOVREL-NEXT: s_mov_b32 s12, s14 ; MOVREL-NEXT: s_mov_b32 s12, s14
; MOVREL-NEXT: s_mov_b32 s14, s16 ; MOVREL-NEXT: s_mov_b32 s14, s16
; MOVREL-NEXT: v_mov_b32_e32 v16, s15 ; MOVREL-NEXT: v_mov_b32_e32 v16, s15
; MOVREL-NEXT: v_mov_b32_e32 v2, s1 ; MOVREL-NEXT: v_mov_b32_e32 v2, s1
; MOVREL-NEXT: v_mov_b32_e32 v1, s0
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
; MOVREL-NEXT: v_mov_b32_e32 v1, s0
; MOVREL-NEXT: v_mov_b32_e32 v15, s14 ; MOVREL-NEXT: v_mov_b32_e32 v15, s14
; MOVREL-NEXT: v_mov_b32_e32 v14, s13 ; MOVREL-NEXT: v_mov_b32_e32 v14, s13
; MOVREL-NEXT: v_mov_b32_e32 v13, s12 ; MOVREL-NEXT: v_mov_b32_e32 v13, s12
; MOVREL-NEXT: v_mov_b32_e32 v12, s11 ; MOVREL-NEXT: v_mov_b32_e32 v12, s11
; MOVREL-NEXT: v_mov_b32_e32 v11, s10 ; MOVREL-NEXT: v_mov_b32_e32 v11, s10
; MOVREL-NEXT: v_mov_b32_e32 v10, s9 ; MOVREL-NEXT: v_mov_b32_e32 v10, s9
; MOVREL-NEXT: v_mov_b32_e32 v9, s8 ; MOVREL-NEXT: v_mov_b32_e32 v9, s8
; MOVREL-NEXT: v_mov_b32_e32 v8, s7 ; MOVREL-NEXT: v_mov_b32_e32 v8, s7
; MOVREL-NEXT: v_mov_b32_e32 v7, s6 ; MOVREL-NEXT: v_mov_b32_e32 v7, s6
; MOVREL-NEXT: v_mov_b32_e32 v6, s5 ; MOVREL-NEXT: v_mov_b32_e32 v6, s5
; MOVREL-NEXT: v_mov_b32_e32 v5, s4 ; MOVREL-NEXT: v_mov_b32_e32 v5, s4
; MOVREL-NEXT: v_mov_b32_e32 v4, s3 ; MOVREL-NEXT: v_mov_b32_e32 v4, s3
; MOVREL-NEXT: v_mov_b32_e32 v3, s2 ; MOVREL-NEXT: v_mov_b32_e32 v3, s2
; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, 1, v0 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, 1, v0
; MOVREL-NEXT: s_mov_b32 s30, s18
; MOVREL-NEXT: s_mov_b32 s31, s19
; MOVREL-NEXT: v_cmp_eq_u32_e64 s1, 2, v0 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s1, 2, v0
; MOVREL-NEXT: v_cndmask_b32_e64 v1, v1, s30, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e64 v1, v1, s18, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e64 v2, v2, s31, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e64 v2, v2, s19, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; MOVREL-NEXT: v_cmp_eq_u32_e64 s2, 5, v0 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s2, 5, v0
; MOVREL-NEXT: v_cndmask_b32_e64 v3, v3, s30, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v3, v3, s18, s0
; MOVREL-NEXT: v_cndmask_b32_e64 v4, v4, s31, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v4, v4, s19, s0
; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, 4, v0 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, 4, v0
; MOVREL-NEXT: v_cmp_eq_u32_e64 s3, 6, v0 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s3, 6, v0
; MOVREL-NEXT: v_cmp_eq_u32_e64 s4, 7, v0 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s4, 7, v0
; MOVREL-NEXT: v_cndmask_b32_e64 v5, v5, s30, s1 ; MOVREL-NEXT: v_cndmask_b32_e64 v5, v5, s18, s1
; MOVREL-NEXT: v_cndmask_b32_e64 v6, v6, s31, s1 ; MOVREL-NEXT: v_cndmask_b32_e64 v6, v6, s19, s1
; MOVREL-NEXT: v_cndmask_b32_e64 v7, v7, s30, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e64 v7, v7, s18, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e64 v8, v8, s31, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e64 v8, v8, s19, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e64 v9, v9, s30, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v9, v9, s18, s0
; MOVREL-NEXT: v_cndmask_b32_e64 v10, v10, s31, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v10, v10, s19, s0
; MOVREL-NEXT: v_cndmask_b32_e64 v11, v11, s30, s2 ; MOVREL-NEXT: v_cndmask_b32_e64 v11, v11, s18, s2
; MOVREL-NEXT: v_cndmask_b32_e64 v12, v12, s31, s2 ; MOVREL-NEXT: v_cndmask_b32_e64 v12, v12, s19, s2
; MOVREL-NEXT: v_cndmask_b32_e64 v13, v13, s30, s3 ; MOVREL-NEXT: v_cndmask_b32_e64 v13, v13, s18, s3
; MOVREL-NEXT: v_cndmask_b32_e64 v14, v14, s31, s3 ; MOVREL-NEXT: v_cndmask_b32_e64 v14, v14, s19, s3
; MOVREL-NEXT: v_cndmask_b32_e64 v15, v15, s30, s4 ; MOVREL-NEXT: v_cndmask_b32_e64 v15, v15, s18, s4
; MOVREL-NEXT: v_cndmask_b32_e64 v16, v16, s31, s4 ; MOVREL-NEXT: v_cndmask_b32_e64 v16, v16, s19, s4
; MOVREL-NEXT: global_store_dwordx4 v[0:1], v[1:4], off ; MOVREL-NEXT: global_store_dwordx4 v[0:1], v[1:4], off
; MOVREL-NEXT: s_waitcnt_vscnt null, 0x0 ; MOVREL-NEXT: s_waitcnt_vscnt null, 0x0
; MOVREL-NEXT: global_store_dwordx4 v[0:1], v[5:8], off ; MOVREL-NEXT: global_store_dwordx4 v[0:1], v[5:8], off
; MOVREL-NEXT: s_waitcnt_vscnt null, 0x0 ; MOVREL-NEXT: s_waitcnt_vscnt null, 0x0
; MOVREL-NEXT: global_store_dwordx4 v[0:1], v[9:12], off ; MOVREL-NEXT: global_store_dwordx4 v[0:1], v[9:12], off
; MOVREL-NEXT: s_waitcnt_vscnt null, 0x0 ; MOVREL-NEXT: s_waitcnt_vscnt null, 0x0
; MOVREL-NEXT: global_store_dwordx4 v[0:1], v[13:16], off ; MOVREL-NEXT: global_store_dwordx4 v[0:1], v[13:16], off
; MOVREL-NEXT: s_waitcnt_vscnt null, 0x0 ; MOVREL-NEXT: s_waitcnt_vscnt null, 0x0
▲ Show 20 Lines • Show All 480 Lines • ▼ Show 20 Lines
; GPRIDX-NEXT: global_store_dwordx4 v[0:1], v[8:11], off ; GPRIDX-NEXT: global_store_dwordx4 v[0:1], v[8:11], off
; GPRIDX-NEXT: s_waitcnt vmcnt(0) ; GPRIDX-NEXT: s_waitcnt vmcnt(0)
; GPRIDX-NEXT: global_store_dwordx4 v[0:1], v[12:15], off ; GPRIDX-NEXT: global_store_dwordx4 v[0:1], v[12:15], off
; GPRIDX-NEXT: s_waitcnt vmcnt(0) ; GPRIDX-NEXT: s_waitcnt vmcnt(0)
; GPRIDX-NEXT: s_endpgm ; GPRIDX-NEXT: s_endpgm
; ;
; MOVREL-LABEL: dyn_insertelement_v8f64_v_v_v: ; MOVREL-LABEL: dyn_insertelement_v8f64_v_v_v:
; MOVREL: ; %bb.0: ; %entry ; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, 1, v18
; MOVREL-NEXT: v_mov_b32_e32 v19, v0
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v18
; MOVREL-NEXT: v_mov_b32_e32 v23, v1 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, 1, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s1, 2, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s1, 2, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s2, 3, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s2, 3, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s3, 4, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s3, 4, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s4, 5, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s4, 5, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s5, 7, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s5, 7, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s6, 6, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s6, 6, v18
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v19, v16, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v16, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e64 v2, v2, v16, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v2, v2, v16, s0
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v23, v17, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v17, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e64 v3, v3, v17, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v3, v3, v17, s0
; MOVREL-NEXT: v_cndmask_b32_e64 v4, v4, v16, s1 ; MOVREL-NEXT: v_cndmask_b32_e64 v4, v4, v16, s1
; MOVREL-NEXT: v_cndmask_b32_e64 v6, v6, v16, s2 ; MOVREL-NEXT: v_cndmask_b32_e64 v6, v6, v16, s2
; MOVREL-NEXT: v_cndmask_b32_e64 v5, v5, v17, s1 ; MOVREL-NEXT: v_cndmask_b32_e64 v5, v5, v17, s1
; MOVREL-NEXT: v_cndmask_b32_e64 v7, v7, v17, s2 ; MOVREL-NEXT: v_cndmask_b32_e64 v7, v7, v17, s2
; MOVREL-NEXT: v_cndmask_b32_e64 v8, v8, v16, s3 ; MOVREL-NEXT: v_cndmask_b32_e64 v8, v8, v16, s3
; MOVREL-NEXT: v_cndmask_b32_e64 v10, v10, v16, s4 ; MOVREL-NEXT: v_cndmask_b32_e64 v10, v10, v16, s4
; MOVREL-NEXT: v_cndmask_b32_e64 v9, v9, v17, s3 ; MOVREL-NEXT: v_cndmask_b32_e64 v9, v9, v17, s3
▲ Show 20 Lines • Show All 607 Lines • ▼ Show 20 Lines
; GPRIDX-NEXT: s_waitcnt vmcnt(0) ; GPRIDX-NEXT: s_waitcnt vmcnt(0)
; GPRIDX-NEXT: global_store_dwordx4 v[0:1], v[12:15], off ; GPRIDX-NEXT: global_store_dwordx4 v[0:1], v[12:15], off
; GPRIDX-NEXT: s_waitcnt vmcnt(0) ; GPRIDX-NEXT: s_waitcnt vmcnt(0)
; GPRIDX-NEXT: s_endpgm ; GPRIDX-NEXT: s_endpgm
; ;
; MOVREL-LABEL: dyn_insertelement_v8f64_v_v_v_add_1: ; MOVREL-LABEL: dyn_insertelement_v8f64_v_v_v_add_1:
; MOVREL: ; %bb.0: ; %entry ; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: v_add_nc_u32_e32 v18, 1, v18 ; MOVREL-NEXT: v_add_nc_u32_e32 v18, 1, v18
; MOVREL-NEXT: v_mov_b32_e32 v19, v0
; MOVREL-NEXT: v_mov_b32_e32 v23, v1
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, 1, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, 1, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s1, 2, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s1, 2, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s2, 3, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s2, 3, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s3, 4, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s3, 4, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s4, 5, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s4, 5, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s5, 7, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s5, 7, v18
; MOVREL-NEXT: v_cmp_eq_u32_e64 s6, 6, v18 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s6, 6, v18
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v19, v16, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v16, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e64 v2, v2, v16, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v2, v2, v16, s0
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v23, v17, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v17, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e64 v3, v3, v17, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v3, v3, v17, s0
; MOVREL-NEXT: v_cndmask_b32_e64 v4, v4, v16, s1 ; MOVREL-NEXT: v_cndmask_b32_e64 v4, v4, v16, s1
; MOVREL-NEXT: v_cndmask_b32_e64 v6, v6, v16, s2 ; MOVREL-NEXT: v_cndmask_b32_e64 v6, v6, v16, s2
; MOVREL-NEXT: v_cndmask_b32_e64 v5, v5, v17, s1 ; MOVREL-NEXT: v_cndmask_b32_e64 v5, v5, v17, s1
; MOVREL-NEXT: v_cndmask_b32_e64 v7, v7, v17, s2 ; MOVREL-NEXT: v_cndmask_b32_e64 v7, v7, v17, s2
; MOVREL-NEXT: v_cndmask_b32_e64 v8, v8, v16, s3 ; MOVREL-NEXT: v_cndmask_b32_e64 v8, v8, v16, s3
; MOVREL-NEXT: v_cndmask_b32_e64 v10, v10, v16, s4 ; MOVREL-NEXT: v_cndmask_b32_e64 v10, v10, v16, s4
; MOVREL-NEXT: v_cndmask_b32_e64 v9, v9, v17, s3 ; MOVREL-NEXT: v_cndmask_b32_e64 v9, v9, v17, s3
▲ Show 20 Lines • Show All 1,360 Lines • ▼ Show 20 Lines
; MOVREL: ; %bb.0: ; %entry ; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 s0, s2 ; MOVREL-NEXT: s_mov_b32 s0, s2
; MOVREL-NEXT: s_mov_b32 s1, s3 ; MOVREL-NEXT: s_mov_b32 s1, s3
; MOVREL-NEXT: s_mov_b32 s2, s4 ; MOVREL-NEXT: s_mov_b32 s2, s4
; MOVREL-NEXT: s_mov_b32 s3, s5 ; MOVREL-NEXT: s_mov_b32 s3, s5
; MOVREL-NEXT: s_mov_b32 s4, s6 ; MOVREL-NEXT: s_mov_b32 s4, s6
; MOVREL-NEXT: s_mov_b32 s5, s7 ; MOVREL-NEXT: s_mov_b32 s5, s7
; MOVREL-NEXT: s_mov_b32 s6, s8 ; MOVREL-NEXT: s_mov_b32 s6, s8
; MOVREL-NEXT: v_mov_b32_e32 v16, s7 ; MOVREL-NEXT: v_mov_b32_e32 v14, s7
; MOVREL-NEXT: v_mov_b32_e32 v9, s0 ; MOVREL-NEXT: v_mov_b32_e32 v7, s0
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 0 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 0
; MOVREL-NEXT: v_mov_b32_e32 v10, s1 ; MOVREL-NEXT: v_mov_b32_e32 v8, s1
; MOVREL-NEXT: v_mov_b32_e32 v11, s2 ; MOVREL-NEXT: v_mov_b32_e32 v9, s2
; MOVREL-NEXT: v_mov_b32_e32 v12, s3 ; MOVREL-NEXT: v_mov_b32_e32 v10, s3
; MOVREL-NEXT: v_mov_b32_e32 v13, s4 ; MOVREL-NEXT: v_mov_b32_e32 v11, s4
; MOVREL-NEXT: v_cndmask_b32_e32 v7, v9, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v7, v7, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 1 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 1
; MOVREL-NEXT: v_mov_b32_e32 v14, s5 ; MOVREL-NEXT: v_mov_b32_e32 v12, s5
; MOVREL-NEXT: v_mov_b32_e32 v15, s6 ; MOVREL-NEXT: v_mov_b32_e32 v13, s6
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v10, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v1, v8, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 2 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 2
; MOVREL-NEXT: v_cndmask_b32_e32 v2, v11, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v2, v9, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 3 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 3
; MOVREL-NEXT: v_cndmask_b32_e32 v3, v12, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v3, v10, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 4 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 4
; MOVREL-NEXT: v_cndmask_b32_e32 v4, v13, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v4, v11, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 5 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 5
; MOVREL-NEXT: v_cndmask_b32_e32 v5, v14, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v5, v12, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 6 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s9, 6
; MOVREL-NEXT: v_cndmask_b32_e32 v6, v15, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v6, v13, v0, vcc_lo
; MOVREL-NEXT: v_mov_b32_e32 v0, v7 ; MOVREL-NEXT: v_mov_b32_e32 v0, v7
; MOVREL-NEXT: ; return to shader part epilog ; MOVREL-NEXT: ; return to shader part epilog
entry: entry:
%insert = insertelement <7 x float> %vec, float %val, i32 %idx %insert = insertelement <7 x float> %vec, float %val, i32 %idx
ret <7 x float> %insert ret <7 x float> %insert
} }
define amdgpu_ps <7 x float> @dyn_insertelement_v7f32_s_v_v(<7 x float> inreg %vec, float %val, i32 %idx) { define amdgpu_ps <7 x float> @dyn_insertelement_v7f32_s_v_v(<7 x float> inreg %vec, float %val, i32 %idx) {
Show All 36 Lines
; MOVREL: ; %bb.0: ; %entry ; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 s0, s2 ; MOVREL-NEXT: s_mov_b32 s0, s2
; MOVREL-NEXT: s_mov_b32 s1, s3 ; MOVREL-NEXT: s_mov_b32 s1, s3
; MOVREL-NEXT: s_mov_b32 s2, s4 ; MOVREL-NEXT: s_mov_b32 s2, s4
; MOVREL-NEXT: s_mov_b32 s3, s5 ; MOVREL-NEXT: s_mov_b32 s3, s5
; MOVREL-NEXT: s_mov_b32 s4, s6 ; MOVREL-NEXT: s_mov_b32 s4, s6
; MOVREL-NEXT: s_mov_b32 s5, s7 ; MOVREL-NEXT: s_mov_b32 s5, s7
; MOVREL-NEXT: s_mov_b32 s6, s8 ; MOVREL-NEXT: s_mov_b32 s6, s8
; MOVREL-NEXT: v_mov_b32_e32 v16, s7 ; MOVREL-NEXT: v_mov_b32_e32 v15, s7
; MOVREL-NEXT: v_mov_b32_e32 v9, s0 ; MOVREL-NEXT: v_mov_b32_e32 v8, s0
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; MOVREL-NEXT: v_mov_b32_e32 v10, s1 ; MOVREL-NEXT: v_mov_b32_e32 v9, s1
; MOVREL-NEXT: v_mov_b32_e32 v11, s2 ; MOVREL-NEXT: v_mov_b32_e32 v10, s2
; MOVREL-NEXT: v_mov_b32_e32 v12, s3 ; MOVREL-NEXT: v_mov_b32_e32 v11, s3
; MOVREL-NEXT: v_mov_b32_e32 v13, s4 ; MOVREL-NEXT: v_mov_b32_e32 v12, s4
; MOVREL-NEXT: v_cndmask_b32_e32 v8, v9, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v8, v8, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
; MOVREL-NEXT: v_mov_b32_e32 v14, s5 ; MOVREL-NEXT: v_mov_b32_e32 v13, s5
; MOVREL-NEXT: v_mov_b32_e32 v15, s6 ; MOVREL-NEXT: v_mov_b32_e32 v14, s6
; MOVREL-NEXT: v_cndmask_b32_e32 v7, v10, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v7, v9, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v1 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v1
; MOVREL-NEXT: v_cndmask_b32_e32 v2, v11, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v2, v10, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v1 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v1
; MOVREL-NEXT: v_cndmask_b32_e32 v3, v12, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v3, v11, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v1 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v1
; MOVREL-NEXT: v_cndmask_b32_e32 v4, v13, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v4, v12, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v1 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v1
; MOVREL-NEXT: v_cndmask_b32_e32 v5, v14, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v5, v13, v0, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v1 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v1
; MOVREL-NEXT: v_mov_b32_e32 v1, v7 ; MOVREL-NEXT: v_mov_b32_e32 v1, v7
; MOVREL-NEXT: v_cndmask_b32_e32 v6, v15, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v6, v14, v0, vcc_lo
; MOVREL-NEXT: v_mov_b32_e32 v0, v8 ; MOVREL-NEXT: v_mov_b32_e32 v0, v8
; MOVREL-NEXT: ; return to shader part epilog ; MOVREL-NEXT: ; return to shader part epilog
entry: entry:
%insert = insertelement <7 x float> %vec, float %val, i32 %idx %insert = insertelement <7 x float> %vec, float %val, i32 %idx
ret <7 x float> %insert ret <7 x float> %insert
} }
define amdgpu_ps <7 x float> @dyn_insertelement_v7f32_v_v_s(<7 x float> %vec, float %val, i32 inreg %idx) { define amdgpu_ps <7 x float> @dyn_insertelement_v7f32_v_v_s(<7 x float> %vec, float %val, i32 inreg %idx) {
▲ Show 20 Lines • Show All 465 Lines • ▼ Show 20 Lines
; GPRIDX-NEXT: v_readfirstlane_b32 s11, v11 ; GPRIDX-NEXT: v_readfirstlane_b32 s11, v11
; GPRIDX-NEXT: v_readfirstlane_b32 s12, v12 ; GPRIDX-NEXT: v_readfirstlane_b32 s12, v12
; GPRIDX-NEXT: v_readfirstlane_b32 s13, v13 ; GPRIDX-NEXT: v_readfirstlane_b32 s13, v13
; GPRIDX-NEXT: ; return to shader part epilog ; GPRIDX-NEXT: ; return to shader part epilog
; ;
; MOVREL-LABEL: dyn_insertelement_v7f64_v_v_v: ; MOVREL-LABEL: dyn_insertelement_v7f64_v_v_v:
; MOVREL: ; %bb.0: ; %entry ; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v16 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v16
; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, 1, v16
; MOVREL-NEXT: v_cmp_eq_u32_e64 s1, 2, v16 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s1, 2, v16
; MOVREL-NEXT: v_cmp_eq_u32_e64 s2, 3, v16 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s2, 3, v16
; MOVREL-NEXT: v_cmp_eq_u32_e64 s3, 4, v16 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s3, 4, v16
; MOVREL-NEXT: v_cmp_eq_u32_e64 s4, 5, v16 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s4, 5, v16
; MOVREL-NEXT: v_cmp_eq_u32_e64 s5, 6, v16 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s5, 6, v16
; MOVREL-NEXT: v_mov_b32_e32 v19, v2 ; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, 1, v16
; MOVREL-NEXT: v_mov_b32_e32 v18, v3
; MOVREL-NEXT: v_cndmask_b32_e64 v6, v6, v14, s2 ; MOVREL-NEXT: v_cndmask_b32_e64 v6, v6, v14, s2
; MOVREL-NEXT: v_cndmask_b32_e64 v8, v8, v14, s3 ; MOVREL-NEXT: v_cndmask_b32_e64 v8, v8, v14, s3
; MOVREL-NEXT: v_cndmask_b32_e64 v10, v10, v14, s4 ; MOVREL-NEXT: v_cndmask_b32_e64 v10, v10, v14, s4
; MOVREL-NEXT: v_cndmask_b32_e64 v12, v12, v14, s5 ; MOVREL-NEXT: v_cndmask_b32_e64 v12, v12, v14, s5
; MOVREL-NEXT: v_cndmask_b32_e64 v7, v7, v15, s2 ; MOVREL-NEXT: v_cndmask_b32_e64 v7, v7, v15, s2
; MOVREL-NEXT: v_cndmask_b32_e64 v2, v19, v14, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v2, v2, v14, s0
; MOVREL-NEXT: v_cndmask_b32_e64 v3, v3, v15, s0
; MOVREL-NEXT: v_cndmask_b32_e64 v9, v9, v15, s3 ; MOVREL-NEXT: v_cndmask_b32_e64 v9, v9, v15, s3
; MOVREL-NEXT: v_cndmask_b32_e64 v3, v18, v15, s0
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e64 v11, v11, v15, s4 ; MOVREL-NEXT: v_cndmask_b32_e64 v11, v11, v15, s4
; MOVREL-NEXT: v_cndmask_b32_e64 v4, v4, v14, s1 ; MOVREL-NEXT: v_cndmask_b32_e64 v4, v4, v14, s1
; MOVREL-NEXT: v_cndmask_b32_e64 v13, v13, v15, s5 ; MOVREL-NEXT: v_cndmask_b32_e64 v13, v13, v15, s5
; MOVREL-NEXT: v_cndmask_b32_e64 v5, v5, v15, s1 ; MOVREL-NEXT: v_cndmask_b32_e64 v5, v5, v15, s1
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc_lo
; MOVREL-NEXT: v_readfirstlane_b32 s0, v0 ; MOVREL-NEXT: v_readfirstlane_b32 s0, v0
; MOVREL-NEXT: v_readfirstlane_b32 s2, v2 ; MOVREL-NEXT: v_readfirstlane_b32 s2, v2
; MOVREL-NEXT: v_readfirstlane_b32 s3, v3 ; MOVREL-NEXT: v_readfirstlane_b32 s3, v3
▲ Show 20 Lines • Show All 110 Lines • ▼ Show 20 Lines
; MOVREL-NEXT: s_mov_b32 s2, s4 ; MOVREL-NEXT: s_mov_b32 s2, s4
; MOVREL-NEXT: s_mov_b32 s3, s5 ; MOVREL-NEXT: s_mov_b32 s3, s5
; MOVREL-NEXT: s_mov_b32 s4, s6 ; MOVREL-NEXT: s_mov_b32 s4, s6
; MOVREL-NEXT: s_mov_b32 s5, s7 ; MOVREL-NEXT: s_mov_b32 s5, s7
; MOVREL-NEXT: s_mov_b32 s6, s8 ; MOVREL-NEXT: s_mov_b32 s6, s8
; MOVREL-NEXT: s_mov_b32 s7, s9 ; MOVREL-NEXT: s_mov_b32 s7, s9
; MOVREL-NEXT: s_mov_b32 s8, s10 ; MOVREL-NEXT: s_mov_b32 s8, s10
; MOVREL-NEXT: s_mov_b32 s9, s11 ; MOVREL-NEXT: s_mov_b32 s9, s11
; MOVREL-NEXT: v_mov_b32_e32 v20, s15 ; MOVREL-NEXT: v_mov_b32_e32 v17, s15
; MOVREL-NEXT: v_mov_b32_e32 v19, s14 ; MOVREL-NEXT: v_mov_b32_e32 v16, s14
; MOVREL-NEXT: v_mov_b32_e32 v18, s13 ; MOVREL-NEXT: v_mov_b32_e32 v15, s13
; MOVREL-NEXT: v_mov_b32_e32 v17, s12 ; MOVREL-NEXT: v_mov_b32_e32 v14, s12
; MOVREL-NEXT: v_mov_b32_e32 v16, s11 ; MOVREL-NEXT: v_mov_b32_e32 v13, s11
; MOVREL-NEXT: v_mov_b32_e32 v15, s10 ; MOVREL-NEXT: v_mov_b32_e32 v12, s10
; MOVREL-NEXT: v_mov_b32_e32 v14, s9 ; MOVREL-NEXT: v_mov_b32_e32 v11, s9
; MOVREL-NEXT: v_mov_b32_e32 v13, s8 ; MOVREL-NEXT: v_mov_b32_e32 v10, s8
; MOVREL-NEXT: v_mov_b32_e32 v12, s7 ; MOVREL-NEXT: v_mov_b32_e32 v9, s7
; MOVREL-NEXT: v_mov_b32_e32 v11, s6 ; MOVREL-NEXT: v_mov_b32_e32 v8, s6
; MOVREL-NEXT: v_mov_b32_e32 v10, s5 ; MOVREL-NEXT: v_mov_b32_e32 v7, s5
; MOVREL-NEXT: v_mov_b32_e32 v9, s4 ; MOVREL-NEXT: v_mov_b32_e32 v6, s4
; MOVREL-NEXT: v_mov_b32_e32 v8, s3 ; MOVREL-NEXT: v_mov_b32_e32 v5, s3
; MOVREL-NEXT: v_mov_b32_e32 v7, s2 ; MOVREL-NEXT: v_mov_b32_e32 v4, s2
; MOVREL-NEXT: v_mov_b32_e32 v6, s1 ; MOVREL-NEXT: v_mov_b32_e32 v3, s1
; MOVREL-NEXT: v_mov_b32_e32 v5, s0 ; MOVREL-NEXT: v_mov_b32_e32 v2, s0
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s12, 0 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s12, 0
; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, s12, 1 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, s12, 1
; MOVREL-NEXT: v_cmp_eq_u32_e64 s1, s12, 4 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s1, s12, 4
; MOVREL-NEXT: v_cndmask_b32_e32 v2, v5, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e32 v3, v6, v1, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v3, v3, v1, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e64 v4, v7, v0, s0
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s12, 2 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s12, 2
; MOVREL-NEXT: v_cndmask_b32_e64 v5, v8, v1, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v4, v4, v0, s0
; MOVREL-NEXT: v_cndmask_b32_e64 v5, v5, v1, s0
; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, s12, 3 ; MOVREL-NEXT: v_cmp_eq_u32_e64 s0, s12, 3
; MOVREL-NEXT: v_cndmask_b32_e32 v6, v6, v0, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e32 v7, v7, v1, vcc_lo
; MOVREL-NEXT: v_readfirstlane_b32 s2, v4 ; MOVREL-NEXT: v_readfirstlane_b32 s2, v4
; MOVREL-NEXT: v_cndmask_b32_e32 v6, v9, v0, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e64 v8, v8, v0, s0
; MOVREL-NEXT: v_cndmask_b32_e32 v7, v10, v1, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e64 v9, v9, v1, s0
; MOVREL-NEXT: v_cndmask_b32_e64 v8, v11, v0, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v0, v10, v0, s1
; MOVREL-NEXT: v_cndmask_b32_e64 v9, v12, v1, s0 ; MOVREL-NEXT: v_cndmask_b32_e64 v1, v11, v1, s1
; MOVREL-NEXT: v_cndmask_b32_e64 v0, v13, v0, s1
; MOVREL-NEXT: v_cndmask_b32_e64 v1, v14, v1, s1
; MOVREL-NEXT: v_readfirstlane_b32 s0, v2 ; MOVREL-NEXT: v_readfirstlane_b32 s0, v2
; MOVREL-NEXT: v_readfirstlane_b32 s1, v3 ; MOVREL-NEXT: v_readfirstlane_b32 s1, v3
; MOVREL-NEXT: v_readfirstlane_b32 s3, v5 ; MOVREL-NEXT: v_readfirstlane_b32 s3, v5
; MOVREL-NEXT: v_readfirstlane_b32 s4, v6 ; MOVREL-NEXT: v_readfirstlane_b32 s4, v6
; MOVREL-NEXT: v_readfirstlane_b32 s5, v7 ; MOVREL-NEXT: v_readfirstlane_b32 s5, v7
; MOVREL-NEXT: v_readfirstlane_b32 s6, v8 ; MOVREL-NEXT: v_readfirstlane_b32 s6, v8
; MOVREL-NEXT: v_readfirstlane_b32 s7, v9 ; MOVREL-NEXT: v_readfirstlane_b32 s7, v9
; MOVREL-NEXT: v_readfirstlane_b32 s8, v0 ; MOVREL-NEXT: v_readfirstlane_b32 s8, v0
▲ Show 20 Lines • Show All 147 Lines • ▼ Show 20 Lines
; GPRIDX-NEXT: v_readfirstlane_b32 s7, v7 ; GPRIDX-NEXT: v_readfirstlane_b32 s7, v7
; GPRIDX-NEXT: v_readfirstlane_b32 s8, v8 ; GPRIDX-NEXT: v_readfirstlane_b32 s8, v8
; GPRIDX-NEXT: v_readfirstlane_b32 s9, v9 ; GPRIDX-NEXT: v_readfirstlane_b32 s9, v9
; GPRIDX-NEXT: ; return to shader part epilog ; GPRIDX-NEXT: ; return to shader part epilog
; ;
; MOVREL-LABEL: dyn_insertelement_v5f64_v_v_s: ; MOVREL-LABEL: dyn_insertelement_v5f64_v_v_s:
; MOVREL: ; %bb.0: ; %entry ; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 0 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 0
; MOVREL-NEXT: v_mov_b32_e32 v15, v2
; MOVREL-NEXT: v_mov_b32_e32 v14, v3
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1
; MOVREL-NEXT: v_readfirstlane_b32 s0, v0 ; MOVREL-NEXT: v_readfirstlane_b32 s0, v0
; MOVREL-NEXT: v_readfirstlane_b32 s1, v1 ; MOVREL-NEXT: v_readfirstlane_b32 s1, v1
; MOVREL-NEXT: v_cndmask_b32_e32 v3, v14, v11, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v3, v3, v11, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e32 v2, v15, v10, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 2 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 2
; MOVREL-NEXT: v_readfirstlane_b32 s3, v3 ; MOVREL-NEXT: v_readfirstlane_b32 s3, v3
; MOVREL-NEXT: v_cndmask_b32_e32 v4, v4, v10, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v4, v4, v10, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e32 v5, v5, v11, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v5, v5, v11, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 3 ; MOVREL-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 3
; MOVREL-NEXT: v_readfirstlane_b32 s4, v4 ; MOVREL-NEXT: v_readfirstlane_b32 s4, v4
; MOVREL-NEXT: v_readfirstlane_b32 s5, v5 ; MOVREL-NEXT: v_readfirstlane_b32 s5, v5
; MOVREL-NEXT: v_cndmask_b32_e32 v6, v6, v10, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v6, v6, v10, vcc_lo
Show All 40 Lines
; GPRIDX-NEXT: v_readfirstlane_b32 s7, v7 ; GPRIDX-NEXT: v_readfirstlane_b32 s7, v7
; GPRIDX-NEXT: v_readfirstlane_b32 s8, v8 ; GPRIDX-NEXT: v_readfirstlane_b32 s8, v8
; GPRIDX-NEXT: v_readfirstlane_b32 s9, v9 ; GPRIDX-NEXT: v_readfirstlane_b32 s9, v9
; GPRIDX-NEXT: ; return to shader part epilog ; GPRIDX-NEXT: ; return to shader part epilog
; ;
; MOVREL-LABEL: dyn_insertelement_v5f64_v_v_v: ; MOVREL-LABEL: dyn_insertelement_v5f64_v_v_v:
; MOVREL: ; %bb.0: ; %entry ; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v12 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v12
; MOVREL-NEXT: v_mov_b32_e32 v15, v2
; MOVREL-NEXT: v_mov_b32_e32 v14, v3
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v12 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v12
; MOVREL-NEXT: v_readfirstlane_b32 s0, v0 ; MOVREL-NEXT: v_readfirstlane_b32 s0, v0
; MOVREL-NEXT: v_readfirstlane_b32 s1, v1 ; MOVREL-NEXT: v_readfirstlane_b32 s1, v1
; MOVREL-NEXT: v_cndmask_b32_e32 v2, v15, v10, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e32 v3, v14, v11, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v3, v3, v11, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v12 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v12
; MOVREL-NEXT: v_readfirstlane_b32 s2, v2 ; MOVREL-NEXT: v_readfirstlane_b32 s2, v2
; MOVREL-NEXT: v_readfirstlane_b32 s3, v3 ; MOVREL-NEXT: v_readfirstlane_b32 s3, v3
; MOVREL-NEXT: v_cndmask_b32_e32 v4, v4, v10, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v4, v4, v10, vcc_lo
; MOVREL-NEXT: v_cndmask_b32_e32 v5, v5, v11, vcc_lo ; MOVREL-NEXT: v_cndmask_b32_e32 v5, v5, v11, vcc_lo
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v12 ; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v12
; MOVREL-NEXT: v_readfirstlane_b32 s4, v4 ; MOVREL-NEXT: v_readfirstlane_b32 s4, v4
; MOVREL-NEXT: v_readfirstlane_b32 s5, v5 ; MOVREL-NEXT: v_readfirstlane_b32 s5, v5
Show All 14 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.atomic.inc.ll

Show First 20 Lines • Show All 1,822 Lines • ▼ Show 20 Lines
; ;
; GFX10-LABEL: flat_atomic_inc_ret_i64_offset_addr64: ; GFX10-LABEL: flat_atomic_inc_ret_i64_offset_addr64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0 ; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 3, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 3, v0
; GFX10-NEXT: v_mov_b32_e32 v2, 42 ; GFX10-NEXT: v_mov_b32_e32 v2, 42
; GFX10-NEXT: v_mov_b32_e32 v3, 0 ; GFX10-NEXT: v_mov_b32_e32 v3, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v8, s3 ; GFX10-NEXT: v_mov_b32_e32 v0, s2
; GFX10-NEXT: v_mov_b32_e32 v7, s2 ; GFX10-NEXT: v_mov_b32_e32 v1, s3
; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v7, v4 ; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, 0, v8, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, 40 ; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, 40
; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT: flat_atomic_inc_x2 v[0:1], v[0:1], v[2:3] glc ; GFX10-NEXT: flat_atomic_inc_x2 v[0:1], v[0:1], v[2:3] glc
; GFX10-NEXT: v_mov_b32_e32 v3, s1 ; GFX10-NEXT: v_mov_b32_e32 v3, s1
; GFX10-NEXT: v_mov_b32_e32 v2, s0 ; GFX10-NEXT: v_mov_b32_e32 v2, s0
; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, v4 ; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
▲ Show 20 Lines • Show All 162 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.gather4.a16.dim.ll

Show All 23 Lines
; GFX9-NEXT: v_and_or_b32 v0, v0, v2, v1 ; GFX9-NEXT: v_and_or_b32 v0, v0, v2, v1
; GFX9-NEXT: s_and_b64 exec, exec, s[14:15] ; GFX9-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX9-NEXT: image_gather4 v[0:3], v0, s[0:7], s[8:11] dmask:0x1 a16 ; GFX9-NEXT: image_gather4 v[0:3], v0, s[0:7], s[8:11] dmask:0x1 a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10NSA-LABEL: gather4_2d: ; GFX10NSA-LABEL: gather4_2d:
; GFX10NSA: ; %bb.0: ; %main_body ; GFX10NSA: ; %bb.0: ; %main_body
; GFX10NSA-NEXT: s_mov_b32 s28, exec_lo ; GFX10NSA-NEXT: s_mov_b32 s14, exec_lo
; GFX10NSA-NEXT: s_mov_b32 s0, s2 ; GFX10NSA-NEXT: s_mov_b32 s0, s2
; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10NSA-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10NSA-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10NSA-NEXT: s_mov_b32 s1, s3 ; GFX10NSA-NEXT: s_mov_b32 s1, s3
; GFX10NSA-NEXT: s_mov_b32 s2, s4 ; GFX10NSA-NEXT: s_mov_b32 s2, s4
; GFX10NSA-NEXT: s_mov_b32 s3, s5 ; GFX10NSA-NEXT: s_mov_b32 s3, s5
; GFX10NSA-NEXT: s_mov_b32 s4, s6 ; GFX10NSA-NEXT: s_mov_b32 s4, s6
; GFX10NSA-NEXT: s_mov_b32 s5, s7 ; GFX10NSA-NEXT: s_mov_b32 s5, s7
; GFX10NSA-NEXT: s_mov_b32 s6, s8 ; GFX10NSA-NEXT: s_mov_b32 s6, s8
; GFX10NSA-NEXT: s_mov_b32 s7, s9 ; GFX10NSA-NEXT: s_mov_b32 s7, s9
; GFX10NSA-NEXT: s_mov_b32 s8, s10 ; GFX10NSA-NEXT: s_mov_b32 s8, s10
; GFX10NSA-NEXT: s_mov_b32 s9, s11 ; GFX10NSA-NEXT: s_mov_b32 s9, s11
; GFX10NSA-NEXT: s_mov_b32 s10, s12 ; GFX10NSA-NEXT: s_mov_b32 s10, s12
; GFX10NSA-NEXT: s_mov_b32 s11, s13 ; GFX10NSA-NEXT: s_mov_b32 s11, s13
; GFX10NSA-NEXT: v_and_or_b32 v0, v0, 0xffff, v1 ; GFX10NSA-NEXT: v_and_or_b32 v0, v0, 0xffff, v1
; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10NSA-NEXT: image_gather4 v[0:3], v0, s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 ; GFX10NSA-NEXT: image_gather4 v[0:3], v0, s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
; GFX10NSA-NEXT: s_waitcnt vmcnt(0) ; GFX10NSA-NEXT: s_waitcnt vmcnt(0)
; GFX10NSA-NEXT: ; return to shader part epilog ; GFX10NSA-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.gather4.2d.v4f32.f16(i32 1, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.gather4.2d.v4f32.f16(i32 1, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
Show All 21 Lines
; GFX9-NEXT: v_and_or_b32 v1, v2, v3, s12 ; GFX9-NEXT: v_and_or_b32 v1, v2, v3, s12
; GFX9-NEXT: s_and_b64 exec, exec, s[14:15] ; GFX9-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX9-NEXT: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16 da ; GFX9-NEXT: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16 da
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10NSA-LABEL: gather4_cube: ; GFX10NSA-LABEL: gather4_cube:
; GFX10NSA: ; %bb.0: ; %main_body ; GFX10NSA: ; %bb.0: ; %main_body
; GFX10NSA-NEXT: s_mov_b32 s28, exec_lo ; GFX10NSA-NEXT: s_mov_b32 s14, exec_lo
; GFX10NSA-NEXT: s_mov_b32 s0, s2 ; GFX10NSA-NEXT: s_mov_b32 s0, s2
; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10NSA-NEXT: v_mov_b32_e32 v3, 0xffff ; GFX10NSA-NEXT: v_mov_b32_e32 v3, 0xffff
; GFX10NSA-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10NSA-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10NSA-NEXT: s_mov_b32 s2, s4 ; GFX10NSA-NEXT: s_mov_b32 s2, s4
; GFX10NSA-NEXT: s_mov_b32 s4, s6 ; GFX10NSA-NEXT: s_mov_b32 s4, s6
; GFX10NSA-NEXT: s_mov_b32 s6, s8 ; GFX10NSA-NEXT: s_mov_b32 s6, s8
; GFX10NSA-NEXT: s_mov_b32 s8, s10 ; GFX10NSA-NEXT: s_mov_b32 s8, s10
; GFX10NSA-NEXT: s_mov_b32 s10, s12 ; GFX10NSA-NEXT: s_mov_b32 s10, s12
; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16 ; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16
; GFX10NSA-NEXT: s_mov_b32 s1, s3 ; GFX10NSA-NEXT: s_mov_b32 s1, s3
; GFX10NSA-NEXT: s_mov_b32 s3, s5 ; GFX10NSA-NEXT: s_mov_b32 s3, s5
; GFX10NSA-NEXT: s_mov_b32 s5, s7 ; GFX10NSA-NEXT: s_mov_b32 s5, s7
; GFX10NSA-NEXT: s_mov_b32 s7, s9 ; GFX10NSA-NEXT: s_mov_b32 s7, s9
; GFX10NSA-NEXT: s_mov_b32 s9, s11 ; GFX10NSA-NEXT: s_mov_b32 s9, s11
; GFX10NSA-NEXT: v_and_or_b32 v0, v0, v3, v1 ; GFX10NSA-NEXT: v_and_or_b32 v0, v0, v3, v1
; GFX10NSA-NEXT: s_mov_b32 s11, s13 ; GFX10NSA-NEXT: s_mov_b32 s11, s13
; GFX10NSA-NEXT: v_and_or_b32 v1, v2, v3, s12 ; GFX10NSA-NEXT: v_and_or_b32 v1, v2, v3, s12
; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10NSA-NEXT: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_CUBE a16 ; GFX10NSA-NEXT: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_CUBE a16
; GFX10NSA-NEXT: s_waitcnt vmcnt(0) ; GFX10NSA-NEXT: s_waitcnt vmcnt(0)
; GFX10NSA-NEXT: ; return to shader part epilog ; GFX10NSA-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.gather4.cube.v4f32.f16(i32 1, half %s, half %t, half %face, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.gather4.cube.v4f32.f16(i32 1, half %s, half %t, half %face, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
Show All 21 Lines
; GFX9-NEXT: v_and_or_b32 v1, v2, v3, s12 ; GFX9-NEXT: v_and_or_b32 v1, v2, v3, s12
; GFX9-NEXT: s_and_b64 exec, exec, s[14:15] ; GFX9-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX9-NEXT: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16 da ; GFX9-NEXT: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16 da
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10NSA-LABEL: gather4_2darray: ; GFX10NSA-LABEL: gather4_2darray:
; GFX10NSA: ; %bb.0: ; %main_body ; GFX10NSA: ; %bb.0: ; %main_body
; GFX10NSA-NEXT: s_mov_b32 s28, exec_lo ; GFX10NSA-NEXT: s_mov_b32 s14, exec_lo
; GFX10NSA-NEXT: s_mov_b32 s0, s2 ; GFX10NSA-NEXT: s_mov_b32 s0, s2
; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10NSA-NEXT: v_mov_b32_e32 v3, 0xffff ; GFX10NSA-NEXT: v_mov_b32_e32 v3, 0xffff
; GFX10NSA-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10NSA-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10NSA-NEXT: s_mov_b32 s2, s4 ; GFX10NSA-NEXT: s_mov_b32 s2, s4
; GFX10NSA-NEXT: s_mov_b32 s4, s6 ; GFX10NSA-NEXT: s_mov_b32 s4, s6
; GFX10NSA-NEXT: s_mov_b32 s6, s8 ; GFX10NSA-NEXT: s_mov_b32 s6, s8
; GFX10NSA-NEXT: s_mov_b32 s8, s10 ; GFX10NSA-NEXT: s_mov_b32 s8, s10
; GFX10NSA-NEXT: s_mov_b32 s10, s12 ; GFX10NSA-NEXT: s_mov_b32 s10, s12
; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16 ; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16
; GFX10NSA-NEXT: s_mov_b32 s1, s3 ; GFX10NSA-NEXT: s_mov_b32 s1, s3
; GFX10NSA-NEXT: s_mov_b32 s3, s5 ; GFX10NSA-NEXT: s_mov_b32 s3, s5
; GFX10NSA-NEXT: s_mov_b32 s5, s7 ; GFX10NSA-NEXT: s_mov_b32 s5, s7
; GFX10NSA-NEXT: s_mov_b32 s7, s9 ; GFX10NSA-NEXT: s_mov_b32 s7, s9
; GFX10NSA-NEXT: s_mov_b32 s9, s11 ; GFX10NSA-NEXT: s_mov_b32 s9, s11
; GFX10NSA-NEXT: v_and_or_b32 v0, v0, v3, v1 ; GFX10NSA-NEXT: v_and_or_b32 v0, v0, v3, v1
; GFX10NSA-NEXT: s_mov_b32 s11, s13 ; GFX10NSA-NEXT: s_mov_b32 s11, s13
; GFX10NSA-NEXT: v_and_or_b32 v1, v2, v3, s12 ; GFX10NSA-NEXT: v_and_or_b32 v1, v2, v3, s12
; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10NSA-NEXT: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D_ARRAY a16 ; GFX10NSA-NEXT: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D_ARRAY a16
; GFX10NSA-NEXT: s_waitcnt vmcnt(0) ; GFX10NSA-NEXT: s_waitcnt vmcnt(0)
; GFX10NSA-NEXT: ; return to shader part epilog ; GFX10NSA-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.gather4.2darray.v4f32.f16(i32 1, half %s, half %t, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.gather4.2darray.v4f32.f16(i32 1, half %s, half %t, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
Show All 19 Lines
; GFX9-NEXT: v_and_or_b32 v1, v1, v3, v2 ; GFX9-NEXT: v_and_or_b32 v1, v1, v3, v2
; GFX9-NEXT: s_and_b64 exec, exec, s[14:15] ; GFX9-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX9-NEXT: image_gather4_c v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16 ; GFX9-NEXT: image_gather4_c v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10NSA-LABEL: gather4_c_2d: ; GFX10NSA-LABEL: gather4_c_2d:
; GFX10NSA: ; %bb.0: ; %main_body ; GFX10NSA: ; %bb.0: ; %main_body
; GFX10NSA-NEXT: s_mov_b32 s28, exec_lo ; GFX10NSA-NEXT: s_mov_b32 s14, exec_lo
; GFX10NSA-NEXT: s_mov_b32 s0, s2 ; GFX10NSA-NEXT: s_mov_b32 s0, s2
; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10NSA-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10NSA-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10NSA-NEXT: s_mov_b32 s1, s3 ; GFX10NSA-NEXT: s_mov_b32 s1, s3
; GFX10NSA-NEXT: s_mov_b32 s2, s4 ; GFX10NSA-NEXT: s_mov_b32 s2, s4
; GFX10NSA-NEXT: s_mov_b32 s3, s5 ; GFX10NSA-NEXT: s_mov_b32 s3, s5
; GFX10NSA-NEXT: s_mov_b32 s4, s6 ; GFX10NSA-NEXT: s_mov_b32 s4, s6
; GFX10NSA-NEXT: s_mov_b32 s5, s7 ; GFX10NSA-NEXT: s_mov_b32 s5, s7
; GFX10NSA-NEXT: s_mov_b32 s6, s8 ; GFX10NSA-NEXT: s_mov_b32 s6, s8
; GFX10NSA-NEXT: s_mov_b32 s7, s9 ; GFX10NSA-NEXT: s_mov_b32 s7, s9
; GFX10NSA-NEXT: s_mov_b32 s8, s10 ; GFX10NSA-NEXT: s_mov_b32 s8, s10
; GFX10NSA-NEXT: s_mov_b32 s9, s11 ; GFX10NSA-NEXT: s_mov_b32 s9, s11
; GFX10NSA-NEXT: s_mov_b32 s10, s12 ; GFX10NSA-NEXT: s_mov_b32 s10, s12
; GFX10NSA-NEXT: s_mov_b32 s11, s13 ; GFX10NSA-NEXT: s_mov_b32 s11, s13
; GFX10NSA-NEXT: v_and_or_b32 v1, v1, 0xffff, v2 ; GFX10NSA-NEXT: v_and_or_b32 v1, v1, 0xffff, v2
; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10NSA-NEXT: image_gather4_c v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 ; GFX10NSA-NEXT: image_gather4_c v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
; GFX10NSA-NEXT: s_waitcnt vmcnt(0) ; GFX10NSA-NEXT: s_waitcnt vmcnt(0)
; GFX10NSA-NEXT: ; return to shader part epilog ; GFX10NSA-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.gather4.c.2d.v4f32.f16(i32 1, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.gather4.c.2d.v4f32.f16(i32 1, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
Show All 21 Lines
; GFX9-NEXT: v_and_or_b32 v1, v2, v3, s12 ; GFX9-NEXT: v_and_or_b32 v1, v2, v3, s12
; GFX9-NEXT: s_and_b64 exec, exec, s[14:15] ; GFX9-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX9-NEXT: image_gather4_cl v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16 ; GFX9-NEXT: image_gather4_cl v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10NSA-LABEL: gather4_cl_2d: ; GFX10NSA-LABEL: gather4_cl_2d:
; GFX10NSA: ; %bb.0: ; %main_body ; GFX10NSA: ; %bb.0: ; %main_body
; GFX10NSA-NEXT: s_mov_b32 s28, exec_lo ; GFX10NSA-NEXT: s_mov_b32 s14, exec_lo
; GFX10NSA-NEXT: s_mov_b32 s0, s2 ; GFX10NSA-NEXT: s_mov_b32 s0, s2
; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10NSA-NEXT: v_mov_b32_e32 v3, 0xffff ; GFX10NSA-NEXT: v_mov_b32_e32 v3, 0xffff
; GFX10NSA-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10NSA-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10NSA-NEXT: s_mov_b32 s2, s4 ; GFX10NSA-NEXT: s_mov_b32 s2, s4
; GFX10NSA-NEXT: s_mov_b32 s4, s6 ; GFX10NSA-NEXT: s_mov_b32 s4, s6
; GFX10NSA-NEXT: s_mov_b32 s6, s8 ; GFX10NSA-NEXT: s_mov_b32 s6, s8
; GFX10NSA-NEXT: s_mov_b32 s8, s10 ; GFX10NSA-NEXT: s_mov_b32 s8, s10
; GFX10NSA-NEXT: s_mov_b32 s10, s12 ; GFX10NSA-NEXT: s_mov_b32 s10, s12
; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16 ; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16
; GFX10NSA-NEXT: s_mov_b32 s1, s3 ; GFX10NSA-NEXT: s_mov_b32 s1, s3
; GFX10NSA-NEXT: s_mov_b32 s3, s5 ; GFX10NSA-NEXT: s_mov_b32 s3, s5
; GFX10NSA-NEXT: s_mov_b32 s5, s7 ; GFX10NSA-NEXT: s_mov_b32 s5, s7
; GFX10NSA-NEXT: s_mov_b32 s7, s9 ; GFX10NSA-NEXT: s_mov_b32 s7, s9
; GFX10NSA-NEXT: s_mov_b32 s9, s11 ; GFX10NSA-NEXT: s_mov_b32 s9, s11
; GFX10NSA-NEXT: v_and_or_b32 v0, v0, v3, v1 ; GFX10NSA-NEXT: v_and_or_b32 v0, v0, v3, v1
; GFX10NSA-NEXT: s_mov_b32 s11, s13 ; GFX10NSA-NEXT: s_mov_b32 s11, s13
; GFX10NSA-NEXT: v_and_or_b32 v1, v2, v3, s12 ; GFX10NSA-NEXT: v_and_or_b32 v1, v2, v3, s12
; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10NSA-NEXT: image_gather4_cl v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 ; GFX10NSA-NEXT: image_gather4_cl v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
; GFX10NSA-NEXT: s_waitcnt vmcnt(0) ; GFX10NSA-NEXT: s_waitcnt vmcnt(0)
; GFX10NSA-NEXT: ; return to shader part epilog ; GFX10NSA-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.gather4.cl.2d.v4f32.f16(i32 1, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.gather4.cl.2d.v4f32.f16(i32 1, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
Show All 21 Lines
; GFX9-NEXT: v_and_or_b32 v2, v3, v4, s12 ; GFX9-NEXT: v_and_or_b32 v2, v3, v4, s12
; GFX9-NEXT: s_and_b64 exec, exec, s[14:15] ; GFX9-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX9-NEXT: image_gather4_c_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 a16 ; GFX9-NEXT: image_gather4_c_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10NSA-LABEL: gather4_c_cl_2d: ; GFX10NSA-LABEL: gather4_c_cl_2d:
; GFX10NSA: ; %bb.0: ; %main_body ; GFX10NSA: ; %bb.0: ; %main_body
; GFX10NSA-NEXT: s_mov_b32 s28, exec_lo ; GFX10NSA-NEXT: s_mov_b32 s14, exec_lo
; GFX10NSA-NEXT: s_mov_b32 s0, s2 ; GFX10NSA-NEXT: s_mov_b32 s0, s2
; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10NSA-NEXT: v_mov_b32_e32 v4, 0xffff ; GFX10NSA-NEXT: v_mov_b32_e32 v4, 0xffff
; GFX10NSA-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10NSA-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10NSA-NEXT: s_mov_b32 s2, s4 ; GFX10NSA-NEXT: s_mov_b32 s2, s4
; GFX10NSA-NEXT: s_mov_b32 s4, s6 ; GFX10NSA-NEXT: s_mov_b32 s4, s6
; GFX10NSA-NEXT: s_mov_b32 s6, s8 ; GFX10NSA-NEXT: s_mov_b32 s6, s8
; GFX10NSA-NEXT: s_mov_b32 s8, s10 ; GFX10NSA-NEXT: s_mov_b32 s8, s10
; GFX10NSA-NEXT: s_mov_b32 s10, s12 ; GFX10NSA-NEXT: s_mov_b32 s10, s12
; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16 ; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16
; GFX10NSA-NEXT: s_mov_b32 s1, s3 ; GFX10NSA-NEXT: s_mov_b32 s1, s3
; GFX10NSA-NEXT: s_mov_b32 s3, s5 ; GFX10NSA-NEXT: s_mov_b32 s3, s5
; GFX10NSA-NEXT: s_mov_b32 s5, s7 ; GFX10NSA-NEXT: s_mov_b32 s5, s7
; GFX10NSA-NEXT: s_mov_b32 s7, s9 ; GFX10NSA-NEXT: s_mov_b32 s7, s9
; GFX10NSA-NEXT: s_mov_b32 s9, s11 ; GFX10NSA-NEXT: s_mov_b32 s9, s11
; GFX10NSA-NEXT: v_and_or_b32 v1, v1, v4, v2 ; GFX10NSA-NEXT: v_and_or_b32 v1, v1, v4, v2
; GFX10NSA-NEXT: s_mov_b32 s11, s13 ; GFX10NSA-NEXT: s_mov_b32 s11, s13
; GFX10NSA-NEXT: v_and_or_b32 v2, v3, v4, s12 ; GFX10NSA-NEXT: v_and_or_b32 v2, v3, v4, s12
; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10NSA-NEXT: image_gather4_c_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 ; GFX10NSA-NEXT: image_gather4_c_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
; GFX10NSA-NEXT: s_waitcnt vmcnt(0) ; GFX10NSA-NEXT: s_waitcnt vmcnt(0)
; GFX10NSA-NEXT: ; return to shader part epilog ; GFX10NSA-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.gather4.c.cl.2d.v4f32.f16(i32 1, float %zcompare, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.gather4.c.cl.2d.v4f32.f16(i32 1, float %zcompare, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
Show All 19 Lines
; GFX9-NEXT: v_and_or_b32 v1, v1, v3, v2 ; GFX9-NEXT: v_and_or_b32 v1, v1, v3, v2
; GFX9-NEXT: s_and_b64 exec, exec, s[14:15] ; GFX9-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX9-NEXT: image_gather4_b v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16 ; GFX9-NEXT: image_gather4_b v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10NSA-LABEL: gather4_b_2d: ; GFX10NSA-LABEL: gather4_b_2d:
; GFX10NSA: ; %bb.0: ; %main_body ; GFX10NSA: ; %bb.0: ; %main_body
; GFX10NSA-NEXT: s_mov_b32 s28, exec_lo ; GFX10NSA-NEXT: s_mov_b32 s14, exec_lo
; GFX10NSA-NEXT: s_mov_b32 s0, s2 ; GFX10NSA-NEXT: s_mov_b32 s0, s2
; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10NSA-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10NSA-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10NSA-NEXT: s_mov_b32 s1, s3 ; GFX10NSA-NEXT: s_mov_b32 s1, s3
; GFX10NSA-NEXT: s_mov_b32 s2, s4 ; GFX10NSA-NEXT: s_mov_b32 s2, s4
; GFX10NSA-NEXT: s_mov_b32 s3, s5 ; GFX10NSA-NEXT: s_mov_b32 s3, s5
; GFX10NSA-NEXT: s_mov_b32 s4, s6 ; GFX10NSA-NEXT: s_mov_b32 s4, s6
; GFX10NSA-NEXT: s_mov_b32 s5, s7 ; GFX10NSA-NEXT: s_mov_b32 s5, s7
; GFX10NSA-NEXT: s_mov_b32 s6, s8 ; GFX10NSA-NEXT: s_mov_b32 s6, s8
; GFX10NSA-NEXT: s_mov_b32 s7, s9 ; GFX10NSA-NEXT: s_mov_b32 s7, s9
; GFX10NSA-NEXT: s_mov_b32 s8, s10 ; GFX10NSA-NEXT: s_mov_b32 s8, s10
; GFX10NSA-NEXT: s_mov_b32 s9, s11 ; GFX10NSA-NEXT: s_mov_b32 s9, s11
; GFX10NSA-NEXT: s_mov_b32 s10, s12 ; GFX10NSA-NEXT: s_mov_b32 s10, s12
; GFX10NSA-NEXT: s_mov_b32 s11, s13 ; GFX10NSA-NEXT: s_mov_b32 s11, s13
; GFX10NSA-NEXT: v_and_or_b32 v1, v1, 0xffff, v2 ; GFX10NSA-NEXT: v_and_or_b32 v1, v1, 0xffff, v2
; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10NSA-NEXT: image_gather4_b v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 ; GFX10NSA-NEXT: image_gather4_b v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
; GFX10NSA-NEXT: s_waitcnt vmcnt(0) ; GFX10NSA-NEXT: s_waitcnt vmcnt(0)
; GFX10NSA-NEXT: ; return to shader part epilog ; GFX10NSA-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.gather4.b.2d.v4f32.f32.f16(i32 1, float %bias, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.gather4.b.2d.v4f32.f32.f16(i32 1, float %bias, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
Show All 19 Lines
; GFX9-NEXT: v_and_or_b32 v2, v2, v4, v3 ; GFX9-NEXT: v_and_or_b32 v2, v2, v4, v3
; GFX9-NEXT: s_and_b64 exec, exec, s[14:15] ; GFX9-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX9-NEXT: image_gather4_c_b v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 a16 ; GFX9-NEXT: image_gather4_c_b v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10NSA-LABEL: gather4_c_b_2d: ; GFX10NSA-LABEL: gather4_c_b_2d:
; GFX10NSA: ; %bb.0: ; %main_body ; GFX10NSA: ; %bb.0: ; %main_body
; GFX10NSA-NEXT: s_mov_b32 s28, exec_lo ; GFX10NSA-NEXT: s_mov_b32 s14, exec_lo
; GFX10NSA-NEXT: s_mov_b32 s0, s2 ; GFX10NSA-NEXT: s_mov_b32 s0, s2
; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10NSA-NEXT: v_lshlrev_b32_e32 v3, 16, v3 ; GFX10NSA-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; GFX10NSA-NEXT: s_mov_b32 s1, s3 ; GFX10NSA-NEXT: s_mov_b32 s1, s3
; GFX10NSA-NEXT: s_mov_b32 s2, s4 ; GFX10NSA-NEXT: s_mov_b32 s2, s4
; GFX10NSA-NEXT: s_mov_b32 s3, s5 ; GFX10NSA-NEXT: s_mov_b32 s3, s5
; GFX10NSA-NEXT: s_mov_b32 s4, s6 ; GFX10NSA-NEXT: s_mov_b32 s4, s6
; GFX10NSA-NEXT: s_mov_b32 s5, s7 ; GFX10NSA-NEXT: s_mov_b32 s5, s7
; GFX10NSA-NEXT: s_mov_b32 s6, s8 ; GFX10NSA-NEXT: s_mov_b32 s6, s8
; GFX10NSA-NEXT: s_mov_b32 s7, s9 ; GFX10NSA-NEXT: s_mov_b32 s7, s9
; GFX10NSA-NEXT: s_mov_b32 s8, s10 ; GFX10NSA-NEXT: s_mov_b32 s8, s10
; GFX10NSA-NEXT: s_mov_b32 s9, s11 ; GFX10NSA-NEXT: s_mov_b32 s9, s11
; GFX10NSA-NEXT: s_mov_b32 s10, s12 ; GFX10NSA-NEXT: s_mov_b32 s10, s12
; GFX10NSA-NEXT: s_mov_b32 s11, s13 ; GFX10NSA-NEXT: s_mov_b32 s11, s13
; GFX10NSA-NEXT: v_and_or_b32 v2, v2, 0xffff, v3 ; GFX10NSA-NEXT: v_and_or_b32 v2, v2, 0xffff, v3
; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10NSA-NEXT: image_gather4_c_b v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 ; GFX10NSA-NEXT: image_gather4_c_b v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
; GFX10NSA-NEXT: s_waitcnt vmcnt(0) ; GFX10NSA-NEXT: s_waitcnt vmcnt(0)
; GFX10NSA-NEXT: ; return to shader part epilog ; GFX10NSA-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.2d.v4f32.f32.f16(i32 1, float %bias, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.2d.v4f32.f32.f16(i32 1, float %bias, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
Show All 21 Lines
; GFX9-NEXT: v_and_or_b32 v2, v3, v4, s12 ; GFX9-NEXT: v_and_or_b32 v2, v3, v4, s12
; GFX9-NEXT: s_and_b64 exec, exec, s[14:15] ; GFX9-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX9-NEXT: image_gather4_b_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 a16 ; GFX9-NEXT: image_gather4_b_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10NSA-LABEL: gather4_b_cl_2d: ; GFX10NSA-LABEL: gather4_b_cl_2d:
; GFX10NSA: ; %bb.0: ; %main_body ; GFX10NSA: ; %bb.0: ; %main_body
; GFX10NSA-NEXT: s_mov_b32 s28, exec_lo ; GFX10NSA-NEXT: s_mov_b32 s14, exec_lo
; GFX10NSA-NEXT: s_mov_b32 s0, s2 ; GFX10NSA-NEXT: s_mov_b32 s0, s2
; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10NSA-NEXT: v_mov_b32_e32 v4, 0xffff ; GFX10NSA-NEXT: v_mov_b32_e32 v4, 0xffff
; GFX10NSA-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10NSA-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10NSA-NEXT: s_mov_b32 s2, s4 ; GFX10NSA-NEXT: s_mov_b32 s2, s4
; GFX10NSA-NEXT: s_mov_b32 s4, s6 ; GFX10NSA-NEXT: s_mov_b32 s4, s6
; GFX10NSA-NEXT: s_mov_b32 s6, s8 ; GFX10NSA-NEXT: s_mov_b32 s6, s8
; GFX10NSA-NEXT: s_mov_b32 s8, s10 ; GFX10NSA-NEXT: s_mov_b32 s8, s10
; GFX10NSA-NEXT: s_mov_b32 s10, s12 ; GFX10NSA-NEXT: s_mov_b32 s10, s12
; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16 ; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16
; GFX10NSA-NEXT: s_mov_b32 s1, s3 ; GFX10NSA-NEXT: s_mov_b32 s1, s3
; GFX10NSA-NEXT: s_mov_b32 s3, s5 ; GFX10NSA-NEXT: s_mov_b32 s3, s5
; GFX10NSA-NEXT: s_mov_b32 s5, s7 ; GFX10NSA-NEXT: s_mov_b32 s5, s7
; GFX10NSA-NEXT: s_mov_b32 s7, s9 ; GFX10NSA-NEXT: s_mov_b32 s7, s9
; GFX10NSA-NEXT: s_mov_b32 s9, s11 ; GFX10NSA-NEXT: s_mov_b32 s9, s11
; GFX10NSA-NEXT: v_and_or_b32 v1, v1, v4, v2 ; GFX10NSA-NEXT: v_and_or_b32 v1, v1, v4, v2
; GFX10NSA-NEXT: s_mov_b32 s11, s13 ; GFX10NSA-NEXT: s_mov_b32 s11, s13
; GFX10NSA-NEXT: v_and_or_b32 v2, v3, v4, s12 ; GFX10NSA-NEXT: v_and_or_b32 v2, v3, v4, s12
; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10NSA-NEXT: image_gather4_b_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 ; GFX10NSA-NEXT: image_gather4_b_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
; GFX10NSA-NEXT: s_waitcnt vmcnt(0) ; GFX10NSA-NEXT: s_waitcnt vmcnt(0)
; GFX10NSA-NEXT: ; return to shader part epilog ; GFX10NSA-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.gather4.b.cl.2d.v4f32.f32.f16(i32 1, float %bias, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.gather4.b.cl.2d.v4f32.f32.f16(i32 1, float %bias, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
Show All 21 Lines
; GFX9-NEXT: v_and_or_b32 v3, v4, v5, s12 ; GFX9-NEXT: v_and_or_b32 v3, v4, v5, s12
; GFX9-NEXT: s_and_b64 exec, exec, s[14:15] ; GFX9-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX9-NEXT: image_gather4_c_b_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 a16 ; GFX9-NEXT: image_gather4_c_b_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10NSA-LABEL: gather4_c_b_cl_2d: ; GFX10NSA-LABEL: gather4_c_b_cl_2d:
; GFX10NSA: ; %bb.0: ; %main_body ; GFX10NSA: ; %bb.0: ; %main_body
; GFX10NSA-NEXT: s_mov_b32 s28, exec_lo ; GFX10NSA-NEXT: s_mov_b32 s14, exec_lo
; GFX10NSA-NEXT: s_mov_b32 s0, s2 ; GFX10NSA-NEXT: s_mov_b32 s0, s2
; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10NSA-NEXT: v_mov_b32_e32 v5, 0xffff ; GFX10NSA-NEXT: v_mov_b32_e32 v5, 0xffff
; GFX10NSA-NEXT: v_lshlrev_b32_e32 v3, 16, v3 ; GFX10NSA-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; GFX10NSA-NEXT: s_mov_b32 s2, s4 ; GFX10NSA-NEXT: s_mov_b32 s2, s4
; GFX10NSA-NEXT: s_mov_b32 s4, s6 ; GFX10NSA-NEXT: s_mov_b32 s4, s6
; GFX10NSA-NEXT: s_mov_b32 s6, s8 ; GFX10NSA-NEXT: s_mov_b32 s6, s8
; GFX10NSA-NEXT: s_mov_b32 s8, s10 ; GFX10NSA-NEXT: s_mov_b32 s8, s10
; GFX10NSA-NEXT: s_mov_b32 s10, s12 ; GFX10NSA-NEXT: s_mov_b32 s10, s12
; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16 ; GFX10NSA-NEXT: s_lshl_b32 s12, s0, 16
; GFX10NSA-NEXT: s_mov_b32 s1, s3 ; GFX10NSA-NEXT: s_mov_b32 s1, s3
; GFX10NSA-NEXT: s_mov_b32 s3, s5 ; GFX10NSA-NEXT: s_mov_b32 s3, s5
; GFX10NSA-NEXT: s_mov_b32 s5, s7 ; GFX10NSA-NEXT: s_mov_b32 s5, s7
; GFX10NSA-NEXT: s_mov_b32 s7, s9 ; GFX10NSA-NEXT: s_mov_b32 s7, s9
; GFX10NSA-NEXT: s_mov_b32 s9, s11 ; GFX10NSA-NEXT: s_mov_b32 s9, s11
; GFX10NSA-NEXT: v_and_or_b32 v2, v2, v5, v3 ; GFX10NSA-NEXT: v_and_or_b32 v2, v2, v5, v3
; GFX10NSA-NEXT: s_mov_b32 s11, s13 ; GFX10NSA-NEXT: s_mov_b32 s11, s13
; GFX10NSA-NEXT: v_and_or_b32 v3, v4, v5, s12 ; GFX10NSA-NEXT: v_and_or_b32 v3, v4, v5, s12
; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10NSA-NEXT: image_gather4_c_b_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 ; GFX10NSA-NEXT: image_gather4_c_b_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
; GFX10NSA-NEXT: s_waitcnt vmcnt(0) ; GFX10NSA-NEXT: s_waitcnt vmcnt(0)
; GFX10NSA-NEXT: ; return to shader part epilog ; GFX10NSA-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.2d.v4f32.f32.f16(i32 1, float %bias, float %zcompare, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.2d.v4f32.f32.f16(i32 1, float %bias, float %zcompare, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
▲ Show 20 Lines • Show All 210 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.gather4.dim.ll

Show First 20 Lines • Show All 74 Lines • ▼ Show 20 Lines
; GFX6-NEXT: v_mov_b32_e32 v4, v0 ; GFX6-NEXT: v_mov_b32_e32 v4, v0
; GFX6-NEXT: s_and_b64 exec, exec, s[14:15] ; GFX6-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX6-NEXT: image_gather4 v[0:4], v[5:6], s[0:7], s[8:11] dmask:0x1 tfe ; GFX6-NEXT: image_gather4 v[0:4], v[5:6], s[0:7], s[8:11] dmask:0x1 tfe
; GFX6-NEXT: s_waitcnt vmcnt(0) ; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog ; GFX6-NEXT: ; return to shader part epilog
; ;
; GFX10NSA-LABEL: gather4_2d_tfe: ; GFX10NSA-LABEL: gather4_2d_tfe:
; GFX10NSA: ; %bb.0: ; %main_body ; GFX10NSA: ; %bb.0: ; %main_body
; GFX10NSA-NEXT: s_mov_b32 s28, exec_lo ; GFX10NSA-NEXT: s_mov_b32 s14, exec_lo
; GFX10NSA-NEXT: s_mov_b32 s0, s2 ; GFX10NSA-NEXT: s_mov_b32 s0, s2
; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10NSA-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10NSA-NEXT: v_mov_b32_e32 v5, v0 ; GFX10NSA-NEXT: v_mov_b32_e32 v5, v0
; GFX10NSA-NEXT: v_mov_b32_e32 v0, 0 ; GFX10NSA-NEXT: v_mov_b32_e32 v0, 0
; GFX10NSA-NEXT: v_mov_b32_e32 v6, v1 ; GFX10NSA-NEXT: v_mov_b32_e32 v6, v1
; GFX10NSA-NEXT: s_mov_b32 s1, s3 ; GFX10NSA-NEXT: s_mov_b32 s1, s3
; GFX10NSA-NEXT: s_mov_b32 s2, s4 ; GFX10NSA-NEXT: s_mov_b32 s2, s4
; GFX10NSA-NEXT: s_mov_b32 s3, s5 ; GFX10NSA-NEXT: s_mov_b32 s3, s5
; GFX10NSA-NEXT: s_mov_b32 s4, s6 ; GFX10NSA-NEXT: s_mov_b32 s4, s6
; GFX10NSA-NEXT: s_mov_b32 s5, s7 ; GFX10NSA-NEXT: s_mov_b32 s5, s7
; GFX10NSA-NEXT: s_mov_b32 s6, s8 ; GFX10NSA-NEXT: s_mov_b32 s6, s8
; GFX10NSA-NEXT: s_mov_b32 s7, s9 ; GFX10NSA-NEXT: s_mov_b32 s7, s9
; GFX10NSA-NEXT: s_mov_b32 s8, s10 ; GFX10NSA-NEXT: s_mov_b32 s8, s10
; GFX10NSA-NEXT: s_mov_b32 s9, s11 ; GFX10NSA-NEXT: s_mov_b32 s9, s11
; GFX10NSA-NEXT: s_mov_b32 s10, s12 ; GFX10NSA-NEXT: s_mov_b32 s10, s12
; GFX10NSA-NEXT: s_mov_b32 s11, s13 ; GFX10NSA-NEXT: s_mov_b32 s11, s13
; GFX10NSA-NEXT: v_mov_b32_e32 v1, v0 ; GFX10NSA-NEXT: v_mov_b32_e32 v1, v0
; GFX10NSA-NEXT: v_mov_b32_e32 v2, v0 ; GFX10NSA-NEXT: v_mov_b32_e32 v2, v0
; GFX10NSA-NEXT: v_mov_b32_e32 v3, v0 ; GFX10NSA-NEXT: v_mov_b32_e32 v3, v0
; GFX10NSA-NEXT: v_mov_b32_e32 v4, v0 ; GFX10NSA-NEXT: v_mov_b32_e32 v4, v0
; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10NSA-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10NSA-NEXT: image_gather4 v[0:4], v[5:6], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D tfe ; GFX10NSA-NEXT: image_gather4 v[0:4], v[5:6], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D tfe
; GFX10NSA-NEXT: s_waitcnt vmcnt(0) ; GFX10NSA-NEXT: s_waitcnt vmcnt(0)
; GFX10NSA-NEXT: ; return to shader part epilog ; GFX10NSA-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call { <4 x float>, i32 } @llvm.amdgcn.image.gather4.2d.sl_v4f32i32s.f32(i32 1, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 1, i32 0) %v = call { <4 x float>, i32 } @llvm.amdgcn.image.gather4.2d.sl_v4f32i32s.f32(i32 1, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 1, i32 0)
%r = extractvalue { <4 x float>, i32 } %v, 0 %r = extractvalue { <4 x float>, i32 } %v, 0
ret <4 x float> %r ret <4 x float> %r
} }
▲ Show 20 Lines • Show All 746 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.load.3d.ll

Show First 20 Lines • Show All 59 Lines • ▼ Show 20 Lines
; GFX6-NEXT: s_mov_b32 s11, 0xf000 ; GFX6-NEXT: s_mov_b32 s11, 0xf000
; GFX6-NEXT: s_waitcnt vmcnt(0) ; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_store_dword v4, off, s[8:11], 0 ; GFX6-NEXT: buffer_store_dword v4, off, s[8:11], 0
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) ; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; GFX6-NEXT: ; return to shader part epilog ; GFX6-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: load_3d_v4f32_xyzw_tfe: ; GFX10-LABEL: load_3d_v4f32_xyzw_tfe:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: v_mov_b32_e32 v11, 0 ; GFX10-NEXT: v_mov_b32_e32 v8, 0
; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; GFX10-NEXT: v_mov_b32_e32 v5, v0
; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; GFX10-NEXT: v_mov_b32_e32 v6, v1
; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; GFX10-NEXT: v_mov_b32_e32 v7, v2
; GFX10-NEXT: s_mov_b32 s0, s2 ; GFX10-NEXT: s_mov_b32 s0, s2
; GFX10-NEXT: v_mov_b32_e32 v12, v11 ; GFX10-NEXT: v_mov_b32_e32 v9, v8
; GFX10-NEXT: v_mov_b32_e32 v13, v11 ; GFX10-NEXT: v_mov_b32_e32 v10, v8
; GFX10-NEXT: v_mov_b32_e32 v14, v11 ; GFX10-NEXT: v_mov_b32_e32 v11, v8
; GFX10-NEXT: v_mov_b32_e32 v15, v11 ; GFX10-NEXT: v_mov_b32_e32 v12, v8
; GFX10-NEXT: v_mov_b32_e32 v0, v11 ; GFX10-NEXT: v_mov_b32_e32 v0, v8
; GFX10-NEXT: s_mov_b32 s1, s3 ; GFX10-NEXT: s_mov_b32 s1, s3
; GFX10-NEXT: s_mov_b32 s2, s4 ; GFX10-NEXT: s_mov_b32 s2, s4
; GFX10-NEXT: s_mov_b32 s3, s5 ; GFX10-NEXT: s_mov_b32 s3, s5
; GFX10-NEXT: s_mov_b32 s4, s6 ; GFX10-NEXT: s_mov_b32 s4, s6
; GFX10-NEXT: s_mov_b32 s5, s7 ; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8 ; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: s_mov_b32 s7, s9 ; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: v_mov_b32_e32 v1, v12 ; GFX10-NEXT: v_mov_b32_e32 v1, v9
; GFX10-NEXT: v_mov_b32_e32 v2, v13 ; GFX10-NEXT: v_mov_b32_e32 v2, v10
; GFX10-NEXT: v_mov_b32_e32 v3, v14 ; GFX10-NEXT: v_mov_b32_e32 v3, v11
; GFX10-NEXT: v_mov_b32_e32 v4, v15 ; GFX10-NEXT: v_mov_b32_e32 v4, v12
; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm tfe ; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm tfe
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: global_store_dword v11, v4, s[10:11] ; GFX10-NEXT: global_store_dword v8, v4, s[10:11]
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.3d.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 1, i32 0) %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.3d.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 1, i32 0)
%v.vec = extractvalue { <4 x float>, i32 } %v, 0 %v.vec = extractvalue { <4 x float>, i32 } %v, 0
%v.err = extractvalue { <4 x float>, i32 } %v, 1 %v.err = extractvalue { <4 x float>, i32 } %v, 1
store i32 %v.err, i32 addrspace(1)* %out, align 4 store i32 %v.err, i32 addrspace(1)* %out, align 4
ret <4 x float> %v.vec ret <4 x float> %v.vec
} }
Show All 24 Lines
; GFX6-NEXT: s_mov_b32 s11, 0xf000 ; GFX6-NEXT: s_mov_b32 s11, 0xf000
; GFX6-NEXT: s_waitcnt vmcnt(0) ; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_store_dword v4, off, s[8:11], 0 ; GFX6-NEXT: buffer_store_dword v4, off, s[8:11], 0
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) ; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; GFX6-NEXT: ; return to shader part epilog ; GFX6-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: load_3d_v4f32_xyzw_tfe_lwe: ; GFX10-LABEL: load_3d_v4f32_xyzw_tfe_lwe:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: v_mov_b32_e32 v11, 0 ; GFX10-NEXT: v_mov_b32_e32 v8, 0
; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; GFX10-NEXT: v_mov_b32_e32 v5, v0
; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; GFX10-NEXT: v_mov_b32_e32 v6, v1
; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; GFX10-NEXT: v_mov_b32_e32 v7, v2
; GFX10-NEXT: s_mov_b32 s0, s2 ; GFX10-NEXT: s_mov_b32 s0, s2
; GFX10-NEXT: v_mov_b32_e32 v12, v11 ; GFX10-NEXT: v_mov_b32_e32 v9, v8
; GFX10-NEXT: v_mov_b32_e32 v13, v11 ; GFX10-NEXT: v_mov_b32_e32 v10, v8
; GFX10-NEXT: v_mov_b32_e32 v14, v11 ; GFX10-NEXT: v_mov_b32_e32 v11, v8
; GFX10-NEXT: v_mov_b32_e32 v15, v11 ; GFX10-NEXT: v_mov_b32_e32 v12, v8
; GFX10-NEXT: v_mov_b32_e32 v0, v11 ; GFX10-NEXT: v_mov_b32_e32 v0, v8
; GFX10-NEXT: s_mov_b32 s1, s3 ; GFX10-NEXT: s_mov_b32 s1, s3
; GFX10-NEXT: s_mov_b32 s2, s4 ; GFX10-NEXT: s_mov_b32 s2, s4
; GFX10-NEXT: s_mov_b32 s3, s5 ; GFX10-NEXT: s_mov_b32 s3, s5
; GFX10-NEXT: s_mov_b32 s4, s6 ; GFX10-NEXT: s_mov_b32 s4, s6
; GFX10-NEXT: s_mov_b32 s5, s7 ; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8 ; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: s_mov_b32 s7, s9 ; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: v_mov_b32_e32 v1, v12 ; GFX10-NEXT: v_mov_b32_e32 v1, v9
; GFX10-NEXT: v_mov_b32_e32 v2, v13 ; GFX10-NEXT: v_mov_b32_e32 v2, v10
; GFX10-NEXT: v_mov_b32_e32 v3, v14 ; GFX10-NEXT: v_mov_b32_e32 v3, v11
; GFX10-NEXT: v_mov_b32_e32 v4, v15 ; GFX10-NEXT: v_mov_b32_e32 v4, v12
; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm tfe lwe ; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm tfe lwe
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: global_store_dword v11, v4, s[10:11] ; GFX10-NEXT: global_store_dword v8, v4, s[10:11]
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.3d.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 3, i32 0) %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.3d.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 3, i32 0)
%v.vec = extractvalue { <4 x float>, i32 } %v, 0 %v.vec = extractvalue { <4 x float>, i32 } %v, 0
%v.err = extractvalue { <4 x float>, i32 } %v, 1 %v.err = extractvalue { <4 x float>, i32 } %v, 1
store i32 %v.err, i32 addrspace(1)* %out, align 4 store i32 %v.err, i32 addrspace(1)* %out, align 4
ret <4 x float> %v.vec ret <4 x float> %v.vec
} }
declare <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i32(i32 immarg, i32, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0 declare <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i32(i32 immarg, i32, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
declare { <4 x float>, i32 } @llvm.amdgcn.image.load.3d.sl_v4f32i32s.i32(i32 immarg, i32, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0 declare { <4 x float>, i32 } @llvm.amdgcn.image.load.3d.sl_v4f32i32s.i32(i32 immarg, i32, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
attributes #0 = { nounwind readonly } attributes #0 = { nounwind readonly }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.sample.g16.ll

Show All 16 Lines
} }
define amdgpu_ps <4 x float> @sample_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_d_2d: ; GFX10-LABEL: sample_d_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v6, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v6, 0xffff
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT: v_and_or_b32 v7, v0, v6, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, v6, v1
; GFX10-NEXT: v_and_or_b32 v2, v2, v6, v3 ; GFX10-NEXT: v_and_or_b32 v1, v2, v6, v3
; GFX10-NEXT: image_sample_d_g16 v[0:3], [v7, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_d_g16 v[0:3], [v0, v1, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_d_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r) { define amdgpu_ps <4 x float> @sample_d_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r) {
; GFX10-LABEL: sample_d_3d: ; GFX10-LABEL: sample_d_3d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v11, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v9, 0xffff
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; GFX10-NEXT: s_lshl_b32 s12, s0, 16 ; GFX10-NEXT: s_lshl_b32 s12, s0, 16
; GFX10-NEXT: v_and_or_b32 v0, v0, v11, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, v9, v1
; GFX10-NEXT: v_and_or_b32 v1, v2, v11, s12 ; GFX10-NEXT: v_and_or_b32 v1, v2, v9, s12
; GFX10-NEXT: v_and_or_b32 v2, v3, v11, v4 ; GFX10-NEXT: v_and_or_b32 v2, v3, v9, v4
; GFX10-NEXT: v_and_or_b32 v3, v5, v11, s12 ; GFX10-NEXT: v_and_or_b32 v3, v5, v9, s12
; GFX10-NEXT: image_sample_d_g16 v[0:3], [v0, v1, v2, v3, v6, v7, v8], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_3D ; GFX10-NEXT: image_sample_d_g16 v[0:3], [v0, v1, v2, v3, v6, v7, v8], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_3D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
Show All 13 Lines
} }
define amdgpu_ps <4 x float> @sample_c_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_c_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_c_d_2d: ; GFX10-LABEL: sample_c_d_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; GFX10-NEXT: v_and_or_b32 v11, v1, v7, v2 ; GFX10-NEXT: v_and_or_b32 v1, v1, v7, v2
; GFX10-NEXT: v_and_or_b32 v2, v3, v7, v4 ; GFX10-NEXT: v_and_or_b32 v2, v3, v7, v4
; GFX10-NEXT: image_sample_c_d_g16 v[0:3], [v0, v11, v2, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_c_d_g16 v[0:3], [v0, v1, v2, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, float %s, float %clamp) { define amdgpu_ps <4 x float> @sample_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, float %s, float %clamp) {
; GFX10-LABEL: sample_d_cl_1d: ; GFX10-LABEL: sample_d_cl_1d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v4, 0xffff
; GFX10-NEXT: s_lshl_b32 s12, s0, 16 ; GFX10-NEXT: s_lshl_b32 s12, s0, 16
; GFX10-NEXT: v_and_or_b32 v0, v0, v7, s12 ; GFX10-NEXT: v_and_or_b32 v0, v0, v4, s12
; GFX10-NEXT: v_and_or_b32 v1, v1, v7, s12 ; GFX10-NEXT: v_and_or_b32 v1, v1, v4, s12
; GFX10-NEXT: image_sample_d_cl_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-NEXT: image_sample_d_cl_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) { define amdgpu_ps <4 x float> @sample_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
; GFX10-LABEL: sample_d_cl_2d: ; GFX10-LABEL: sample_d_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 16, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT: v_and_or_b32 v11, v0, v7, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, v7, v1
; GFX10-NEXT: v_and_or_b32 v1, v2, v7, v9 ; GFX10-NEXT: v_and_or_b32 v1, v2, v7, v3
; GFX10-NEXT: image_sample_d_cl_g16 v[0:3], [v11, v1, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_d_cl_g16 v[0:3], [v0, v1, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) { define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
; GFX10-LABEL: sample_c_d_cl_1d: ; GFX10-LABEL: sample_c_d_cl_1d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v5, 0xffff
; GFX10-NEXT: s_lshl_b32 s12, s0, 16 ; GFX10-NEXT: s_lshl_b32 s12, s0, 16
; GFX10-NEXT: v_and_or_b32 v1, v1, v7, s12 ; GFX10-NEXT: v_and_or_b32 v1, v1, v5, s12
; GFX10-NEXT: v_and_or_b32 v2, v2, v7, s12 ; GFX10-NEXT: v_and_or_b32 v2, v2, v5, s12
; GFX10-NEXT: image_sample_c_d_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-NEXT: image_sample_c_d_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) { define amdgpu_ps <4 x float> @sample_c_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
; GFX10-LABEL: sample_c_d_cl_2d: ; GFX10-LABEL: sample_c_d_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v8, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v8, 0xffff
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 16, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; GFX10-NEXT: v_and_or_b32 v1, v1, v8, v2 ; GFX10-NEXT: v_and_or_b32 v1, v1, v8, v2
; GFX10-NEXT: v_and_or_b32 v2, v3, v8, v10 ; GFX10-NEXT: v_and_or_b32 v2, v3, v8, v4
; GFX10-NEXT: image_sample_c_d_cl_g16 v[0:3], [v0, v1, v2, v5, v6, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_c_d_cl_g16 v[0:3], [v0, v1, v2, v5, v6, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
Show All 13 Lines
} }
define amdgpu_ps <4 x float> @sample_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_cd_2d: ; GFX10-LABEL: sample_cd_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v6, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v6, 0xffff
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT: v_and_or_b32 v7, v0, v6, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, v6, v1
; GFX10-NEXT: v_and_or_b32 v2, v2, v6, v3 ; GFX10-NEXT: v_and_or_b32 v1, v2, v6, v3
; GFX10-NEXT: image_sample_cd_g16 v[0:3], [v7, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_cd_g16 v[0:3], [v0, v1, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_cd_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s) { define amdgpu_ps <4 x float> @sample_c_cd_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s) {
Show All 12 Lines
} }
define amdgpu_ps <4 x float> @sample_c_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_c_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_c_cd_2d: ; GFX10-LABEL: sample_c_cd_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; GFX10-NEXT: v_and_or_b32 v11, v1, v7, v2 ; GFX10-NEXT: v_and_or_b32 v1, v1, v7, v2
; GFX10-NEXT: v_and_or_b32 v2, v3, v7, v4 ; GFX10-NEXT: v_and_or_b32 v2, v3, v7, v4
; GFX10-NEXT: image_sample_c_cd_g16 v[0:3], [v0, v11, v2, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_c_cd_g16 v[0:3], [v0, v1, v2, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, float %s, float %clamp) { define amdgpu_ps <4 x float> @sample_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, float %s, float %clamp) {
; GFX10-LABEL: sample_cd_cl_1d: ; GFX10-LABEL: sample_cd_cl_1d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v4, 0xffff
; GFX10-NEXT: s_lshl_b32 s12, s0, 16 ; GFX10-NEXT: s_lshl_b32 s12, s0, 16
; GFX10-NEXT: v_and_or_b32 v0, v0, v7, s12 ; GFX10-NEXT: v_and_or_b32 v0, v0, v4, s12
; GFX10-NEXT: v_and_or_b32 v1, v1, v7, s12 ; GFX10-NEXT: v_and_or_b32 v1, v1, v4, s12
; GFX10-NEXT: image_sample_cd_cl_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-NEXT: image_sample_cd_cl_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) { define amdgpu_ps <4 x float> @sample_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
; GFX10-LABEL: sample_cd_cl_2d: ; GFX10-LABEL: sample_cd_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 16, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT: v_and_or_b32 v11, v0, v7, v1 ; GFX10-NEXT: v_and_or_b32 v0, v0, v7, v1
; GFX10-NEXT: v_and_or_b32 v1, v2, v7, v9 ; GFX10-NEXT: v_and_or_b32 v1, v2, v7, v3
; GFX10-NEXT: image_sample_cd_cl_g16 v[0:3], [v11, v1, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_cd_cl_g16 v[0:3], [v0, v1, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) { define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
; GFX10-LABEL: sample_c_cd_cl_1d: ; GFX10-LABEL: sample_c_cd_cl_1d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v5, 0xffff
; GFX10-NEXT: s_lshl_b32 s12, s0, 16 ; GFX10-NEXT: s_lshl_b32 s12, s0, 16
; GFX10-NEXT: v_and_or_b32 v1, v1, v7, s12 ; GFX10-NEXT: v_and_or_b32 v1, v1, v5, s12
; GFX10-NEXT: v_and_or_b32 v2, v2, v7, s12 ; GFX10-NEXT: v_and_or_b32 v2, v2, v5, s12
; GFX10-NEXT: image_sample_c_cd_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-NEXT: image_sample_c_cd_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) { define amdgpu_ps <4 x float> @sample_c_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
; GFX10-LABEL: sample_c_cd_cl_2d: ; GFX10-LABEL: sample_c_cd_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v8, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v8, 0xffff
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 16, v4 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; GFX10-NEXT: v_and_or_b32 v1, v1, v8, v2 ; GFX10-NEXT: v_and_or_b32 v1, v1, v8, v2
; GFX10-NEXT: v_and_or_b32 v2, v3, v8, v10 ; GFX10-NEXT: v_and_or_b32 v2, v3, v8, v4
; GFX10-NEXT: image_sample_c_cd_cl_g16 v[0:3], [v0, v1, v2, v5, v6, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_c_cd_cl_g16 v[0:3], [v0, v1, v2, v5, v6, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps float @sample_c_d_o_2darray_V1(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice) { define amdgpu_ps float @sample_c_d_o_2darray_V1(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice) {
; GFX10-LABEL: sample_c_d_o_2darray_V1: ; GFX10-LABEL: sample_c_d_o_2darray_V1:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v9, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v9, 0xffff
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 16, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; GFX10-NEXT: v_and_or_b32 v2, v2, v9, v3 ; GFX10-NEXT: v_and_or_b32 v2, v2, v9, v3
; GFX10-NEXT: v_and_or_b32 v3, v4, v9, v11 ; GFX10-NEXT: v_and_or_b32 v3, v4, v9, v5
; GFX10-NEXT: image_sample_c_d_o_g16 v0, [v0, v1, v2, v3, v6, v7, v8], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D_ARRAY ; GFX10-NEXT: image_sample_c_d_o_g16 v0, [v0, v1, v2, v3, v6, v7, v8], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D_ARRAY
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call float @llvm.amdgcn.image.sample.c.d.o.2darray.f16.f32.f32(i32 4, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call float @llvm.amdgcn.image.sample.c.d.o.2darray.f16.f32.f32(i32 4, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret float %v ret float %v
} }
define amdgpu_ps <2 x float> @sample_c_d_o_2darray_V2(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice) { define amdgpu_ps <2 x float> @sample_c_d_o_2darray_V2(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice) {
; GFX10-LABEL: sample_c_d_o_2darray_V2: ; GFX10-LABEL: sample_c_d_o_2darray_V2:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v9, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v9, 0xffff
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 16, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; GFX10-NEXT: v_and_or_b32 v2, v2, v9, v3 ; GFX10-NEXT: v_and_or_b32 v2, v2, v9, v3
; GFX10-NEXT: v_and_or_b32 v3, v4, v9, v11 ; GFX10-NEXT: v_and_or_b32 v3, v4, v9, v5
; GFX10-NEXT: image_sample_c_d_o_g16 v[0:1], [v0, v1, v2, v3, v6, v7, v8], s[0:7], s[8:11] dmask:0x6 dim:SQ_RSRC_IMG_2D_ARRAY ; GFX10-NEXT: image_sample_c_d_o_g16 v[0:1], [v0, v1, v2, v3, v6, v7, v8], s[0:7], s[8:11] dmask:0x6 dim:SQ_RSRC_IMG_2D_ARRAY
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <2 x float> @llvm.amdgcn.image.sample.c.d.o.2darray.v2f32.f16.f32(i32 6, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <2 x float> @llvm.amdgcn.image.sample.c.d.o.2darray.v2f32.f16.f32(i32 6, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <2 x float> %v ret <2 x float> %v
} }
Show All 25 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.sdot4.ll

Show First 20 Lines • Show All 73 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_e32 v5, s5, v6 ; GFX10-NEXT: v_and_b32_e32 v5, s5, v6
; GFX10-NEXT: v_and_b32_e32 v6, s5, v7 ; GFX10-NEXT: v_and_b32_e32 v6, s5, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 24, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 24, v2
; GFX10-NEXT: v_and_or_b32 v3, v4, s5, v3 ; GFX10-NEXT: v_and_or_b32 v3, v4, s5, v3
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v5
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v6 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v6
; GFX10-NEXT: v_or3_b32 v7, v0, v1, v2 ; GFX10-NEXT: v_or3_b32 v0, v0, v1, v2
; GFX10-NEXT: v_or3_b32 v1, v3, v4, v5 ; GFX10-NEXT: v_or3_b32 v1, v3, v4, v5
; GFX10-NEXT: v_dot4_i32_i8 v0, v7, v1, v8 ; GFX10-NEXT: v_dot4_i32_i8 v0, v0, v1, v8
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%a.cast = bitcast <4 x i8> %a to i32 %a.cast = bitcast <4 x i8> %a to i32
%b.cast = bitcast <4 x i8> %b to i32 %b.cast = bitcast <4 x i8> %b to i32
%r = call i32 @llvm.amdgcn.sdot4(i32 %a.cast, i32 %b.cast, i32 %c, i1 false) %r = call i32 @llvm.amdgcn.sdot4(i32 %a.cast, i32 %b.cast, i32 %c, i1 false)
ret i32 %r ret i32 %r
} }
define i32 @v_sdot4_fnegf32_a(float %a, i32 %b, i32 %c) { define i32 @v_sdot4_fnegf32_a(float %a, i32 %b, i32 %c) {
▲ Show 20 Lines • Show All 44 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.udot4.ll

Show First 20 Lines • Show All 73 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v3, s4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_e32 v5, s5, v6 ; GFX10-NEXT: v_and_b32_e32 v5, s5, v6
; GFX10-NEXT: v_and_b32_e32 v6, s5, v7 ; GFX10-NEXT: v_and_b32_e32 v6, s5, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 24, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 24, v2
; GFX10-NEXT: v_and_or_b32 v3, v4, s5, v3 ; GFX10-NEXT: v_and_or_b32 v3, v4, s5, v3
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 16, v5
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v6 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 24, v6
; GFX10-NEXT: v_or3_b32 v7, v0, v1, v2 ; GFX10-NEXT: v_or3_b32 v0, v0, v1, v2
; GFX10-NEXT: v_or3_b32 v1, v3, v4, v5 ; GFX10-NEXT: v_or3_b32 v1, v3, v4, v5
; GFX10-NEXT: v_dot4_u32_u8 v0, v7, v1, v8 ; GFX10-NEXT: v_dot4_u32_u8 v0, v0, v1, v8
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%a.cast = bitcast <4 x i8> %a to i32 %a.cast = bitcast <4 x i8> %a to i32
%b.cast = bitcast <4 x i8> %b to i32 %b.cast = bitcast <4 x i8> %b to i32
%r = call i32 @llvm.amdgcn.udot4(i32 %a.cast, i32 %b.cast, i32 %c, i1 false) %r = call i32 @llvm.amdgcn.udot4(i32 %a.cast, i32 %b.cast, i32 %c, i1 false)
ret i32 %r ret i32 %r
} }
define i32 @v_udot4_fnegf32_a(float %a, i32 %b, i32 %c) { define i32 @v_udot4_fnegf32_a(float %a, i32 %b, i32 %c) {
▲ Show 20 Lines • Show All 44 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.wqm.demote.ll

Show First 20 Lines • Show All 345 Lines • ▼ Show 20 Lines
; GFX10-32-NEXT: v_cmp_ngt_f32_e32 vcc_lo, 0, v1 ; GFX10-32-NEXT: v_cmp_ngt_f32_e32 vcc_lo, 0, v1
; GFX10-32-NEXT: s_and_saveexec_b32 s13, vcc_lo ; GFX10-32-NEXT: s_and_saveexec_b32 s13, vcc_lo
; GFX10-32-NEXT: s_xor_b32 s13, exec_lo, s13 ; GFX10-32-NEXT: s_xor_b32 s13, exec_lo, s13
; GFX10-32-NEXT: s_cbranch_execz BB3_3 ; GFX10-32-NEXT: s_cbranch_execz BB3_3
; GFX10-32-NEXT: ; %bb.1: ; %.demote ; GFX10-32-NEXT: ; %bb.1: ; %.demote
; GFX10-32-NEXT: s_andn2_b32 s12, s12, exec_lo ; GFX10-32-NEXT: s_andn2_b32 s12, s12, exec_lo
; GFX10-32-NEXT: s_cbranch_scc0 BB3_4 ; GFX10-32-NEXT: s_cbranch_scc0 BB3_4
; GFX10-32-NEXT: ; %bb.2: ; %.demote ; GFX10-32-NEXT: ; %bb.2: ; %.demote
; GFX10-32-NEXT: s_wqm_b32 s28, s12 ; GFX10-32-NEXT: s_wqm_b32 s14, s12
; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10-32-NEXT: BB3_3: ; %.continue ; GFX10-32-NEXT: BB3_3: ; %.continue
; GFX10-32-NEXT: s_or_b32 exec_lo, exec_lo, s13 ; GFX10-32-NEXT: s_or_b32 exec_lo, exec_lo, s13
; GFX10-32-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-32-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-32-NEXT: s_waitcnt vmcnt(0) ; GFX10-32-NEXT: s_waitcnt vmcnt(0)
; GFX10-32-NEXT: v_add_f32_e32 v0, v0, v0 ; GFX10-32-NEXT: v_add_f32_e32 v0, v0, v0
; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s12 ; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s12
; GFX10-32-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-32-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-32-NEXT: s_waitcnt vmcnt(0) ; GFX10-32-NEXT: s_waitcnt vmcnt(0)
; GFX10-32-NEXT: s_branch BB3_5 ; GFX10-32-NEXT: s_branch BB3_5
; GFX10-32-NEXT: BB3_4: ; GFX10-32-NEXT: BB3_4:
; GFX10-32-NEXT: s_mov_b32 exec_lo, 0 ; GFX10-32-NEXT: s_mov_b32 exec_lo, 0
; GFX10-32-NEXT: exp null off, off, off, off done vm ; GFX10-32-NEXT: exp null off, off, off, off done vm
; GFX10-32-NEXT: s_endpgm ; GFX10-32-NEXT: s_endpgm
; GFX10-32-NEXT: BB3_5: ; GFX10-32-NEXT: BB3_5:
; ;
; GFX10-64-LABEL: wqm_demote_1: ; GFX10-64-LABEL: wqm_demote_1:
; GFX10-64: ; %bb.0: ; %.entry ; GFX10-64: ; %bb.0: ; %.entry
; GFX10-64-NEXT: s_mov_b64 s[12:13], exec ; GFX10-64-NEXT: s_mov_b64 s[12:13], exec
; GFX10-64-NEXT: s_wqm_b64 exec, exec ; GFX10-64-NEXT: s_wqm_b64 exec, exec
; GFX10-64-NEXT: v_cmp_ngt_f32_e32 vcc, 0, v1 ; GFX10-64-NEXT: v_cmp_ngt_f32_e32 vcc, 0, v1
; GFX10-64-NEXT: s_and_saveexec_b64 s[14:15], vcc ; GFX10-64-NEXT: s_and_saveexec_b64 s[14:15], vcc
; GFX10-64-NEXT: s_xor_b64 s[28:29], exec, s[14:15] ; GFX10-64-NEXT: s_xor_b64 s[14:15], exec, s[14:15]
; GFX10-64-NEXT: s_cbranch_execz BB3_3 ; GFX10-64-NEXT: s_cbranch_execz BB3_3
; GFX10-64-NEXT: ; %bb.1: ; %.demote ; GFX10-64-NEXT: ; %bb.1: ; %.demote
; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], exec ; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], exec
; GFX10-64-NEXT: s_cbranch_scc0 BB3_4 ; GFX10-64-NEXT: s_cbranch_scc0 BB3_4
; GFX10-64-NEXT: ; %bb.2: ; %.demote ; GFX10-64-NEXT: ; %bb.2: ; %.demote
; GFX10-64-NEXT: s_wqm_b64 s[16:17], s[12:13] ; GFX10-64-NEXT: s_wqm_b64 s[16:17], s[12:13]
; GFX10-64-NEXT: s_and_b64 exec, exec, s[16:17] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[16:17]
; GFX10-64-NEXT: BB3_3: ; %.continue ; GFX10-64-NEXT: BB3_3: ; %.continue
; GFX10-64-NEXT: s_or_b64 exec, exec, s[28:29] ; GFX10-64-NEXT: s_or_b64 exec, exec, s[14:15]
; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0 ; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0
; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13]
; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: s_branch BB3_5 ; GFX10-64-NEXT: s_branch BB3_5
; GFX10-64-NEXT: BB3_4: ; GFX10-64-NEXT: BB3_4:
▲ Show 20 Lines • Show All 87 Lines • ▼ Show 20 Lines
; GFX10-32-NEXT: v_cmp_ngt_f32_e32 vcc_lo, 0, v0 ; GFX10-32-NEXT: v_cmp_ngt_f32_e32 vcc_lo, 0, v0
; GFX10-32-NEXT: s_and_saveexec_b32 s13, vcc_lo ; GFX10-32-NEXT: s_and_saveexec_b32 s13, vcc_lo
; GFX10-32-NEXT: s_xor_b32 s13, exec_lo, s13 ; GFX10-32-NEXT: s_xor_b32 s13, exec_lo, s13
; GFX10-32-NEXT: s_cbranch_execz BB4_3 ; GFX10-32-NEXT: s_cbranch_execz BB4_3
; GFX10-32-NEXT: ; %bb.1: ; %.demote ; GFX10-32-NEXT: ; %bb.1: ; %.demote
; GFX10-32-NEXT: s_andn2_b32 s12, s12, exec_lo ; GFX10-32-NEXT: s_andn2_b32 s12, s12, exec_lo
; GFX10-32-NEXT: s_cbranch_scc0 BB4_4 ; GFX10-32-NEXT: s_cbranch_scc0 BB4_4
; GFX10-32-NEXT: ; %bb.2: ; %.demote ; GFX10-32-NEXT: ; %bb.2: ; %.demote
; GFX10-32-NEXT: s_wqm_b32 s28, s12 ; GFX10-32-NEXT: s_wqm_b32 s14, s12
; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10-32-NEXT: BB4_3: ; %.continue ; GFX10-32-NEXT: BB4_3: ; %.continue
; GFX10-32-NEXT: s_or_b32 exec_lo, exec_lo, s13 ; GFX10-32-NEXT: s_or_b32 exec_lo, exec_lo, s13
; GFX10-32-NEXT: v_add_f32_e32 v0, v0, v0 ; GFX10-32-NEXT: v_add_f32_e32 v0, v0, v0
; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s12 ; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s12
; GFX10-32-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-32-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-32-NEXT: s_waitcnt vmcnt(0) ; GFX10-32-NEXT: s_waitcnt vmcnt(0)
; GFX10-32-NEXT: s_branch BB4_5 ; GFX10-32-NEXT: s_branch BB4_5
; GFX10-32-NEXT: BB4_4: ; GFX10-32-NEXT: BB4_4:
; GFX10-32-NEXT: s_mov_b32 exec_lo, 0 ; GFX10-32-NEXT: s_mov_b32 exec_lo, 0
; GFX10-32-NEXT: exp null off, off, off, off done vm ; GFX10-32-NEXT: exp null off, off, off, off done vm
; GFX10-32-NEXT: s_endpgm ; GFX10-32-NEXT: s_endpgm
; GFX10-32-NEXT: BB4_5: ; GFX10-32-NEXT: BB4_5:
; ;
; GFX10-64-LABEL: wqm_demote_2: ; GFX10-64-LABEL: wqm_demote_2:
; GFX10-64: ; %bb.0: ; %.entry ; GFX10-64: ; %bb.0: ; %.entry
; GFX10-64-NEXT: s_mov_b64 s[12:13], exec ; GFX10-64-NEXT: s_mov_b64 s[12:13], exec
; GFX10-64-NEXT: s_wqm_b64 exec, exec ; GFX10-64-NEXT: s_wqm_b64 exec, exec
; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: v_cmp_ngt_f32_e32 vcc, 0, v0 ; GFX10-64-NEXT: v_cmp_ngt_f32_e32 vcc, 0, v0
; GFX10-64-NEXT: s_and_saveexec_b64 s[14:15], vcc ; GFX10-64-NEXT: s_and_saveexec_b64 s[14:15], vcc
; GFX10-64-NEXT: s_xor_b64 s[28:29], exec, s[14:15] ; GFX10-64-NEXT: s_xor_b64 s[14:15], exec, s[14:15]
; GFX10-64-NEXT: s_cbranch_execz BB4_3 ; GFX10-64-NEXT: s_cbranch_execz BB4_3
; GFX10-64-NEXT: ; %bb.1: ; %.demote ; GFX10-64-NEXT: ; %bb.1: ; %.demote
; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], exec ; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], exec
; GFX10-64-NEXT: s_cbranch_scc0 BB4_4 ; GFX10-64-NEXT: s_cbranch_scc0 BB4_4
; GFX10-64-NEXT: ; %bb.2: ; %.demote ; GFX10-64-NEXT: ; %bb.2: ; %.demote
; GFX10-64-NEXT: s_wqm_b64 s[16:17], s[12:13] ; GFX10-64-NEXT: s_wqm_b64 s[16:17], s[12:13]
; GFX10-64-NEXT: s_and_b64 exec, exec, s[16:17] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[16:17]
; GFX10-64-NEXT: BB4_3: ; %.continue ; GFX10-64-NEXT: BB4_3: ; %.continue
; GFX10-64-NEXT: s_or_b64 exec, exec, s[28:29] ; GFX10-64-NEXT: s_or_b64 exec, exec, s[14:15]
; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0 ; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0
; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13]
; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: s_branch BB4_5 ; GFX10-64-NEXT: s_branch BB4_5
; GFX10-64-NEXT: BB4_4: ; GFX10-64-NEXT: BB4_4:
; GFX10-64-NEXT: s_mov_b64 exec, 0 ; GFX10-64-NEXT: s_mov_b64 exec, 0
; GFX10-64-NEXT: exp null off, off, off, off done vm ; GFX10-64-NEXT: exp null off, off, off, off done vm
▲ Show 20 Lines • Show All 96 Lines • ▼ Show 20 Lines
; GFX10-64-NEXT: s_wqm_b64 exec, exec ; GFX10-64-NEXT: s_wqm_b64 exec, exec
; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: v_cmp_gt_f32_e32 vcc, 0, v0 ; GFX10-64-NEXT: v_cmp_gt_f32_e32 vcc, 0, v0
; GFX10-64-NEXT: s_xor_b64 s[14:15], vcc, exec ; GFX10-64-NEXT: s_xor_b64 s[14:15], vcc, exec
; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], s[14:15] ; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], s[14:15]
; GFX10-64-NEXT: s_cbranch_scc0 BB5_2 ; GFX10-64-NEXT: s_cbranch_scc0 BB5_2
; GFX10-64-NEXT: ; %bb.1: ; %.entry ; GFX10-64-NEXT: ; %bb.1: ; %.entry
; GFX10-64-NEXT: s_wqm_b64 s[28:29], s[12:13] ; GFX10-64-NEXT: s_wqm_b64 s[14:15], s[12:13]
; GFX10-64-NEXT: s_and_b64 exec, exec, s[28:29] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0 ; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0
; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13]
; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: s_branch BB5_3 ; GFX10-64-NEXT: s_branch BB5_3
; GFX10-64-NEXT: BB5_2: ; GFX10-64-NEXT: BB5_2:
; GFX10-64-NEXT: s_mov_b64 exec, 0 ; GFX10-64-NEXT: s_mov_b64 exec, 0
; GFX10-64-NEXT: exp null off, off, off, off done vm ; GFX10-64-NEXT: exp null off, off, off, off done vm
▲ Show 20 Lines • Show All 566 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/load-local.128.ll

Show First 20 Lines • Show All 186 Lines • ▼ Show 20 Lines
; GFX10-NEXT: ds_read_u8 v5, v0 offset:6 ; GFX10-NEXT: ds_read_u8 v5, v0 offset:6
; GFX10-NEXT: ds_read_u8 v6, v0 offset:7 ; GFX10-NEXT: ds_read_u8 v6, v0 offset:7
; GFX10-NEXT: ds_read_u8 v7, v0 offset:9 ; GFX10-NEXT: ds_read_u8 v7, v0 offset:9
; GFX10-NEXT: ds_read_u8 v8, v0 ; GFX10-NEXT: ds_read_u8 v8, v0
; GFX10-NEXT: ds_read_u8 v9, v0 offset:4 ; GFX10-NEXT: ds_read_u8 v9, v0 offset:4
; GFX10-NEXT: ds_read_u8 v10, v0 offset:8 ; GFX10-NEXT: ds_read_u8 v10, v0 offset:8
; GFX10-NEXT: ds_read_u8 v12, v0 offset:10 ; GFX10-NEXT: ds_read_u8 v12, v0 offset:10
; GFX10-NEXT: ds_read_u8 v13, v0 offset:11 ; GFX10-NEXT: ds_read_u8 v13, v0 offset:11
; GFX10-NEXT: ds_read_u8 v25, v0 offset:12 ; GFX10-NEXT: ds_read_u8 v14, v0 offset:12
; GFX10-NEXT: ds_read_u8 v15, v0 offset:13 ; GFX10-NEXT: ds_read_u8 v15, v0 offset:13
; GFX10-NEXT: ds_read_u8 v16, v0 offset:14 ; GFX10-NEXT: ds_read_u8 v16, v0 offset:14
; GFX10-NEXT: ds_read_u8 v0, v0 offset:15 ; GFX10-NEXT: ds_read_u8 v0, v0 offset:15
; GFX10-NEXT: v_mov_b32_e32 v17, 8 ; GFX10-NEXT: v_mov_b32_e32 v17, 8
; GFX10-NEXT: s_mov_b32 s5, 8 ; GFX10-NEXT: s_mov_b32 s5, 8
; GFX10-NEXT: v_mov_b32_e32 v11, 0xff ; GFX10-NEXT: v_mov_b32_e32 v11, 0xff
; GFX10-NEXT: s_movk_i32 s4, 0xff ; GFX10-NEXT: s_movk_i32 s4, 0xff
; GFX10-NEXT: s_waitcnt lgkmcnt(15) ; GFX10-NEXT: s_waitcnt lgkmcnt(15)
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(14) ; GFX10-NEXT: s_waitcnt lgkmcnt(14)
; GFX10-NEXT: v_and_b32_e32 v2, s4, v2 ; GFX10-NEXT: v_and_b32_e32 v2, s4, v2
; GFX10-NEXT: s_waitcnt lgkmcnt(13) ; GFX10-NEXT: s_waitcnt lgkmcnt(13)
; GFX10-NEXT: v_and_b32_e32 v3, s4, v3 ; GFX10-NEXT: v_and_b32_e32 v3, s4, v3
; GFX10-NEXT: s_waitcnt lgkmcnt(12) ; GFX10-NEXT: s_waitcnt lgkmcnt(12)
; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s5, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s5, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(11) ; GFX10-NEXT: s_waitcnt lgkmcnt(11)
; GFX10-NEXT: v_and_b32_e32 v5, v5, v11 ; GFX10-NEXT: v_and_b32_e32 v5, v5, v11
; GFX10-NEXT: s_waitcnt lgkmcnt(10) ; GFX10-NEXT: s_waitcnt lgkmcnt(10)
; GFX10-NEXT: v_and_b32_e32 v6, v6, v11 ; GFX10-NEXT: v_and_b32_e32 v6, v6, v11
; GFX10-NEXT: s_waitcnt lgkmcnt(9) ; GFX10-NEXT: s_waitcnt lgkmcnt(9)
; GFX10-NEXT: v_lshlrev_b32_sdwa v21, v17, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v17, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(8) ; GFX10-NEXT: s_waitcnt lgkmcnt(8)
; GFX10-NEXT: v_and_or_b32 v1, v8, s4, v1 ; GFX10-NEXT: v_and_or_b32 v1, v8, s4, v1
; GFX10-NEXT: s_waitcnt lgkmcnt(5) ; GFX10-NEXT: s_waitcnt lgkmcnt(5)
; GFX10-NEXT: v_and_b32_e32 v8, v12, v11 ; GFX10-NEXT: v_and_b32_e32 v8, v12, v11
; GFX10-NEXT: v_and_or_b32 v4, v9, s4, v4 ; GFX10-NEXT: v_and_or_b32 v4, v9, s4, v4
; GFX10-NEXT: s_waitcnt lgkmcnt(4) ; GFX10-NEXT: s_waitcnt lgkmcnt(4)
; GFX10-NEXT: v_and_b32_e32 v9, v13, v11 ; GFX10-NEXT: v_and_b32_e32 v9, v13, v11
; GFX10-NEXT: v_and_or_b32 v7, v10, v11, v21 ; GFX10-NEXT: v_and_or_b32 v7, v10, v11, v7
; GFX10-NEXT: s_waitcnt lgkmcnt(2) ; GFX10-NEXT: s_waitcnt lgkmcnt(2)
; GFX10-NEXT: v_lshlrev_b32_sdwa v10, v17, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v10, v17, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(1) ; GFX10-NEXT: s_waitcnt lgkmcnt(1)
; GFX10-NEXT: v_and_b32_e32 v12, v16, v11 ; GFX10-NEXT: v_and_b32_e32 v12, v16, v11
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_and_b32_e32 v0, v0, v11 ; GFX10-NEXT: v_and_b32_e32 v0, v0, v11
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; GFX10-NEXT: v_and_or_b32 v10, v25, v11, v10 ; GFX10-NEXT: v_and_or_b32 v10, v14, v11, v10
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 16, v12 ; GFX10-NEXT: v_lshlrev_b32_e32 v11, 16, v12
; GFX10-NEXT: v_lshlrev_b32_e32 v12, 24, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v12, 24, v0
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6 ; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 16, v8 ; GFX10-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9 ; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9
; GFX10-NEXT: v_or3_b32 v0, v1, v2, v3 ; GFX10-NEXT: v_or3_b32 v0, v1, v2, v3
; GFX10-NEXT: v_or3_b32 v3, v10, v11, v12 ; GFX10-NEXT: v_or3_b32 v3, v10, v11, v12
▲ Show 20 Lines • Show All 198 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/load-local.96.ll

Show First 20 Lines • Show All 152 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: ds_read_u8 v1, v0 offset:1 ; GFX10-NEXT: ds_read_u8 v1, v0 offset:1
; GFX10-NEXT: ds_read_u8 v2, v0 offset:2 ; GFX10-NEXT: ds_read_u8 v2, v0 offset:2
; GFX10-NEXT: ds_read_u8 v3, v0 offset:3 ; GFX10-NEXT: ds_read_u8 v3, v0 offset:3
; GFX10-NEXT: ds_read_u8 v4, v0 offset:5 ; GFX10-NEXT: ds_read_u8 v4, v0 offset:5
; GFX10-NEXT: ds_read_u8 v5, v0 offset:6 ; GFX10-NEXT: ds_read_u8 v5, v0 offset:6
; GFX10-NEXT: ds_read_u8 v6, v0 offset:7 ; GFX10-NEXT: ds_read_u8 v6, v0 offset:7
; GFX10-NEXT: ds_read_u8 v7, v0 offset:9 ; GFX10-NEXT: ds_read_u8 v7, v0 offset:9
; GFX10-NEXT: ds_read_u8 v15, v0 offset:10 ; GFX10-NEXT: ds_read_u8 v8, v0 offset:10
; GFX10-NEXT: ds_read_u8 v9, v0 offset:11 ; GFX10-NEXT: ds_read_u8 v9, v0 offset:11
; GFX10-NEXT: ds_read_u8 v10, v0 ; GFX10-NEXT: ds_read_u8 v10, v0
; GFX10-NEXT: ds_read_u8 v11, v0 offset:4 ; GFX10-NEXT: ds_read_u8 v11, v0 offset:4
; GFX10-NEXT: ds_read_u8 v14, v0 offset:8 ; GFX10-NEXT: ds_read_u8 v0, v0 offset:8
; GFX10-NEXT: v_mov_b32_e32 v12, 0xff ; GFX10-NEXT: v_mov_b32_e32 v12, 0xff
; GFX10-NEXT: v_mov_b32_e32 v13, 8 ; GFX10-NEXT: v_mov_b32_e32 v13, 8
; GFX10-NEXT: s_movk_i32 s4, 0xff ; GFX10-NEXT: s_movk_i32 s4, 0xff
; GFX10-NEXT: s_mov_b32 s5, 8 ; GFX10-NEXT: s_mov_b32 s5, 8
; GFX10-NEXT: s_waitcnt lgkmcnt(11) ; GFX10-NEXT: s_waitcnt lgkmcnt(11)
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(10) ; GFX10-NEXT: s_waitcnt lgkmcnt(10)
; GFX10-NEXT: v_and_b32_e32 v2, s4, v2 ; GFX10-NEXT: v_and_b32_e32 v2, s4, v2
; GFX10-NEXT: s_waitcnt lgkmcnt(9) ; GFX10-NEXT: s_waitcnt lgkmcnt(9)
; GFX10-NEXT: v_and_b32_e32 v3, s4, v3 ; GFX10-NEXT: v_and_b32_e32 v3, s4, v3
; GFX10-NEXT: s_waitcnt lgkmcnt(8) ; GFX10-NEXT: s_waitcnt lgkmcnt(8)
; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s5, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s5, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(7) ; GFX10-NEXT: s_waitcnt lgkmcnt(7)
; GFX10-NEXT: v_and_b32_e32 v5, v5, v12 ; GFX10-NEXT: v_and_b32_e32 v5, v5, v12
; GFX10-NEXT: s_waitcnt lgkmcnt(6) ; GFX10-NEXT: s_waitcnt lgkmcnt(6)
; GFX10-NEXT: v_and_b32_e32 v6, v6, v12 ; GFX10-NEXT: v_and_b32_e32 v6, v6, v12
; GFX10-NEXT: s_waitcnt lgkmcnt(5) ; GFX10-NEXT: s_waitcnt lgkmcnt(5)
; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v13, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v13, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(4) ; GFX10-NEXT: s_waitcnt lgkmcnt(4)
; GFX10-NEXT: v_and_b32_e32 v8, v15, v12 ; GFX10-NEXT: v_and_b32_e32 v8, v8, v12
; GFX10-NEXT: s_waitcnt lgkmcnt(3) ; GFX10-NEXT: s_waitcnt lgkmcnt(3)
; GFX10-NEXT: v_and_b32_e32 v9, v9, v12 ; GFX10-NEXT: v_and_b32_e32 v9, v9, v12
; GFX10-NEXT: s_waitcnt lgkmcnt(2) ; GFX10-NEXT: s_waitcnt lgkmcnt(2)
; GFX10-NEXT: v_and_or_b32 v1, v10, s4, v1 ; GFX10-NEXT: v_and_or_b32 v1, v10, s4, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; GFX10-NEXT: s_waitcnt lgkmcnt(1) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_and_or_b32 v7, v0, v12, v7
; GFX10-NEXT: v_and_or_b32 v4, v11, s4, v4 ; GFX10-NEXT: v_and_or_b32 v4, v11, s4, v4
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6 ; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_and_or_b32 v7, v14, v12, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 16, v8 ; GFX10-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9 ; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9
; GFX10-NEXT: v_or3_b32 v0, v1, v2, v3 ; GFX10-NEXT: v_or3_b32 v0, v1, v2, v3
; GFX10-NEXT: v_or3_b32 v1, v4, v5, v6 ; GFX10-NEXT: v_or3_b32 v1, v4, v5, v6
; GFX10-NEXT: v_or3_b32 v2, v7, v8, v9 ; GFX10-NEXT: v_or3_b32 v2, v7, v8, v9
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%load = load <3 x i32>, <3 x i32> addrspace(3)* %ptr, align 1 %load = load <3 x i32>, <3 x i32> addrspace(3)* %ptr, align 1
ret <3 x i32> %load ret <3 x i32> %load
▲ Show 20 Lines • Show All 55 Lines • ▼ Show 20 Lines
; ;
; GFX10-LABEL: load_lds_v3i32_align2: ; GFX10-LABEL: load_lds_v3i32_align2:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: ds_read_u16 v1, v0 offset:2 ; GFX10-NEXT: ds_read_u16 v1, v0 offset:2
; GFX10-NEXT: ds_read_u16 v2, v0 offset:6 ; GFX10-NEXT: ds_read_u16 v2, v0 offset:6
; GFX10-NEXT: ds_read_u16 v3, v0 offset:10 ; GFX10-NEXT: ds_read_u16 v3, v0 offset:10
; GFX10-NEXT: ds_read_u16 v7, v0 ; GFX10-NEXT: ds_read_u16 v4, v0
; GFX10-NEXT: ds_read_u16 v11, v0 offset:4 ; GFX10-NEXT: ds_read_u16 v5, v0 offset:4
; GFX10-NEXT: ds_read_u16 v15, v0 offset:8 ; GFX10-NEXT: ds_read_u16 v6, v0 offset:8
; GFX10-NEXT: s_mov_b32 s4, 0xffff ; GFX10-NEXT: s_mov_b32 s4, 0xffff
; GFX10-NEXT: s_waitcnt lgkmcnt(5) ; GFX10-NEXT: s_waitcnt lgkmcnt(5)
; GFX10-NEXT: v_and_b32_e32 v0, s4, v1 ; GFX10-NEXT: v_and_b32_e32 v0, s4, v1
; GFX10-NEXT: s_waitcnt lgkmcnt(4) ; GFX10-NEXT: s_waitcnt lgkmcnt(4)
; GFX10-NEXT: v_and_b32_e32 v1, s4, v2 ; GFX10-NEXT: v_and_b32_e32 v1, s4, v2
; GFX10-NEXT: s_waitcnt lgkmcnt(3) ; GFX10-NEXT: s_waitcnt lgkmcnt(3)
; GFX10-NEXT: v_and_b32_e32 v2, s4, v3 ; GFX10-NEXT: v_and_b32_e32 v2, s4, v3
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1 ; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT: s_waitcnt lgkmcnt(2) ; GFX10-NEXT: s_waitcnt lgkmcnt(2)
; GFX10-NEXT: v_and_or_b32 v0, v7, s4, v0 ; GFX10-NEXT: v_and_or_b32 v0, v4, s4, v0
; GFX10-NEXT: s_waitcnt lgkmcnt(1) ; GFX10-NEXT: s_waitcnt lgkmcnt(1)
; GFX10-NEXT: v_and_or_b32 v1, v11, s4, v1 ; GFX10-NEXT: v_and_or_b32 v1, v5, s4, v1
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_and_or_b32 v2, v15, s4, v2 ; GFX10-NEXT: v_and_or_b32 v2, v6, s4, v2
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%load = load <3 x i32>, <3 x i32> addrspace(3)* %ptr, align 2 %load = load <3 x i32>, <3 x i32> addrspace(3)* %ptr, align 2
ret <3 x i32> %load ret <3 x i32> %load
} }
define <3 x i32> @load_lds_v3i32_align4(<3 x i32> addrspace(3)* %ptr) { define <3 x i32> @load_lds_v3i32_align4(<3 x i32> addrspace(3)* %ptr) {
; GFX9-LABEL: load_lds_v3i32_align4: ; GFX9-LABEL: load_lds_v3i32_align4:
; GFX9: ; %bb.0: ; GFX9: ; %bb.0:
▲ Show 20 Lines • Show All 89 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/load-unaligned.ll

Show First 20 Lines • Show All 102 Lines • ▼ Show 20 Lines
; GFX10-NEXT: ds_read_u8 v5, v0 offset:6 ; GFX10-NEXT: ds_read_u8 v5, v0 offset:6
; GFX10-NEXT: ds_read_u8 v6, v0 offset:7 ; GFX10-NEXT: ds_read_u8 v6, v0 offset:7
; GFX10-NEXT: ds_read_u8 v7, v0 offset:9 ; GFX10-NEXT: ds_read_u8 v7, v0 offset:9
; GFX10-NEXT: ds_read_u8 v8, v0 ; GFX10-NEXT: ds_read_u8 v8, v0
; GFX10-NEXT: ds_read_u8 v9, v0 offset:4 ; GFX10-NEXT: ds_read_u8 v9, v0 offset:4
; GFX10-NEXT: ds_read_u8 v10, v0 offset:8 ; GFX10-NEXT: ds_read_u8 v10, v0 offset:8
; GFX10-NEXT: ds_read_u8 v12, v0 offset:10 ; GFX10-NEXT: ds_read_u8 v12, v0 offset:10
; GFX10-NEXT: ds_read_u8 v13, v0 offset:11 ; GFX10-NEXT: ds_read_u8 v13, v0 offset:11
; GFX10-NEXT: ds_read_u8 v25, v0 offset:12 ; GFX10-NEXT: ds_read_u8 v14, v0 offset:12
; GFX10-NEXT: ds_read_u8 v15, v0 offset:13 ; GFX10-NEXT: ds_read_u8 v15, v0 offset:13
; GFX10-NEXT: ds_read_u8 v16, v0 offset:14 ; GFX10-NEXT: ds_read_u8 v16, v0 offset:14
; GFX10-NEXT: ds_read_u8 v0, v0 offset:15 ; GFX10-NEXT: ds_read_u8 v0, v0 offset:15
; GFX10-NEXT: v_mov_b32_e32 v17, 8 ; GFX10-NEXT: v_mov_b32_e32 v17, 8
; GFX10-NEXT: s_mov_b32 s5, 8 ; GFX10-NEXT: s_mov_b32 s5, 8
; GFX10-NEXT: v_mov_b32_e32 v11, 0xff ; GFX10-NEXT: v_mov_b32_e32 v11, 0xff
; GFX10-NEXT: s_movk_i32 s4, 0xff ; GFX10-NEXT: s_movk_i32 s4, 0xff
; GFX10-NEXT: s_waitcnt lgkmcnt(15) ; GFX10-NEXT: s_waitcnt lgkmcnt(15)
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(14) ; GFX10-NEXT: s_waitcnt lgkmcnt(14)
; GFX10-NEXT: v_and_b32_e32 v2, s4, v2 ; GFX10-NEXT: v_and_b32_e32 v2, s4, v2
; GFX10-NEXT: s_waitcnt lgkmcnt(13) ; GFX10-NEXT: s_waitcnt lgkmcnt(13)
; GFX10-NEXT: v_and_b32_e32 v3, s4, v3 ; GFX10-NEXT: v_and_b32_e32 v3, s4, v3
; GFX10-NEXT: s_waitcnt lgkmcnt(12) ; GFX10-NEXT: s_waitcnt lgkmcnt(12)
; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s5, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s5, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(11) ; GFX10-NEXT: s_waitcnt lgkmcnt(11)
; GFX10-NEXT: v_and_b32_e32 v5, v5, v11 ; GFX10-NEXT: v_and_b32_e32 v5, v5, v11
; GFX10-NEXT: s_waitcnt lgkmcnt(10) ; GFX10-NEXT: s_waitcnt lgkmcnt(10)
; GFX10-NEXT: v_and_b32_e32 v6, v6, v11 ; GFX10-NEXT: v_and_b32_e32 v6, v6, v11
; GFX10-NEXT: s_waitcnt lgkmcnt(9) ; GFX10-NEXT: s_waitcnt lgkmcnt(9)
; GFX10-NEXT: v_lshlrev_b32_sdwa v21, v17, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v17, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(8) ; GFX10-NEXT: s_waitcnt lgkmcnt(8)
; GFX10-NEXT: v_and_or_b32 v1, v8, s4, v1 ; GFX10-NEXT: v_and_or_b32 v1, v8, s4, v1
; GFX10-NEXT: s_waitcnt lgkmcnt(5) ; GFX10-NEXT: s_waitcnt lgkmcnt(5)
; GFX10-NEXT: v_and_b32_e32 v8, v12, v11 ; GFX10-NEXT: v_and_b32_e32 v8, v12, v11
; GFX10-NEXT: v_and_or_b32 v4, v9, s4, v4 ; GFX10-NEXT: v_and_or_b32 v4, v9, s4, v4
; GFX10-NEXT: s_waitcnt lgkmcnt(4) ; GFX10-NEXT: s_waitcnt lgkmcnt(4)
; GFX10-NEXT: v_and_b32_e32 v9, v13, v11 ; GFX10-NEXT: v_and_b32_e32 v9, v13, v11
; GFX10-NEXT: v_and_or_b32 v7, v10, v11, v21 ; GFX10-NEXT: v_and_or_b32 v7, v10, v11, v7
; GFX10-NEXT: s_waitcnt lgkmcnt(2) ; GFX10-NEXT: s_waitcnt lgkmcnt(2)
; GFX10-NEXT: v_lshlrev_b32_sdwa v10, v17, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v10, v17, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(1) ; GFX10-NEXT: s_waitcnt lgkmcnt(1)
; GFX10-NEXT: v_and_b32_e32 v12, v16, v11 ; GFX10-NEXT: v_and_b32_e32 v12, v16, v11
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_and_b32_e32 v0, v0, v11 ; GFX10-NEXT: v_and_b32_e32 v0, v0, v11
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; GFX10-NEXT: v_and_or_b32 v10, v25, v11, v10 ; GFX10-NEXT: v_and_or_b32 v10, v14, v11, v10
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 16, v12 ; GFX10-NEXT: v_lshlrev_b32_e32 v11, 16, v12
; GFX10-NEXT: v_lshlrev_b32_e32 v12, 24, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v12, 24, v0
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6 ; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 16, v8 ; GFX10-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9 ; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9
; GFX10-NEXT: v_or3_b32 v0, v1, v2, v3 ; GFX10-NEXT: v_or3_b32 v0, v1, v2, v3
; GFX10-NEXT: v_or3_b32 v3, v10, v11, v12 ; GFX10-NEXT: v_or3_b32 v3, v10, v11, v12
▲ Show 20 Lines • Show All 79 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: ds_read_u8 v1, v0 offset:1 ; GFX10-NEXT: ds_read_u8 v1, v0 offset:1
; GFX10-NEXT: ds_read_u8 v2, v0 offset:2 ; GFX10-NEXT: ds_read_u8 v2, v0 offset:2
; GFX10-NEXT: ds_read_u8 v3, v0 offset:3 ; GFX10-NEXT: ds_read_u8 v3, v0 offset:3
; GFX10-NEXT: ds_read_u8 v4, v0 offset:5 ; GFX10-NEXT: ds_read_u8 v4, v0 offset:5
; GFX10-NEXT: ds_read_u8 v5, v0 offset:6 ; GFX10-NEXT: ds_read_u8 v5, v0 offset:6
; GFX10-NEXT: ds_read_u8 v6, v0 offset:7 ; GFX10-NEXT: ds_read_u8 v6, v0 offset:7
; GFX10-NEXT: ds_read_u8 v7, v0 offset:9 ; GFX10-NEXT: ds_read_u8 v7, v0 offset:9
; GFX10-NEXT: ds_read_u8 v15, v0 offset:10 ; GFX10-NEXT: ds_read_u8 v8, v0 offset:10
; GFX10-NEXT: ds_read_u8 v9, v0 offset:11 ; GFX10-NEXT: ds_read_u8 v9, v0 offset:11
; GFX10-NEXT: ds_read_u8 v10, v0 ; GFX10-NEXT: ds_read_u8 v10, v0
; GFX10-NEXT: ds_read_u8 v11, v0 offset:4 ; GFX10-NEXT: ds_read_u8 v11, v0 offset:4
; GFX10-NEXT: ds_read_u8 v14, v0 offset:8 ; GFX10-NEXT: ds_read_u8 v0, v0 offset:8
; GFX10-NEXT: v_mov_b32_e32 v12, 0xff ; GFX10-NEXT: v_mov_b32_e32 v12, 0xff
; GFX10-NEXT: v_mov_b32_e32 v13, 8 ; GFX10-NEXT: v_mov_b32_e32 v13, 8
; GFX10-NEXT: s_movk_i32 s4, 0xff ; GFX10-NEXT: s_movk_i32 s4, 0xff
; GFX10-NEXT: s_mov_b32 s5, 8 ; GFX10-NEXT: s_mov_b32 s5, 8
; GFX10-NEXT: s_waitcnt lgkmcnt(11) ; GFX10-NEXT: s_waitcnt lgkmcnt(11)
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(10) ; GFX10-NEXT: s_waitcnt lgkmcnt(10)
; GFX10-NEXT: v_and_b32_e32 v2, s4, v2 ; GFX10-NEXT: v_and_b32_e32 v2, s4, v2
; GFX10-NEXT: s_waitcnt lgkmcnt(9) ; GFX10-NEXT: s_waitcnt lgkmcnt(9)
; GFX10-NEXT: v_and_b32_e32 v3, s4, v3 ; GFX10-NEXT: v_and_b32_e32 v3, s4, v3
; GFX10-NEXT: s_waitcnt lgkmcnt(8) ; GFX10-NEXT: s_waitcnt lgkmcnt(8)
; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s5, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v4, s5, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(7) ; GFX10-NEXT: s_waitcnt lgkmcnt(7)
; GFX10-NEXT: v_and_b32_e32 v5, v5, v12 ; GFX10-NEXT: v_and_b32_e32 v5, v5, v12
; GFX10-NEXT: s_waitcnt lgkmcnt(6) ; GFX10-NEXT: s_waitcnt lgkmcnt(6)
; GFX10-NEXT: v_and_b32_e32 v6, v6, v12 ; GFX10-NEXT: v_and_b32_e32 v6, v6, v12
; GFX10-NEXT: s_waitcnt lgkmcnt(5) ; GFX10-NEXT: s_waitcnt lgkmcnt(5)
; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v13, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX10-NEXT: v_lshlrev_b32_sdwa v7, v13, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: s_waitcnt lgkmcnt(4) ; GFX10-NEXT: s_waitcnt lgkmcnt(4)
; GFX10-NEXT: v_and_b32_e32 v8, v15, v12 ; GFX10-NEXT: v_and_b32_e32 v8, v8, v12
; GFX10-NEXT: s_waitcnt lgkmcnt(3) ; GFX10-NEXT: s_waitcnt lgkmcnt(3)
; GFX10-NEXT: v_and_b32_e32 v9, v9, v12 ; GFX10-NEXT: v_and_b32_e32 v9, v9, v12
; GFX10-NEXT: s_waitcnt lgkmcnt(2) ; GFX10-NEXT: s_waitcnt lgkmcnt(2)
; GFX10-NEXT: v_and_or_b32 v1, v10, s4, v1 ; GFX10-NEXT: v_and_or_b32 v1, v10, s4, v1
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2 ; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; GFX10-NEXT: s_waitcnt lgkmcnt(1) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_and_or_b32 v7, v0, v12, v7
; GFX10-NEXT: v_and_or_b32 v4, v11, s4, v4 ; GFX10-NEXT: v_and_or_b32 v4, v11, s4, v4
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v5 ; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6 ; GFX10-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_and_or_b32 v7, v14, v12, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v8, 16, v8 ; GFX10-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9 ; GFX10-NEXT: v_lshlrev_b32_e32 v9, 24, v9
; GFX10-NEXT: v_or3_b32 v0, v1, v2, v3 ; GFX10-NEXT: v_or3_b32 v0, v1, v2, v3
; GFX10-NEXT: v_or3_b32 v1, v4, v5, v6 ; GFX10-NEXT: v_or3_b32 v1, v4, v5, v6
; GFX10-NEXT: v_or3_b32 v2, v7, v8, v9 ; GFX10-NEXT: v_or3_b32 v2, v7, v8, v9
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%load = load <3 x i32>, <3 x i32> addrspace(3)* %ptr, align 1 %load = load <3 x i32>, <3 x i32> addrspace(3)* %ptr, align 1
ret <3 x i32> %load ret <3 x i32> %load
▲ Show 20 Lines • Show All 115 Lines • ▼ Show 20 Lines
; GFX7-NEXT: ds_write_b8 v0, v4 offset:11 ; GFX7-NEXT: ds_write_b8 v0, v4 offset:11
; GFX7-NEXT: s_waitcnt lgkmcnt(0) ; GFX7-NEXT: s_waitcnt lgkmcnt(0)
; GFX7-NEXT: s_setpc_b64 s[30:31] ; GFX7-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: store_lds_v3i32_align1: ; GFX10-LABEL: store_lds_v3i32_align1:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 16, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 16, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 24, v1
; GFX10-NEXT: ds_write_b8 v0, v1 ; GFX10-NEXT: ds_write_b8 v0, v1
; GFX10-NEXT: ds_write_b8 v0, v7 offset:1 ; GFX10-NEXT: ds_write_b8 v0, v4 offset:1
; GFX10-NEXT: ds_write_b8 v0, v5 offset:2 ; GFX10-NEXT: ds_write_b8 v0, v5 offset:2
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 16, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v2
; GFX10-NEXT: ds_write_b8 v0, v6 offset:3 ; GFX10-NEXT: ds_write_b8 v0, v6 offset:3
; GFX10-NEXT: ds_write_b8 v0, v2 offset:4 ; GFX10-NEXT: ds_write_b8 v0, v2 offset:4
; GFX10-NEXT: ds_write_b8 v0, v1 offset:5 ; GFX10-NEXT: ds_write_b8 v0, v1 offset:5
; GFX10-NEXT: ds_write_b8 v0, v7 offset:6 ; GFX10-NEXT: ds_write_b8 v0, v4 offset:6
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v4, 24, v3
; GFX10-NEXT: ds_write_b8 v0, v5 offset:7 ; GFX10-NEXT: ds_write_b8 v0, v5 offset:7
; GFX10-NEXT: ds_write_b8 v0, v3 offset:8 ; GFX10-NEXT: ds_write_b8 v0, v3 offset:8
; GFX10-NEXT: ds_write_b8 v0, v1 offset:9 ; GFX10-NEXT: ds_write_b8 v0, v1 offset:9
; GFX10-NEXT: ds_write_b8 v0, v2 offset:10 ; GFX10-NEXT: ds_write_b8 v0, v2 offset:10
; GFX10-NEXT: ds_write_b8 v0, v7 offset:11 ; GFX10-NEXT: ds_write_b8 v0, v4 offset:11
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
store <3 x i32> %x, <3 x i32> addrspace(3)* %out, align 1 store <3 x i32> %x, <3 x i32> addrspace(3)* %out, align 1
ret void ret void
} }

llvm/test/CodeGen/AMDGPU/GlobalISel/lshr.ll

Show First 20 Lines • Show All 1,602 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_lshrrev_b64 v[0:1], v4, v[0:1] ; GFX9-NEXT: v_lshrrev_b64 v[0:1], v4, v[0:1]
; GFX9-NEXT: v_lshrrev_b64 v[2:3], v6, v[2:3] ; GFX9-NEXT: v_lshrrev_b64 v[2:3], v6, v[2:3]
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_lshr_v2i64: ; GFX10-LABEL: v_lshr_v2i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v10, v0 ; GFX10-NEXT: v_lshrrev_b64 v[0:1], v4, v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v11, v1 ; GFX10-NEXT: v_lshrrev_b64 v[2:3], v6, v[2:3]
; GFX10-NEXT: v_mov_b32_e32 v7, v2
; GFX10-NEXT: v_mov_b32_e32 v8, v3
; GFX10-NEXT: v_lshrrev_b64 v[0:1], v4, v[10:11]
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v6, v[7:8]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = lshr <2 x i64> %value, %amount %result = lshr <2 x i64> %value, %amount
ret <2 x i64> %result ret <2 x i64> %result
} }
define <2 x i64> @v_lshr_v2i64_31(<2 x i64> %value) { define <2 x i64> @v_lshr_v2i64_31(<2 x i64> %value) {
; GFX6-LABEL: v_lshr_v2i64_31: ; GFX6-LABEL: v_lshr_v2i64_31:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 45 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/mul.ll

Show First 20 Lines • Show All 579 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_mul_hi_u32 v8, v0, v3 ; GFX10-NEXT: v_mul_hi_u32 v8, v0, v3
; GFX10-NEXT: v_mul_lo_u32 v9, v1, v4 ; GFX10-NEXT: v_mul_lo_u32 v9, v1, v4
; GFX10-NEXT: v_mul_lo_u32 v2, v2, v3 ; GFX10-NEXT: v_mul_lo_u32 v2, v2, v3
; GFX10-NEXT: v_mul_lo_u32 v5, v0, v5 ; GFX10-NEXT: v_mul_lo_u32 v5, v0, v5
; GFX10-NEXT: v_mul_hi_u32 v4, v0, v4 ; GFX10-NEXT: v_mul_hi_u32 v4, v0, v4
; GFX10-NEXT: v_mul_lo_u32 v0, v0, v3 ; GFX10-NEXT: v_mul_lo_u32 v0, v0, v3
; GFX10-NEXT: v_add_co_u32 v6, s4, v6, v7 ; GFX10-NEXT: v_add_co_u32 v6, s4, v6, v7
; GFX10-NEXT: v_mul_hi_u32 v7, v1, v3 ; GFX10-NEXT: v_mul_hi_u32 v7, v1, v3
; GFX10-NEXT: v_cndmask_b32_e64 v11, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v10, 0, 1, s4
; GFX10-NEXT: v_add_nc_u32_e32 v2, v2, v9 ; GFX10-NEXT: v_add_nc_u32_e32 v2, v2, v9
; GFX10-NEXT: v_add_co_u32 v1, s4, v6, v8 ; GFX10-NEXT: v_add_co_u32 v1, s4, v6, v8
; GFX10-NEXT: v_cndmask_b32_e64 v6, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, 0, 1, s4
; GFX10-NEXT: v_add3_u32 v2, v2, v5, v7 ; GFX10-NEXT: v_add3_u32 v2, v2, v5, v7
; GFX10-NEXT: v_add_nc_u32_e32 v3, v11, v6 ; GFX10-NEXT: v_add_nc_u32_e32 v3, v10, v6
; GFX10-NEXT: v_add3_u32 v2, v2, v4, v3 ; GFX10-NEXT: v_add3_u32 v2, v2, v4, v3
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = mul i96 %num, %den %result = mul i96 %num, %den
ret i96 %result ret i96 %result
} }
define amdgpu_ps <4 x i32> @s_mul_i128(i128 inreg %num, i128 inreg %den) { define amdgpu_ps <4 x i32> @s_mul_i128(i128 inreg %num, i128 inreg %den) {
; GFX7-LABEL: s_mul_i128: ; GFX7-LABEL: s_mul_i128:
▲ Show 20 Lines • Show All 390 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_add_co_u32 v9, s5, v10, v11 ; GFX10-NEXT: v_add_co_u32 v9, s5, v10, v11
; GFX10-NEXT: v_mul_hi_u32 v11, v1, v4 ; GFX10-NEXT: v_mul_hi_u32 v11, v1, v4
; GFX10-NEXT: v_cndmask_b32_e64 v14, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v14, 0, 1, s4
; GFX10-NEXT: v_cndmask_b32_e64 v10, 0, 1, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v10, 0, 1, s5
; GFX10-NEXT: v_add_co_u32 v13, s4, v8, v13 ; GFX10-NEXT: v_add_co_u32 v13, s4, v8, v13
; GFX10-NEXT: v_add_co_u32 v8, s5, v9, v12 ; GFX10-NEXT: v_add_co_u32 v8, s5, v9, v12
; GFX10-NEXT: v_cndmask_b32_e64 v12, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v12, 0, 1, s4
; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, s5
; GFX10-NEXT: v_add_co_u32 v18, s4, v13, v11 ; GFX10-NEXT: v_add_co_u32 v11, s4, v13, v11
; GFX10-NEXT: v_cndmask_b32_e64 v13, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v13, 0, 1, s4
; GFX10-NEXT: v_add_nc_u32_e32 v9, v10, v9 ; GFX10-NEXT: v_add_nc_u32_e32 v9, v10, v9
; GFX10-NEXT: v_mul_lo_u32 v10, v2, v5 ; GFX10-NEXT: v_mul_lo_u32 v10, v2, v5
; GFX10-NEXT: v_add_co_u32 v11, s4, v18, v15 ; GFX10-NEXT: v_add_co_u32 v11, s4, v11, v15
; GFX10-NEXT: v_mul_hi_u32 v15, v2, v4 ; GFX10-NEXT: v_mul_hi_u32 v15, v2, v4
; GFX10-NEXT: v_add3_u32 v12, v14, v12, v13 ; GFX10-NEXT: v_add3_u32 v12, v14, v12, v13
; GFX10-NEXT: v_mul_lo_u32 v13, v1, v6 ; GFX10-NEXT: v_mul_lo_u32 v13, v1, v6
; GFX10-NEXT: v_mul_hi_u32 v1, v1, v5 ; GFX10-NEXT: v_mul_hi_u32 v1, v1, v5
; GFX10-NEXT: v_cndmask_b32_e64 v14, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v14, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v2, s4, v11, v9 ; GFX10-NEXT: v_add_co_u32 v2, s4, v11, v9
; GFX10-NEXT: v_add_nc_u32_e32 v10, v3, v10 ; GFX10-NEXT: v_add_nc_u32_e32 v3, v3, v10
; GFX10-NEXT: v_cndmask_b32_e64 v5, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v5, 0, 1, s4
; GFX10-NEXT: v_mul_hi_u32 v6, v0, v6 ; GFX10-NEXT: v_mul_hi_u32 v6, v0, v6
; GFX10-NEXT: v_mul_lo_u32 v0, v0, v4 ; GFX10-NEXT: v_mul_lo_u32 v0, v0, v4
; GFX10-NEXT: v_add3_u32 v10, v10, v13, v7 ; GFX10-NEXT: v_add3_u32 v3, v3, v13, v7
; GFX10-NEXT: v_add3_u32 v4, v12, v14, v5 ; GFX10-NEXT: v_add3_u32 v4, v12, v14, v5
; GFX10-NEXT: v_add3_u32 v1, v10, v15, v1 ; GFX10-NEXT: v_add3_u32 v1, v3, v15, v1
; GFX10-NEXT: v_add3_u32 v3, v1, v6, v4 ; GFX10-NEXT: v_add3_u32 v3, v1, v6, v4
; GFX10-NEXT: v_mov_b32_e32 v1, v8 ; GFX10-NEXT: v_mov_b32_e32 v1, v8
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = mul i128 %num, %den %result = mul i128 %num, %den
ret i128 %result ret i128 %result
} }
define amdgpu_ps <8 x i32> @s_mul_i256(i256 inreg %num, i256 inreg %den) { define amdgpu_ps <8 x i32> @s_mul_i256(i256 inreg %num, i256 inreg %den) {
▲ Show 20 Lines • Show All 1,727 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_mul_lo_u32 v20, v1, v9 ; GFX10-NEXT: v_mul_lo_u32 v20, v1, v9
; GFX10-NEXT: v_mul_hi_u32 v21, v1, v8 ; GFX10-NEXT: v_mul_hi_u32 v21, v1, v8
; GFX10-NEXT: v_mul_lo_u32 v22, v3, v8 ; GFX10-NEXT: v_mul_lo_u32 v22, v3, v8
; GFX10-NEXT: v_mul_lo_u32 v25, v1, v10 ; GFX10-NEXT: v_mul_lo_u32 v25, v1, v10
; GFX10-NEXT: v_mul_hi_u32 v23, v0, v9 ; GFX10-NEXT: v_mul_hi_u32 v23, v0, v9
; GFX10-NEXT: v_add_co_u32 v16, s4, v16, v17 ; GFX10-NEXT: v_add_co_u32 v16, s4, v16, v17
; GFX10-NEXT: v_mul_hi_u32 v27, v0, v10 ; GFX10-NEXT: v_mul_hi_u32 v27, v0, v10
; GFX10-NEXT: v_cndmask_b32_e64 v17, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v17, 0, 1, s4
; GFX10-NEXT: v_mul_lo_u32 v7, v7, v8 ; GFX10-NEXT: v_mul_hi_u32 v29, v3, v9
; GFX10-NEXT: v_mul_lo_u32 v15, v0, v15 ; GFX10-NEXT: v_mul_hi_u32 v31, v4, v9
; GFX10-NEXT: v_add_co_u32 v16, s4, v16, v18 ; GFX10-NEXT: v_add_co_u32 v16, s4, v16, v18
; GFX10-NEXT: v_mul_lo_u32 v7, v7, v8
; GFX10-NEXT: v_cndmask_b32_e64 v18, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v18, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v19, s4, v19, v20 ; GFX10-NEXT: v_add_co_u32 v19, s4, v19, v20
; GFX10-NEXT: v_mul_lo_u32 v20, v2, v9 ; GFX10-NEXT: v_mul_lo_u32 v20, v2, v9
; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4
; GFX10-NEXT: v_mul_lo_u32 v15, v0, v15
; GFX10-NEXT: v_add_nc_u32_e32 v17, v17, v18 ; GFX10-NEXT: v_add_nc_u32_e32 v17, v17, v18
; GFX10-NEXT: v_mul_lo_u32 v18, v0, v10 ; GFX10-NEXT: v_mul_lo_u32 v18, v0, v10
; GFX10-NEXT: v_add_co_u32 v18, s4, v19, v18 ; GFX10-NEXT: v_add_co_u32 v18, s4, v19, v18
; GFX10-NEXT: v_cndmask_b32_e64 v19, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v19, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v20, s4, v22, v20 ; GFX10-NEXT: v_add_co_u32 v20, s4, v22, v20
; GFX10-NEXT: v_mul_lo_u32 v22, v0, v11 ; GFX10-NEXT: v_mul_lo_u32 v22, v0, v11
; GFX10-NEXT: v_add_co_u32 v18, s5, v18, v21 ; GFX10-NEXT: v_add_co_u32 v18, s5, v18, v21
; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, 1, s4
; GFX10-NEXT: v_cndmask_b32_e64 v21, 0, 1, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v21, 0, 1, s5
; GFX10-NEXT: v_add_co_u32 v20, s4, v20, v25 ; GFX10-NEXT: v_add_co_u32 v20, s4, v20, v25
; GFX10-NEXT: v_add_co_u32 v18, s5, v18, v23 ; GFX10-NEXT: v_add_co_u32 v18, s5, v18, v23
; GFX10-NEXT: v_mul_hi_u32 v23, v1, v9 ; GFX10-NEXT: v_mul_hi_u32 v23, v1, v9
; GFX10-NEXT: v_add3_u32 v19, v24, v19, v21 ; GFX10-NEXT: v_add3_u32 v19, v24, v19, v21
; GFX10-NEXT: v_mul_hi_u32 v21, v2, v8 ; GFX10-NEXT: v_mul_hi_u32 v21, v2, v8
; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v20, s4, v20, v22 ; GFX10-NEXT: v_add_co_u32 v20, s4, v20, v22
; GFX10-NEXT: v_cndmask_b32_e64 v29, 0, 1, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v22, 0, 1, s5
; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v17, s5, v18, v17 ; GFX10-NEXT: v_add_co_u32 v17, s5, v18, v17
; GFX10-NEXT: v_cndmask_b32_e64 v18, 0, 1, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v18, 0, 1, s5
; GFX10-NEXT: v_add_co_u32 v20, s4, v20, v21 ; GFX10-NEXT: v_add_co_u32 v20, s4, v20, v21
; GFX10-NEXT: v_add3_u32 v21, v26, v24, v25 ; GFX10-NEXT: v_add3_u32 v21, v26, v24, v25
; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4
; GFX10-NEXT: v_mul_lo_u32 v25, v4, v8 ; GFX10-NEXT: v_mul_lo_u32 v25, v4, v8
; GFX10-NEXT: v_mul_lo_u32 v26, v3, v9 ; GFX10-NEXT: v_mul_lo_u32 v26, v3, v9
; GFX10-NEXT: v_add_co_u32 v20, s4, v20, v23 ; GFX10-NEXT: v_add_co_u32 v20, s4, v20, v23
; GFX10-NEXT: v_add3_u32 v18, v19, v29, v18 ; GFX10-NEXT: v_add3_u32 v18, v19, v22, v18
; GFX10-NEXT: v_cndmask_b32_e64 v23, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v23, 0, 1, s4
; GFX10-NEXT: v_mul_hi_u32 v29, v3, v9
; GFX10-NEXT: v_add_co_u32 v20, s5, v20, v27 ; GFX10-NEXT: v_add_co_u32 v20, s5, v20, v27
; GFX10-NEXT: v_add3_u32 v30, v21, v24, v23 ; GFX10-NEXT: v_add3_u32 v19, v21, v24, v23
; GFX10-NEXT: v_mul_lo_u32 v21, v2, v10 ; GFX10-NEXT: v_mul_lo_u32 v21, v2, v10
; GFX10-NEXT: v_add_co_u32 v22, s4, v25, v26 ; GFX10-NEXT: v_add_co_u32 v22, s4, v25, v26
; GFX10-NEXT: v_mul_lo_u32 v24, v1, v11 ; GFX10-NEXT: v_mul_lo_u32 v24, v1, v11
; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4
; GFX10-NEXT: v_mul_hi_u32 v26, v3, v8 ; GFX10-NEXT: v_mul_hi_u32 v26, v3, v8
; GFX10-NEXT: v_cndmask_b32_e64 v23, 0, 1, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v23, 0, 1, s5
; GFX10-NEXT: v_add_co_u32 v21, s4, v22, v21 ; GFX10-NEXT: v_add_co_u32 v21, s4, v22, v21
; GFX10-NEXT: v_mul_lo_u32 v22, v0, v12 ; GFX10-NEXT: v_mul_lo_u32 v22, v0, v12
; GFX10-NEXT: v_cndmask_b32_e64 v27, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v27, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v21, s4, v21, v24 ; GFX10-NEXT: v_add_co_u32 v21, s4, v21, v24
; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v18, s4, v20, v18 ; GFX10-NEXT: v_add_co_u32 v18, s4, v20, v18
; GFX10-NEXT: v_cndmask_b32_e64 v20, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v20, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v21, s4, v21, v22 ; GFX10-NEXT: v_add_co_u32 v21, s4, v21, v22
; GFX10-NEXT: v_mul_hi_u32 v22, v2, v9 ; GFX10-NEXT: v_mul_hi_u32 v22, v2, v9
; GFX10-NEXT: v_add3_u32 v24, v25, v27, v24 ; GFX10-NEXT: v_add3_u32 v24, v25, v27, v24
; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4
; GFX10-NEXT: v_add3_u32 v19, v30, v23, v20 ; GFX10-NEXT: v_add3_u32 v19, v19, v23, v20
; GFX10-NEXT: v_add_co_u32 v21, s4, v21, v26 ; GFX10-NEXT: v_add_co_u32 v21, s4, v21, v26
; GFX10-NEXT: v_mul_hi_u32 v20, v1, v10 ; GFX10-NEXT: v_mul_hi_u32 v20, v1, v10
; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, 1, s4
; GFX10-NEXT: v_mul_hi_u32 v27, v0, v11 ; GFX10-NEXT: v_mul_hi_u32 v27, v0, v11
; GFX10-NEXT: v_add_co_u32 v21, s4, v21, v22 ; GFX10-NEXT: v_add_co_u32 v21, s4, v21, v22
; GFX10-NEXT: v_mul_lo_u32 v22, v5, v8 ; GFX10-NEXT: v_mul_lo_u32 v22, v5, v8
; GFX10-NEXT: v_add3_u32 v23, v24, v25, v26 ; GFX10-NEXT: v_add3_u32 v23, v24, v25, v26
; GFX10-NEXT: v_mul_lo_u32 v24, v4, v9 ; GFX10-NEXT: v_mul_lo_u32 v24, v4, v9
; GFX10-NEXT: v_cndmask_b32_e64 v30, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v20, s4, v21, v20 ; GFX10-NEXT: v_add_co_u32 v20, s4, v21, v20
; GFX10-NEXT: v_mul_lo_u32 v26, v3, v10 ; GFX10-NEXT: v_mul_lo_u32 v26, v3, v10
; GFX10-NEXT: v_cndmask_b32_e64 v21, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v21, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v20, s5, v20, v27 ; GFX10-NEXT: v_add_co_u32 v20, s5, v20, v27
; GFX10-NEXT: v_add_co_u32 v31, s4, v22, v24 ; GFX10-NEXT: v_add_co_u32 v22, s4, v22, v24
; GFX10-NEXT: v_add3_u32 v35, v23, v30, v21 ; GFX10-NEXT: v_add3_u32 v21, v23, v25, v21
; GFX10-NEXT: v_mul_lo_u32 v23, v2, v11 ; GFX10-NEXT: v_mul_lo_u32 v23, v2, v11
; GFX10-NEXT: v_cndmask_b32_e64 v34, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4
; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s5
; GFX10-NEXT: v_add_co_u32 v22, s4, v31, v26 ; GFX10-NEXT: v_add_co_u32 v22, s4, v22, v26
; GFX10-NEXT: v_mul_lo_u32 v26, v1, v12 ; GFX10-NEXT: v_mul_lo_u32 v26, v1, v12
; GFX10-NEXT: v_cndmask_b32_e64 v27, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v27, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v19, s5, v20, v19 ; GFX10-NEXT: v_add_co_u32 v19, s5, v20, v19
; GFX10-NEXT: v_add_co_u32 v31, s4, v22, v23 ; GFX10-NEXT: v_add_co_u32 v22, s4, v22, v23
; GFX10-NEXT: v_mul_lo_u32 v23, v0, v13 ; GFX10-NEXT: v_mul_lo_u32 v23, v0, v13
; GFX10-NEXT: v_cndmask_b32_e64 v28, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v28, 0, 1, s4
; GFX10-NEXT: v_cndmask_b32_e64 v20, 0, 1, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v20, 0, 1, s5
; GFX10-NEXT: v_add_co_u32 v22, s4, v31, v26 ; GFX10-NEXT: v_add_co_u32 v22, s4, v22, v26
; GFX10-NEXT: v_mul_hi_u32 v26, v4, v8 ; GFX10-NEXT: v_mul_hi_u32 v26, v4, v8
; GFX10-NEXT: v_cndmask_b32_e64 v30, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v30, 0, 1, s4
; GFX10-NEXT: v_add3_u32 v20, v35, v25, v20 ; GFX10-NEXT: v_add3_u32 v20, v21, v25, v20
; GFX10-NEXT: v_add_co_u32 v31, s4, v22, v23 ; GFX10-NEXT: v_add_co_u32 v22, s4, v22, v23
; GFX10-NEXT: v_add3_u32 v23, v34, v27, v28 ; GFX10-NEXT: v_add3_u32 v23, v24, v27, v28
; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4
; GFX10-NEXT: v_mul_hi_u32 v22, v2, v10 ; GFX10-NEXT: v_mul_lo_u32 v27, v6, v8
; GFX10-NEXT: v_mul_lo_u32 v28, v5, v9 ; GFX10-NEXT: v_mul_lo_u32 v28, v5, v9
; GFX10-NEXT: v_add_co_u32 v27, s4, v31, v26 ; GFX10-NEXT: v_add_co_u32 v21, s4, v22, v26
; GFX10-NEXT: v_mul_hi_u32 v26, v1, v11 ; GFX10-NEXT: v_mul_hi_u32 v22, v2, v10
; GFX10-NEXT: v_add3_u32 v23, v23, v30, v24 ; GFX10-NEXT: v_add3_u32 v23, v23, v30, v24
; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v21, s4, v27, v29 ; GFX10-NEXT: v_mul_hi_u32 v26, v1, v11
; GFX10-NEXT: v_mul_lo_u32 v27, v6, v8 ; GFX10-NEXT: v_add_co_u32 v21, s4, v21, v29
; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4
; GFX10-NEXT: v_mul_hi_u32 v29, v0, v12 ; GFX10-NEXT: v_mul_hi_u32 v29, v0, v12
; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v21, s4, v21, v22 ; GFX10-NEXT: v_add_co_u32 v21, s4, v21, v22
; GFX10-NEXT: v_add3_u32 v23, v23, v24, v25 ; GFX10-NEXT: v_add3_u32 v23, v23, v24, v25
; GFX10-NEXT: v_mul_lo_u32 v24, v4, v10 ; GFX10-NEXT: v_mul_lo_u32 v24, v4, v10
; GFX10-NEXT: v_cndmask_b32_e64 v33, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v22, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v25, s4, v27, v28 ; GFX10-NEXT: v_add_co_u32 v25, s4, v27, v28
; GFX10-NEXT: v_add_co_u32 v31, s5, v21, v26 ; GFX10-NEXT: v_add_co_u32 v21, s5, v21, v26
; GFX10-NEXT: v_mul_lo_u32 v27, v3, v11 ; GFX10-NEXT: v_mul_lo_u32 v27, v3, v11
; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, 1, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, 1, s5
; GFX10-NEXT: v_cndmask_b32_e64 v28, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v28, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v24, s4, v25, v24 ; GFX10-NEXT: v_add_co_u32 v24, s4, v25, v24
; GFX10-NEXT: v_add_co_u32 v21, s5, v31, v29 ; GFX10-NEXT: v_add_co_u32 v21, s5, v21, v29
; GFX10-NEXT: v_add3_u32 v39, v23, v33, v26 ; GFX10-NEXT: v_add3_u32 v22, v23, v22, v26
; GFX10-NEXT: v_mul_lo_u32 v23, v2, v12 ; GFX10-NEXT: v_mul_lo_u32 v23, v2, v12
; GFX10-NEXT: v_cndmask_b32_e64 v35, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v24, s4, v24, v27 ; GFX10-NEXT: v_add_co_u32 v24, s4, v24, v27
; GFX10-NEXT: v_mul_lo_u32 v27, v1, v13 ; GFX10-NEXT: v_mul_lo_u32 v27, v1, v13
; GFX10-NEXT: v_cndmask_b32_e64 v29, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v29, 0, 1, s4
; GFX10-NEXT: v_mul_hi_u32 v31, v4, v9
; GFX10-NEXT: v_mul_hi_u32 v25, v3, v10
; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, 1, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, 1, s5
; GFX10-NEXT: v_add_co_u32 v34, s4, v24, v23 ; GFX10-NEXT: v_add_co_u32 v20, s5, v21, v20
; GFX10-NEXT: v_add_co_u32 v23, s4, v24, v23
; GFX10-NEXT: v_mul_lo_u32 v24, v0, v14 ; GFX10-NEXT: v_mul_lo_u32 v24, v0, v14
; GFX10-NEXT: v_cndmask_b32_e64 v30, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v30, 0, 1, s4
; GFX10-NEXT: v_add3_u32 v35, v28, v35, v29 ; GFX10-NEXT: v_cndmask_b32_e64 v21, 0, 1, s5
; GFX10-NEXT: v_add_co_u32 v20, s5, v21, v20 ; GFX10-NEXT: v_add_co_u32 v23, s4, v23, v27
; GFX10-NEXT: v_add_co_u32 v23, s4, v34, v27
; GFX10-NEXT: v_mul_hi_u32 v27, v5, v8 ; GFX10-NEXT: v_mul_hi_u32 v27, v5, v8
; GFX10-NEXT: v_cndmask_b32_e64 v32, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v32, 0, 1, s4
; GFX10-NEXT: v_cndmask_b32_e64 v21, 0, 1, s5 ; GFX10-NEXT: v_add3_u32 v21, v22, v26, v21
; GFX10-NEXT: v_mul_lo_u32 v29, v3, v12
; GFX10-NEXT: v_add_co_u32 v34, s4, v23, v24
; GFX10-NEXT: v_mul_hi_u32 v3, v3, v11
; GFX10-NEXT: v_cndmask_b32_e64 v28, 0, 1, s4
; GFX10-NEXT: v_add3_u32 v22, v35, v30, v32
; GFX10-NEXT: v_add3_u32 v21, v39, v26, v21
; GFX10-NEXT: v_add_co_u32 v34, s4, v34, v27
; GFX10-NEXT: v_mul_hi_u32 v26, v2, v11 ; GFX10-NEXT: v_mul_hi_u32 v26, v2, v11
; GFX10-NEXT: v_add_co_u32 v23, s4, v23, v24
; GFX10-NEXT: v_add3_u32 v24, v28, v25, v29
; GFX10-NEXT: v_cndmask_b32_e64 v28, 0, 1, s4
; GFX10-NEXT: v_mul_hi_u32 v25, v3, v10
; GFX10-NEXT: v_mul_lo_u32 v29, v3, v12
; GFX10-NEXT: v_add_co_u32 v23, s4, v23, v27
; GFX10-NEXT: v_add3_u32 v24, v24, v30, v32
; GFX10-NEXT: v_cndmask_b32_e64 v27, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v27, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v23, s4, v34, v31 ; GFX10-NEXT: v_mul_hi_u32 v3, v3, v11
; GFX10-NEXT: v_add_co_u32 v23, s4, v23, v31
; GFX10-NEXT: v_add3_u32 v22, v24, v28, v27
; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v24, 0, 1, s4
; GFX10-NEXT: v_add3_u32 v22, v22, v28, v27
; GFX10-NEXT: v_mul_lo_u32 v28, v6, v9 ; GFX10-NEXT: v_mul_lo_u32 v28, v6, v9
; GFX10-NEXT: v_add_co_u32 v23, s4, v23, v25 ; GFX10-NEXT: v_add_co_u32 v23, s4, v23, v25
; GFX10-NEXT: v_mul_hi_u32 v27, v1, v12 ; GFX10-NEXT: v_mul_hi_u32 v27, v1, v12
; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v25, 0, 1, s4
; GFX10-NEXT: v_mul_hi_u32 v6, v6, v8 ; GFX10-NEXT: v_mul_hi_u32 v6, v6, v8
; GFX10-NEXT: v_add_co_u32 v30, s4, v23, v26 ; GFX10-NEXT: v_add_co_u32 v23, s4, v23, v26
; GFX10-NEXT: v_add3_u32 v33, v22, v24, v25 ; GFX10-NEXT: v_add3_u32 v22, v22, v24, v25
; GFX10-NEXT: v_mul_lo_u32 v24, v5, v10 ; GFX10-NEXT: v_mul_lo_u32 v24, v5, v10
; GFX10-NEXT: v_mul_lo_u32 v25, v4, v11 ; GFX10-NEXT: v_mul_lo_u32 v25, v4, v11
; GFX10-NEXT: v_add_nc_u32_e32 v7, v7, v28 ; GFX10-NEXT: v_add_nc_u32_e32 v7, v7, v28
; GFX10-NEXT: v_mul_lo_u32 v28, v2, v13 ; GFX10-NEXT: v_mul_lo_u32 v28, v2, v13
; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v23, s4, v30, v27
; GFX10-NEXT: v_mul_hi_u32 v5, v5, v9 ; GFX10-NEXT: v_mul_hi_u32 v5, v5, v9
; GFX10-NEXT: v_cndmask_b32_e64 v27, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, 1, s4
; GFX10-NEXT: v_add_co_u32 v23, s4, v23, v27
; GFX10-NEXT: v_mul_hi_u32 v4, v4, v10 ; GFX10-NEXT: v_mul_hi_u32 v4, v4, v10
; GFX10-NEXT: v_cndmask_b32_e64 v27, 0, 1, s4
; GFX10-NEXT: v_add3_u32 v7, v7, v24, v25 ; GFX10-NEXT: v_add3_u32 v7, v7, v24, v25
; GFX10-NEXT: v_mul_lo_u32 v24, v1, v14 ; GFX10-NEXT: v_mul_lo_u32 v24, v1, v14
; GFX10-NEXT: v_mul_hi_u32 v25, v0, v13 ; GFX10-NEXT: v_mul_hi_u32 v25, v0, v13
; GFX10-NEXT: v_add3_u32 v33, v33, v26, v27
; GFX10-NEXT: v_mul_hi_u32 v2, v2, v12 ; GFX10-NEXT: v_mul_hi_u32 v2, v2, v12
; GFX10-NEXT: v_add3_u32 v26, v7, v29, v28
; GFX10-NEXT: v_mul_hi_u32 v1, v1, v13 ; GFX10-NEXT: v_mul_hi_u32 v1, v1, v13
; GFX10-NEXT: v_add3_u32 v7, v26, v24, v15 ; GFX10-NEXT: v_add3_u32 v7, v7, v29, v28
; GFX10-NEXT: v_add_co_u32 v11, s4, v23, v25 ; GFX10-NEXT: v_add3_u32 v22, v22, v26, v27
; GFX10-NEXT: v_add3_u32 v7, v7, v24, v15
; GFX10-NEXT: v_add_co_u32 v9, s4, v23, v25
; GFX10-NEXT: v_cndmask_b32_e64 v10, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v10, 0, 1, s4
; GFX10-NEXT: v_add3_u32 v5, v7, v6, v5 ; GFX10-NEXT: v_add3_u32 v5, v7, v6, v5
; GFX10-NEXT: v_add_co_u32 v6, s4, v11, v21 ; GFX10-NEXT: v_add_co_u32 v6, s4, v9, v21
; GFX10-NEXT: v_cndmask_b32_e64 v7, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v7, 0, 1, s4
; GFX10-NEXT: v_add3_u32 v3, v5, v4, v3 ; GFX10-NEXT: v_add3_u32 v3, v5, v4, v3
; GFX10-NEXT: v_mul_hi_u32 v4, v0, v14 ; GFX10-NEXT: v_mul_hi_u32 v4, v0, v14
; GFX10-NEXT: v_mul_lo_u32 v0, v0, v8 ; GFX10-NEXT: v_mul_lo_u32 v0, v0, v8
; GFX10-NEXT: v_add3_u32 v5, v33, v10, v7 ; GFX10-NEXT: v_add3_u32 v5, v22, v10, v7
; GFX10-NEXT: v_add3_u32 v3, v3, v2, v1 ; GFX10-NEXT: v_add3_u32 v1, v3, v2, v1
; GFX10-NEXT: v_mov_b32_e32 v1, v16
; GFX10-NEXT: v_mov_b32_e32 v2, v17 ; GFX10-NEXT: v_mov_b32_e32 v2, v17
; GFX10-NEXT: v_add3_u32 v7, v3, v4, v5
; GFX10-NEXT: v_mov_b32_e32 v3, v18 ; GFX10-NEXT: v_mov_b32_e32 v3, v18
; GFX10-NEXT: v_add3_u32 v7, v1, v4, v5
; GFX10-NEXT: v_mov_b32_e32 v1, v16
; GFX10-NEXT: v_mov_b32_e32 v4, v19 ; GFX10-NEXT: v_mov_b32_e32 v4, v19
; GFX10-NEXT: v_mov_b32_e32 v5, v20 ; GFX10-NEXT: v_mov_b32_e32 v5, v20
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = mul i256 %num, %den %result = mul i256 %num, %den
ret i256 %result ret i256 %result
} }

llvm/test/CodeGen/AMDGPU/GlobalISel/roundeven.ll

Show First 20 Lines • Show All 407 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_and_or_b32 v0, v2, v4, v0 ; GFX9-NEXT: v_and_or_b32 v0, v2, v4, v0
; GFX9-NEXT: v_and_or_b32 v1, v3, v4, v1 ; GFX9-NEXT: v_and_or_b32 v1, v3, v4, v1
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_roundeven_v4f16: ; GFX10-LABEL: v_roundeven_v4f16:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_rndne_f16_e32 v3, v1
; GFX10-NEXT: v_rndne_f16_e32 v2, v0 ; GFX10-NEXT: v_rndne_f16_e32 v2, v0
; GFX10-NEXT: v_rndne_f16_sdwa v7, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 ; GFX10-NEXT: v_rndne_f16_e32 v3, v1
; GFX10-NEXT: v_rndne_f16_sdwa v0, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1
; GFX10-NEXT: v_mov_b32_e32 v4, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v4, 0xffff
; GFX10-NEXT: v_rndne_f16_sdwa v1, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 ; GFX10-NEXT: v_rndne_f16_sdwa v1, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1
; GFX10-NEXT: v_and_or_b32 v0, v2, v4, v7 ; GFX10-NEXT: v_and_or_b32 v0, v2, v4, v0
; GFX10-NEXT: v_and_or_b32 v1, v3, v4, v1 ; GFX10-NEXT: v_and_or_b32 v1, v3, v4, v1
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%roundeven = call <4 x half> @llvm.roundeven.v4f16(<4 x half> %x) %roundeven = call <4 x half> @llvm.roundeven.v4f16(<4 x half> %x)
ret <4 x half> %roundeven ret <4 x half> %roundeven
} }
define float @v_roundeven_f32_fabs(float %x) { define float @v_roundeven_f32_fabs(float %x) {
▲ Show 20 Lines • Show All 264 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/saddsat.ll

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 4,193 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v2, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_saddsat_i64: ; GFX10-LABEL: v_saddsat_i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_add_co_u32 v10, vcc_lo, v0, v2 ; GFX10-NEXT: v_add_co_u32 v4, vcc_lo, v0, v2
; GFX10-NEXT: v_add_co_ci_u32_e32 v11, vcc_lo, v1, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, 0, v[2:3] ; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, 0, v[2:3]
; GFX10-NEXT: v_ashrrev_i32_e32 v6, 31, v11 ; GFX10-NEXT: v_ashrrev_i32_e32 v6, 31, v5
; GFX10-NEXT: v_cmp_lt_i64_e64 s4, v[10:11], v[0:1] ; GFX10-NEXT: v_cmp_lt_i64_e64 s4, v[4:5], v[0:1]
; GFX10-NEXT: v_add_co_u32 v0, s5, v6, 0 ; GFX10-NEXT: v_add_co_u32 v0, s5, v6, 0
; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s5, 0x80000000, v6, s5 ; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s5, 0x80000000, v6, s5
; GFX10-NEXT: s_xor_b32 vcc_lo, vcc_lo, s4 ; GFX10-NEXT: s_xor_b32 vcc_lo, vcc_lo, s4
; GFX10-NEXT: v_cndmask_b32_e32 v0, v10, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v11, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i64 @llvm.sadd.sat.i64(i64 %lhs, i64 %rhs) %result = call i64 @llvm.sadd.sat.i64(i64 %lhs, i64 %rhs)
ret i64 %result ret i64 %result
} }
define amdgpu_ps i64 @s_saddsat_i64(i64 inreg %lhs, i64 inreg %rhs) { define amdgpu_ps i64 @s_saddsat_i64(i64 inreg %lhs, i64 inreg %rhs) {
; GFX6-LABEL: s_saddsat_i64: ; GFX6-LABEL: s_saddsat_i64:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 318 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e32 v2, v4, v3, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v2, v4, v3, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, v5, v6, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v3, v5, v6, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_saddsat_v2i64: ; GFX10-LABEL: v_saddsat_v2i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v14, v0 ; GFX10-NEXT: v_add_co_u32 v8, vcc_lo, v0, v4
; GFX10-NEXT: v_mov_b32_e32 v15, v1
; GFX10-NEXT: v_mov_b32_e32 v17, v2
; GFX10-NEXT: v_mov_b32_e32 v18, v3
; GFX10-NEXT: v_cmp_gt_i64_e64 s4, 0, v[4:5] ; GFX10-NEXT: v_cmp_gt_i64_e64 s4, 0, v[4:5]
; GFX10-NEXT: v_add_co_u32 v8, vcc_lo, v14, v4 ; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, v1, v5, vcc_lo
; GFX10-NEXT: v_add_co_u32 v10, vcc_lo, v2, v6
; GFX10-NEXT: v_cmp_gt_i64_e64 s6, 0, v[6:7] ; GFX10-NEXT: v_cmp_gt_i64_e64 s6, 0, v[6:7]
; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, v15, v5, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v11, vcc_lo, v3, v7, vcc_lo
; GFX10-NEXT: v_add_co_u32 v19, vcc_lo, v17, v6
; GFX10-NEXT: v_add_co_ci_u32_e32 v20, vcc_lo, v18, v7, vcc_lo
; GFX10-NEXT: v_ashrrev_i32_e32 v12, 31, v9 ; GFX10-NEXT: v_ashrrev_i32_e32 v12, 31, v9
; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[8:9], v[14:15] ; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[8:9], v[0:1]
; GFX10-NEXT: v_ashrrev_i32_e32 v0, 31, v20 ; GFX10-NEXT: v_ashrrev_i32_e32 v0, 31, v11
; GFX10-NEXT: v_add_co_u32 v1, s5, v12, 0 ; GFX10-NEXT: v_add_co_u32 v1, s5, v12, 0
; GFX10-NEXT: v_add_co_ci_u32_e64 v4, s5, 0x80000000, v12, s5 ; GFX10-NEXT: v_add_co_ci_u32_e64 v4, s5, 0x80000000, v12, s5
; GFX10-NEXT: v_cmp_lt_i64_e64 s5, v[19:20], v[17:18] ; GFX10-NEXT: v_cmp_lt_i64_e64 s5, v[10:11], v[2:3]
; GFX10-NEXT: v_add_co_u32 v2, s7, v0, 0 ; GFX10-NEXT: v_add_co_u32 v2, s7, v0, 0
; GFX10-NEXT: s_xor_b32 vcc_lo, s4, vcc_lo ; GFX10-NEXT: s_xor_b32 vcc_lo, s4, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e64 v3, s7, 0x80000000, v0, s7 ; GFX10-NEXT: v_add_co_ci_u32_e64 v3, s7, 0x80000000, v0, s7
; GFX10-NEXT: v_cndmask_b32_e32 v0, v8, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v8, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v9, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v9, v4, vcc_lo
; GFX10-NEXT: s_xor_b32 vcc_lo, s6, s5 ; GFX10-NEXT: s_xor_b32 vcc_lo, s6, s5
; GFX10-NEXT: v_cndmask_b32_e32 v2, v19, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v10, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v20, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v11, v3, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i64> @llvm.sadd.sat.v2i64(<2 x i64> %lhs, <2 x i64> %rhs) %result = call <2 x i64> @llvm.sadd.sat.v2i64(<2 x i64> %lhs, <2 x i64> %rhs)
ret <2 x i64> %result ret <2 x i64> %result
} }
define amdgpu_ps <2 x i64> @s_saddsat_v2i64(<2 x i64> inreg %lhs, <2 x i64> inreg %rhs) { define amdgpu_ps <2 x i64> @s_saddsat_v2i64(<2 x i64> inreg %lhs, <2 x i64> inreg %rhs) {
; GFX6-LABEL: s_saddsat_v2i64: ; GFX6-LABEL: s_saddsat_v2i64:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 744 Lines • ▼ Show 20 Lines
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: v_add_co_u32 v4, vcc_lo, s0, v0 ; GFX10-NEXT: v_add_co_u32 v4, vcc_lo, s0, v0
; GFX10-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, s1, v1, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, s1, v1, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, s2, v2, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, s2, v2, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, s3, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, s3, v3, vcc_lo
; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[0:1], v[4:5] ; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[0:1], v[4:5]
; GFX10-NEXT: s_movk_i32 s0, 0x7f ; GFX10-NEXT: s_movk_i32 s0, 0x7f
; GFX10-NEXT: s_sub_i32 s1, 64, s0 ; GFX10-NEXT: s_sub_i32 s1, 64, s0
; GFX10-NEXT: v_lshrrev_b64 v[15:16], s0, v[4:5]
; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, s[2:3], v[6:7] ; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, s[2:3], v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, s[2:3], v[6:7] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, s[2:3], v[6:7]
; GFX10-NEXT: v_cndmask_b32_e32 v10, v9, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, v9, v8, vcc_lo
; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, 0, v[0:1] ; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, 0, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[8:9], s1, v[6:7] ; GFX10-NEXT: v_lshlrev_b64 v[8:9], s1, v[6:7]
; GFX10-NEXT: v_lshrrev_b64 v[0:1], s0, v[4:5]
; GFX10-NEXT: s_sub_i32 s1, s0, 64 ; GFX10-NEXT: s_sub_i32 s1, s0, 64
; GFX10-NEXT: s_cmp_lt_u32 s0, 64 ; GFX10-NEXT: s_cmp_lt_u32 s0, 64
; GFX10-NEXT: v_ashrrev_i64 v[0:1], s0, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v11, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v11, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, 0, v[2:3] ; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, 0, v[2:3]
; GFX10-NEXT: v_or_b32_e32 v8, v15, v8 ; GFX10-NEXT: v_or_b32_e32 v8, v0, v8
; GFX10-NEXT: v_or_b32_e32 v9, v16, v9 ; GFX10-NEXT: v_or_b32_e32 v9, v1, v9
; GFX10-NEXT: v_ashrrev_i32_e32 v15, 31, v7 ; GFX10-NEXT: v_ashrrev_i64 v[0:1], s0, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v12, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v12, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, 0, v[2:3] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, 0, v[2:3]
; GFX10-NEXT: v_ashrrev_i64 v[2:3], s1, v[6:7] ; GFX10-NEXT: v_ashrrev_i64 v[2:3], s1, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e32 v11, v12, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v11, v12, v11, vcc_lo
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s0, 0 ; GFX10-NEXT: s_cmp_eq_u32 s0, 0
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v9, vcc_lo
; GFX10-NEXT: s_cselect_b32 s1, 1, 0
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v8, vcc_lo
; GFX10-NEXT: s_cselect_b32 s1, 1, 0
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v9, vcc_lo
; GFX10-NEXT: s_and_b32 s0, 1, s1 ; GFX10-NEXT: s_and_b32 s0, 1, s1
; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX10-NEXT: v_ashrrev_i32_e32 v8, 31, v7
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1 ; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1
; GFX10-NEXT: v_xor_b32_e32 v9, v11, v10 ; GFX10-NEXT: v_xor_b32_e32 v9, v11, v10
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v15, v0, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v8, v0, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v8, v1, s0
; GFX10-NEXT: v_and_b32_e32 v8, 1, v9 ; GFX10-NEXT: v_and_b32_e32 v8, 1, v9
; GFX10-NEXT: v_cndmask_b32_e64 v1, v15, v1, s0
; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0 ; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0
; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, v8 ; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, v8
; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, 0, v0, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, 0, v0, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0x80000000, v1, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0x80000000, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v3, s0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v8, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v8, s0
▲ Show 20 Lines • Show All 191 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v2, v6, v2, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v2, v6, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, v7, v3, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v3, v7, v3, vcc
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: saddsat_i128_vs: ; GFX10-LABEL: saddsat_i128_vs:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; GFX10-NEXT: v_add_co_u32 v4, vcc_lo, v0, s0
; GFX10-NEXT: v_mov_b32_e32 v6, v1
; GFX10-NEXT: v_mov_b32_e32 v9, v2
; GFX10-NEXT: v_mov_b32_e32 v10, v3
; GFX10-NEXT: s_cmp_eq_u64 s[2:3], 0
; GFX10-NEXT: v_add_co_u32 v15, vcc_lo, v5, s0
; GFX10-NEXT: v_cmp_lt_u64_e64 s0, s[0:1], 0 ; GFX10-NEXT: v_cmp_lt_u64_e64 s0, s[0:1], 0
; GFX10-NEXT: v_add_co_ci_u32_e32 v16, vcc_lo, s1, v6, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, s1, v1, vcc_lo
; GFX10-NEXT: s_cmp_eq_u64 s[2:3], 0
; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, s2, v2, vcc_lo
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: v_add_co_ci_u32_e32 v19, vcc_lo, s2, v9, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, s3, v3, vcc_lo
; GFX10-NEXT: s_and_b32 s1, 1, s4 ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[4:5], v[0:1]
; GFX10-NEXT: v_add_co_ci_u32_e32 v20, vcc_lo, s3, v10, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[15:16], v[5:6]
; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, s0
; GFX10-NEXT: v_cmp_lt_i64_e64 s0, s[2:3], 0 ; GFX10-NEXT: v_cmp_lt_i64_e64 s0, s[2:3], 0
; GFX10-NEXT: v_ashrrev_i32_e32 v7, 31, v20 ; GFX10-NEXT: s_and_b32 s1, 1, s4
; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[19:20], v[9:10] ; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[6:7], v[2:3]
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[19:20], v[9:10]
; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, s0
; GFX10-NEXT: s_movk_i32 s0, 0x7f ; GFX10-NEXT: s_movk_i32 s0, 0x7f
; GFX10-NEXT: s_sub_i32 s2, 64, s0 ; GFX10-NEXT: s_sub_i32 s2, 64, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[6:7], v[2:3]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], s2, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e32 v10, v1, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, v1, v0, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[0:1], s0, v[4:5]
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s1 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s1
; GFX10-NEXT: v_lshrrev_b64 v[0:1], s0, v[15:16]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], s2, v[19:20]
; GFX10-NEXT: s_sub_i32 s1, s0, 64 ; GFX10-NEXT: s_sub_i32 s1, s0, 64
; GFX10-NEXT: s_cmp_lt_u32 s0, 64 ; GFX10-NEXT: s_cmp_lt_u32 s0, 64
; GFX10-NEXT: v_cndmask_b32_e32 v11, v9, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v11, v9, v8, vcc_lo
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: v_ashrrev_i64 v[8:9], s1, v[19:20] ; GFX10-NEXT: v_ashrrev_i64 v[8:9], s1, v[6:7]
; GFX10-NEXT: s_cmp_eq_u32 s0, 0 ; GFX10-NEXT: s_cmp_eq_u32 s0, 0
; GFX10-NEXT: v_or_b32_e32 v2, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v2, v0, v2
; GFX10-NEXT: v_or_b32_e32 v3, v1, v3 ; GFX10-NEXT: v_or_b32_e32 v3, v1, v3
; GFX10-NEXT: s_cselect_b32 s1, 1, 0 ; GFX10-NEXT: s_cselect_b32 s1, 1, 0
; GFX10-NEXT: v_ashrrev_i64 v[0:1], s0, v[19:20] ; GFX10-NEXT: v_ashrrev_i64 v[0:1], s0, v[6:7]
; GFX10-NEXT: s_and_b32 s0, 1, s1 ; GFX10-NEXT: s_and_b32 s0, 1, s1
; GFX10-NEXT: v_cndmask_b32_e32 v2, v8, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v8, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v9, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v9, v3, vcc_lo
; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX10-NEXT: v_ashrrev_i32_e32 v8, 31, v7
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1 ; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1
; GFX10-NEXT: v_xor_b32_e32 v9, v11, v10 ; GFX10-NEXT: v_xor_b32_e32 v9, v11, v10
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v15, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v16, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v7, v0, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v8, v0, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v8, v1, s0
; GFX10-NEXT: v_and_b32_e32 v8, 1, v9 ; GFX10-NEXT: v_and_b32_e32 v8, 1, v9
; GFX10-NEXT: v_cndmask_b32_e64 v1, v7, v1, s0
; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0 ; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0
; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, v8 ; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, v8
; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, 0, v0, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, 0, v0, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0x80000000, v1, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0x80000000, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v15, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v16, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v3, s0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v19, v8, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v8, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v20, v9, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v7, v9, s0
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.sadd.sat.i128(i128 %lhs, i128 %rhs) %result = call i128 @llvm.sadd.sat.i128(i128 %lhs, i128 %rhs)
%cast = bitcast i128 %result to <4 x float> %cast = bitcast i128 %result to <4 x float>
ret <4 x float> %cast ret <4 x float> %cast
} }
define <2 x i128> @v_saddsat_v2i128(<2 x i128> %lhs, <2 x i128> %rhs) { define <2 x i128> @v_saddsat_v2i128(<2 x i128> %lhs, <2 x i128> %rhs) {
; GFX6-LABEL: v_saddsat_v2i128: ; GFX6-LABEL: v_saddsat_v2i128:
▲ Show 20 Lines • Show All 316 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e32 v6, v10, v6, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v6, v10, v6, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v7, v11, v7, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v7, v11, v7, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_saddsat_v2i128: ; GFX10-LABEL: v_saddsat_v2i128:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v22, v0 ; GFX10-NEXT: v_add_co_u32 v16, vcc_lo, v0, v8
; GFX10-NEXT: v_mov_b32_e32 v23, v1
; GFX10-NEXT: v_mov_b32_e32 v20, v2
; GFX10-NEXT: v_mov_b32_e32 v21, v3
; GFX10-NEXT: s_movk_i32 s5, 0x7f ; GFX10-NEXT: s_movk_i32 s5, 0x7f
; GFX10-NEXT: v_add_co_u32 v16, vcc_lo, v22, v8 ; GFX10-NEXT: v_add_co_ci_u32_e32 v17, vcc_lo, v1, v9, vcc_lo
; GFX10-NEXT: s_sub_i32 s6, 64, s5 ; GFX10-NEXT: s_sub_i32 s6, 64, s5
; GFX10-NEXT: v_add_co_ci_u32_e32 v17, vcc_lo, v23, v9, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v18, vcc_lo, v2, v10, vcc_lo
; GFX10-NEXT: s_sub_i32 s7, s5, 64 ; GFX10-NEXT: s_sub_i32 s7, s5, 64
; GFX10-NEXT: v_add_co_ci_u32_e32 v18, vcc_lo, v20, v10, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v19, vcc_lo, v3, v11, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[16:17], v[0:1]
; GFX10-NEXT: s_cmp_lt_u32 s5, 64 ; GFX10-NEXT: s_cmp_lt_u32 s5, 64
; GFX10-NEXT: v_add_co_ci_u32_e32 v19, vcc_lo, v21, v11, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[16:17], v[22:23]
; GFX10-NEXT: v_mov_b32_e32 v26, v4
; GFX10-NEXT: v_mov_b32_e32 v27, v5
; GFX10-NEXT: v_mov_b32_e32 v24, v6
; GFX10-NEXT: v_lshlrev_b64 v[2:3], s6, v[18:19]
; GFX10-NEXT: v_mov_b32_e32 v25, v7
; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[18:19], v[20:21] ; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[18:19], v[2:3]
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[18:19], v[20:21] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[18:19], v[2:3]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], s6, v[18:19]
; GFX10-NEXT: v_cndmask_b32_e32 v20, v1, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v20, v1, v0, vcc_lo
; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, 0, v[8:9] ; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, 0, v[8:9]
; GFX10-NEXT: v_lshrrev_b64 v[0:1], s5, v[16:17] ; GFX10-NEXT: v_lshrrev_b64 v[0:1], s5, v[16:17]
; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, 0, v[10:11] ; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, 0, v[10:11]
; GFX10-NEXT: v_or_b32_e32 v2, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v2, v0, v2
; GFX10-NEXT: v_or_b32_e32 v3, v1, v3 ; GFX10-NEXT: v_or_b32_e32 v3, v1, v3
; GFX10-NEXT: v_ashrrev_i64 v[0:1], s5, v[18:19] ; GFX10-NEXT: v_ashrrev_i64 v[0:1], s5, v[18:19]
; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, 0, v[10:11] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, 0, v[10:11]
; GFX10-NEXT: v_ashrrev_i32_e32 v11, 31, v19
; GFX10-NEXT: v_cndmask_b32_e32 v10, v9, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, v9, v8, vcc_lo
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: v_ashrrev_i64 v[8:9], s7, v[18:19] ; GFX10-NEXT: v_ashrrev_i64 v[8:9], s7, v[18:19]
; GFX10-NEXT: s_cmp_eq_u32 s5, 0 ; GFX10-NEXT: s_cmp_eq_u32 s5, 0
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: s_and_b32 s8, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s8, 1, vcc_lo
; GFX10-NEXT: s_and_b32 s4, 1, s4 ; GFX10-NEXT: s_and_b32 s4, 1, s4
; GFX10-NEXT: v_cndmask_b32_e32 v3, v9, v3, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, v8, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v8, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v9, v3, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s4 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s4
; GFX10-NEXT: v_ashrrev_i32_e32 v8, 31, v19
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s8 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s8
; GFX10-NEXT: v_xor_b32_e32 v9, v10, v20 ; GFX10-NEXT: v_xor_b32_e32 v9, v10, v20
; GFX10-NEXT: s_cmp_lt_u32 s5, 64 ; GFX10-NEXT: s_cmp_lt_u32 s5, 64
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v16, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v16, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v17, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v17, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v11, v0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v8, v0, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v11, v1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v8, v1, s4
; GFX10-NEXT: v_and_b32_e32 v8, 1, v9 ; GFX10-NEXT: v_and_b32_e32 v8, 1, v9
; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0 ; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0
; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v20, vcc_lo, 0, v0, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v20, vcc_lo, 0, v0, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v21, vcc_lo, 0x80000000, v1, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v21, vcc_lo, 0x80000000, v1, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v8 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v8
; GFX10-NEXT: v_add_co_u32 v8, s4, v26, v12 ; GFX10-NEXT: v_add_co_u32 v8, s4, v4, v12
; GFX10-NEXT: v_add_co_ci_u32_e64 v9, s4, v27, v13, s4 ; GFX10-NEXT: v_add_co_ci_u32_e64 v9, s4, v5, v13, s4
; GFX10-NEXT: v_add_co_ci_u32_e64 v10, s4, v24, v14, s4 ; GFX10-NEXT: v_add_co_ci_u32_e64 v10, s4, v6, v14, s4
; GFX10-NEXT: v_cndmask_b32_e32 v0, v16, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v16, v2, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e64 v11, s4, v25, v15, s4 ; GFX10-NEXT: v_add_co_ci_u32_e64 v11, s4, v7, v15, s4
; GFX10-NEXT: v_cmp_lt_u64_e64 s4, v[8:9], v[26:27] ; GFX10-NEXT: v_cmp_lt_u64_e64 s4, v[8:9], v[4:5]
; GFX10-NEXT: v_cndmask_b32_e32 v1, v17, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v17, v3, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, v18, v20, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v18, v20, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[3:4], s5, v[8:9] ; GFX10-NEXT: v_lshrrev_b64 v[3:4], s5, v[8:9]
; GFX10-NEXT: v_cndmask_b32_e64 v5, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v5, 0, 1, s4
; GFX10-NEXT: v_cmp_lt_i64_e64 s4, v[10:11], v[24:25] ; GFX10-NEXT: v_cmp_lt_i64_e64 s4, v[10:11], v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v16, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v16, 0, 1, s4
; GFX10-NEXT: v_cmp_gt_u64_e64 s4, 0, v[12:13] ; GFX10-NEXT: v_cmp_gt_u64_e64 s4, 0, v[12:13]
; GFX10-NEXT: v_lshlrev_b64 v[12:13], s6, v[10:11] ; GFX10-NEXT: v_lshlrev_b64 v[12:13], s6, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e64 v17, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v17, 0, 1, s4
; GFX10-NEXT: v_cmp_gt_i64_e64 s4, 0, v[14:15] ; GFX10-NEXT: v_cmp_gt_i64_e64 s4, 0, v[14:15]
; GFX10-NEXT: v_or_b32_e32 v12, v3, v12 ; GFX10-NEXT: v_or_b32_e32 v12, v3, v12
; GFX10-NEXT: v_or_b32_e32 v13, v4, v13 ; GFX10-NEXT: v_or_b32_e32 v13, v4, v13
; GFX10-NEXT: v_ashrrev_i64 v[3:4], s5, v[10:11] ; GFX10-NEXT: v_ashrrev_i64 v[3:4], s5, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e64 v18, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v18, 0, 1, s4
; GFX10-NEXT: v_cmp_eq_u64_e64 s4, v[10:11], v[24:25] ; GFX10-NEXT: v_cmp_eq_u64_e64 s4, v[10:11], v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v7, v16, v5, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v7, v16, v5, s4
; GFX10-NEXT: v_cmp_eq_u64_e64 s4, 0, v[14:15] ; GFX10-NEXT: v_cmp_eq_u64_e64 s4, 0, v[14:15]
; GFX10-NEXT: v_ashrrev_i64 v[5:6], s7, v[10:11] ; GFX10-NEXT: v_ashrrev_i64 v[5:6], s7, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e64 v14, v18, v17, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v14, v18, v17, s4
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s5, 0 ; GFX10-NEXT: s_cmp_eq_u32 s5, 0
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v12, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v12, s4
; GFX10-NEXT: s_cselect_b32 s6, 1, 0 ; GFX10-NEXT: s_cselect_b32 s6, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v13, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v13, s4
; GFX10-NEXT: s_and_b32 s5, 1, s6 ; GFX10-NEXT: s_and_b32 s5, 1, s6
; GFX10-NEXT: s_and_b32 s6, 1, s4 ; GFX10-NEXT: s_and_b32 s6, 1, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5
; GFX10-NEXT: v_xor_b32_e32 v7, v14, v7 ; GFX10-NEXT: v_xor_b32_e32 v7, v14, v7
; GFX10-NEXT: v_ashrrev_i32_e32 v18, 31, v11 ; GFX10-NEXT: v_ashrrev_i32_e32 v12, 31, v11
; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, s6 ; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, s6
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v8, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v8, s4
; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v9, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v9, s4
; GFX10-NEXT: v_and_b32_e32 v7, 1, v7 ; GFX10-NEXT: v_and_b32_e32 v7, 1, v7
; GFX10-NEXT: v_cndmask_b32_e64 v3, v18, v3, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v12, v3, s5
; GFX10-NEXT: v_cndmask_b32_e64 v4, v18, v4, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v12, v4, s5
; GFX10-NEXT: v_add_co_u32 v5, s4, v5, 0 ; GFX10-NEXT: v_add_co_u32 v5, s4, v5, 0
; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s4, 0, v6, s4 ; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s4, 0, v6, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, v7 ; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, v7
; GFX10-NEXT: v_add_co_ci_u32_e64 v7, s4, 0, v3, s4 ; GFX10-NEXT: v_add_co_ci_u32_e64 v7, s4, 0, v3, s4
; GFX10-NEXT: v_cndmask_b32_e32 v3, v19, v21, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v19, v21, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e64 v12, s4, 0x80000000, v4, s4 ; GFX10-NEXT: v_add_co_ci_u32_e64 v12, s4, 0x80000000, v4, s4
; GFX10-NEXT: v_cndmask_b32_e64 v4, v8, v5, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v8, v5, s5
; GFX10-NEXT: v_cndmask_b32_e64 v5, v9, v6, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v9, v6, s5
▲ Show 20 Lines • Show All 520 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_readfirstlane_b32 s6, v2 ; GFX9-NEXT: v_readfirstlane_b32 s6, v2
; GFX9-NEXT: v_readfirstlane_b32 s7, v3 ; GFX9-NEXT: v_readfirstlane_b32 s7, v3
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: s_saddsat_v2i128: ; GFX10-LABEL: s_saddsat_v2i128:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_add_u32 s16, s0, s8 ; GFX10-NEXT: s_add_u32 s16, s0, s8
; GFX10-NEXT: s_cselect_b32 s17, 1, 0 ; GFX10-NEXT: s_cselect_b32 s17, 1, 0
; GFX10-NEXT: s_mov_b32 s46, s0
; GFX10-NEXT: s_and_b32 s17, s17, 1 ; GFX10-NEXT: s_and_b32 s17, s17, 1
; GFX10-NEXT: s_mov_b32 s47, s1
; GFX10-NEXT: s_cmp_lg_u32 s17, 0 ; GFX10-NEXT: s_cmp_lg_u32 s17, 0
; GFX10-NEXT: s_addc_u32 s17, s1, s9 ; GFX10-NEXT: s_addc_u32 s17, s1, s9
; GFX10-NEXT: s_cselect_b32 s18, 1, 0 ; GFX10-NEXT: s_cselect_b32 s18, 1, 0
; GFX10-NEXT: v_cmp_lt_u64_e64 s0, s[16:17], s[46:47] ; GFX10-NEXT: v_cmp_lt_u64_e64 s0, s[16:17], s[0:1]
; GFX10-NEXT: s_and_b32 s18, s18, 1 ; GFX10-NEXT: s_and_b32 s18, s18, 1
; GFX10-NEXT: s_cmp_lg_u32 s18, 0 ; GFX10-NEXT: s_cmp_lg_u32 s18, 0
; GFX10-NEXT: s_addc_u32 s30, s2, s10 ; GFX10-NEXT: s_addc_u32 s18, s2, s10
; GFX10-NEXT: s_cselect_b32 s19, 1, 0 ; GFX10-NEXT: s_cselect_b32 s19, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0
; GFX10-NEXT: s_and_b32 s19, s19, 1 ; GFX10-NEXT: s_and_b32 s19, s19, 1
; GFX10-NEXT: s_cmp_lg_u32 s19, 0 ; GFX10-NEXT: s_cmp_lg_u32 s19, 0
; GFX10-NEXT: s_addc_u32 s31, s3, s11 ; GFX10-NEXT: s_addc_u32 s19, s3, s11
; GFX10-NEXT: v_cmp_lt_i64_e64 s0, s[30:31], s[2:3] ; GFX10-NEXT: v_cmp_lt_i64_e64 s0, s[18:19], s[2:3]
; GFX10-NEXT: s_cmp_eq_u64 s[30:31], s[2:3] ; GFX10-NEXT: s_cmp_eq_u64 s[18:19], s[2:3]
; GFX10-NEXT: v_cmp_lt_u64_e64 s2, s[8:9], 0 ; GFX10-NEXT: v_cmp_lt_u64_e64 s2, s[8:9], 0
; GFX10-NEXT: s_cselect_b32 s20, 1, 0 ; GFX10-NEXT: s_cselect_b32 s20, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s0
; GFX10-NEXT: s_and_b32 s0, 1, s20 ; GFX10-NEXT: s_and_b32 s0, 1, s20
; GFX10-NEXT: s_cmp_eq_u64 s[10:11], 0 ; GFX10-NEXT: s_cmp_eq_u64 s[10:11], 0
; GFX10-NEXT: s_movk_i32 s20, 0x7f ; GFX10-NEXT: s_movk_i32 s20, 0x7f
; GFX10-NEXT: s_cselect_b32 s1, 1, 0 ; GFX10-NEXT: s_cselect_b32 s1, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s2
; GFX10-NEXT: v_cmp_lt_i64_e64 s2, s[10:11], 0 ; GFX10-NEXT: v_cmp_lt_i64_e64 s2, s[10:11], 0
; GFX10-NEXT: s_and_b32 s1, 1, s1 ; GFX10-NEXT: s_and_b32 s1, 1, s1
; GFX10-NEXT: s_sub_i32 s21, s20, 64 ; GFX10-NEXT: s_sub_i32 s21, s20, 64
; GFX10-NEXT: s_sub_i32 s22, 64, s20 ; GFX10-NEXT: s_sub_i32 s22, 64, s20
; GFX10-NEXT: s_cmp_lt_u32 s20, 64 ; GFX10-NEXT: s_cmp_lt_u32 s20, 64
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX10-NEXT: s_cselect_b32 s23, 1, 0 ; GFX10-NEXT: s_cselect_b32 s23, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s20, 0 ; GFX10-NEXT: s_cmp_eq_u32 s20, 0
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s2
; GFX10-NEXT: s_cselect_b32 s24, 1, 0 ; GFX10-NEXT: s_cselect_b32 s24, 1, 0
; GFX10-NEXT: s_lshr_b64 s[2:3], s[16:17], s20 ; GFX10-NEXT: s_lshr_b64 s[2:3], s[16:17], s20
; GFX10-NEXT: s_lshl_b64 s[8:9], s[30:31], s22 ; GFX10-NEXT: s_lshl_b64 s[8:9], s[18:19], s22
; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s1 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s1
; GFX10-NEXT: s_ashr_i32 s10, s31, 31 ; GFX10-NEXT: s_ashr_i32 s10, s19, 31
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[8:9] ; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[8:9]
; GFX10-NEXT: s_ashr_i64 s[0:1], s[30:31], s20 ; GFX10-NEXT: s_ashr_i64 s[0:1], s[18:19], s20
; GFX10-NEXT: s_ashr_i64 s[8:9], s[30:31], s21 ; GFX10-NEXT: s_ashr_i64 s[8:9], s[18:19], s21
; GFX10-NEXT: s_cmp_lg_u32 s23, 0 ; GFX10-NEXT: s_cmp_lg_u32 s23, 0
; GFX10-NEXT: s_mov_b32 s11, s10 ; GFX10-NEXT: s_mov_b32 s11, s10
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[8:9] ; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[8:9]
; GFX10-NEXT: s_cmp_lg_u32 s24, 0 ; GFX10-NEXT: s_cmp_lg_u32 s24, 0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v2, vcc_lo
; GFX10-NEXT: s_cselect_b64 s[2:3], s[16:17], s[2:3] ; GFX10-NEXT: s_cselect_b64 s[2:3], s[16:17], s[2:3]
; GFX10-NEXT: s_cmp_lg_u32 s23, 0 ; GFX10-NEXT: s_cmp_lg_u32 s23, 0
; GFX10-NEXT: v_mov_b32_e32 v2, s17 ; GFX10-NEXT: v_mov_b32_e32 v2, s17
; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[10:11] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[10:11]
; GFX10-NEXT: s_add_u32 s2, s2, 0 ; GFX10-NEXT: s_add_u32 s2, s2, 0
; GFX10-NEXT: s_cselect_b32 s8, 1, 0 ; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: v_xor_b32_e32 v0, v1, v0 ; GFX10-NEXT: v_xor_b32_e32 v0, v1, v0
; GFX10-NEXT: s_and_b32 s8, s8, 1 ; GFX10-NEXT: s_and_b32 s8, s8, 1
; GFX10-NEXT: v_mov_b32_e32 v1, s16 ; GFX10-NEXT: v_mov_b32_e32 v1, s16
; GFX10-NEXT: s_cmp_lg_u32 s8, 0 ; GFX10-NEXT: s_cmp_lg_u32 s8, 0
; GFX10-NEXT: s_brev_b32 s23, 1 ; GFX10-NEXT: s_brev_b32 s23, 1
; GFX10-NEXT: s_addc_u32 s3, s3, 0 ; GFX10-NEXT: s_addc_u32 s3, s3, 0
; GFX10-NEXT: s_cselect_b32 s8, 1, 0 ; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: v_and_b32_e32 v0, 1, v0 ; GFX10-NEXT: v_and_b32_e32 v0, 1, v0
; GFX10-NEXT: s_and_b32 s8, s8, 1 ; GFX10-NEXT: s_and_b32 s8, s8, 1
; GFX10-NEXT: v_mov_b32_e32 v3, s31 ; GFX10-NEXT: v_mov_b32_e32 v3, s19
; GFX10-NEXT: s_cmp_lg_u32 s8, 0 ; GFX10-NEXT: s_cmp_lg_u32 s8, 0
; GFX10-NEXT: s_addc_u32 s0, s0, 0 ; GFX10-NEXT: s_addc_u32 s0, s0, 0
; GFX10-NEXT: s_cselect_b32 s8, 1, 0 ; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0
; GFX10-NEXT: s_and_b32 s8, s8, 1 ; GFX10-NEXT: s_and_b32 s8, s8, 1
; GFX10-NEXT: s_cmp_lg_u32 s8, 0 ; GFX10-NEXT: s_cmp_lg_u32 s8, 0
; GFX10-NEXT: s_addc_u32 s1, s1, s23 ; GFX10-NEXT: s_addc_u32 s1, s1, s23
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s2, vcc_lo
; GFX10-NEXT: s_add_u32 s2, s4, s12 ; GFX10-NEXT: s_add_u32 s2, s4, s12
; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s3, vcc_lo
; GFX10-NEXT: s_cselect_b32 s3, 1, 0 ; GFX10-NEXT: s_cselect_b32 s3, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s1, vcc_lo
; GFX10-NEXT: s_and_b32 s3, s3, 1 ; GFX10-NEXT: s_and_b32 s3, s3, 1
; GFX10-NEXT: v_mov_b32_e32 v2, s30 ; GFX10-NEXT: v_mov_b32_e32 v2, s18
; GFX10-NEXT: s_cmp_lg_u32 s3, 0 ; GFX10-NEXT: s_cmp_lg_u32 s3, 0
; GFX10-NEXT: s_addc_u32 s3, s5, s13 ; GFX10-NEXT: s_addc_u32 s3, s5, s13
; GFX10-NEXT: s_cselect_b32 s8, 1, 0 ; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: v_cmp_lt_u64_e64 s1, s[2:3], s[4:5] ; GFX10-NEXT: v_cmp_lt_u64_e64 s1, s[2:3], s[4:5]
; GFX10-NEXT: s_and_b32 s8, s8, 1 ; GFX10-NEXT: s_and_b32 s8, s8, 1
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s0, vcc_lo
; GFX10-NEXT: s_cmp_lg_u32 s8, 0 ; GFX10-NEXT: s_cmp_lg_u32 s8, 0
; GFX10-NEXT: s_addc_u32 s8, s6, s14 ; GFX10-NEXT: s_addc_u32 s8, s6, s14
▲ Show 20 Lines • Show All 106 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/shl.ll

Show First 20 Lines • Show All 1,565 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_lshlrev_b64 v[0:1], v4, v[0:1] ; GFX9-NEXT: v_lshlrev_b64 v[0:1], v4, v[0:1]
; GFX9-NEXT: v_lshlrev_b64 v[2:3], v6, v[2:3] ; GFX9-NEXT: v_lshlrev_b64 v[2:3], v6, v[2:3]
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_shl_v2i64: ; GFX10-LABEL: v_shl_v2i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v10, v0 ; GFX10-NEXT: v_lshlrev_b64 v[0:1], v4, v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v11, v1 ; GFX10-NEXT: v_lshlrev_b64 v[2:3], v6, v[2:3]
; GFX10-NEXT: v_mov_b32_e32 v7, v2
; GFX10-NEXT: v_mov_b32_e32 v8, v3
; GFX10-NEXT: v_lshlrev_b64 v[0:1], v4, v[10:11]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], v6, v[7:8]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = shl <2 x i64> %value, %amount %result = shl <2 x i64> %value, %amount
ret <2 x i64> %result ret <2 x i64> %result
} }
define <2 x i64> @v_shl_v2i64_31(<2 x i64> %value) { define <2 x i64> @v_shl_v2i64_31(<2 x i64> %value) {
; GFX6-LABEL: v_shl_v2i64_31: ; GFX6-LABEL: v_shl_v2i64_31:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 45 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/ssubsat.ll

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 4,179 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v2, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_ssubsat_i64: ; GFX10-LABEL: v_ssubsat_i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_sub_co_u32 v10, vcc_lo, v0, v2 ; GFX10-NEXT: v_sub_co_u32 v4, vcc_lo, v0, v2
; GFX10-NEXT: v_sub_co_ci_u32_e32 v11, vcc_lo, v1, v3, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v5, vcc_lo, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, 0, v[2:3] ; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, 0, v[2:3]
; GFX10-NEXT: v_ashrrev_i32_e32 v6, 31, v11 ; GFX10-NEXT: v_ashrrev_i32_e32 v6, 31, v5
; GFX10-NEXT: v_cmp_lt_i64_e64 s4, v[10:11], v[0:1] ; GFX10-NEXT: v_cmp_lt_i64_e64 s4, v[4:5], v[0:1]
; GFX10-NEXT: v_add_co_u32 v0, s5, v6, 0 ; GFX10-NEXT: v_add_co_u32 v0, s5, v6, 0
; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s5, 0x80000000, v6, s5 ; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s5, 0x80000000, v6, s5
; GFX10-NEXT: s_xor_b32 vcc_lo, vcc_lo, s4 ; GFX10-NEXT: s_xor_b32 vcc_lo, vcc_lo, s4
; GFX10-NEXT: v_cndmask_b32_e32 v0, v10, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v11, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i64 @llvm.ssub.sat.i64(i64 %lhs, i64 %rhs) %result = call i64 @llvm.ssub.sat.i64(i64 %lhs, i64 %rhs)
ret i64 %result ret i64 %result
} }
define amdgpu_ps i64 @s_ssubsat_i64(i64 inreg %lhs, i64 inreg %rhs) { define amdgpu_ps i64 @s_ssubsat_i64(i64 inreg %lhs, i64 inreg %rhs) {
; GFX6-LABEL: s_ssubsat_i64: ; GFX6-LABEL: s_ssubsat_i64:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 318 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e32 v2, v4, v3, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v2, v4, v3, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, v5, v6, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v3, v5, v6, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_ssubsat_v2i64: ; GFX10-LABEL: v_ssubsat_v2i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v14, v0 ; GFX10-NEXT: v_sub_co_u32 v8, vcc_lo, v0, v4
; GFX10-NEXT: v_mov_b32_e32 v15, v1
; GFX10-NEXT: v_mov_b32_e32 v17, v2
; GFX10-NEXT: v_mov_b32_e32 v18, v3
; GFX10-NEXT: v_cmp_lt_i64_e64 s4, 0, v[4:5] ; GFX10-NEXT: v_cmp_lt_i64_e64 s4, 0, v[4:5]
; GFX10-NEXT: v_sub_co_u32 v8, vcc_lo, v14, v4 ; GFX10-NEXT: v_sub_co_ci_u32_e32 v9, vcc_lo, v1, v5, vcc_lo
; GFX10-NEXT: v_sub_co_u32 v10, vcc_lo, v2, v6
; GFX10-NEXT: v_cmp_lt_i64_e64 s6, 0, v[6:7] ; GFX10-NEXT: v_cmp_lt_i64_e64 s6, 0, v[6:7]
; GFX10-NEXT: v_sub_co_ci_u32_e32 v9, vcc_lo, v15, v5, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v11, vcc_lo, v3, v7, vcc_lo
; GFX10-NEXT: v_sub_co_u32 v19, vcc_lo, v17, v6
; GFX10-NEXT: v_sub_co_ci_u32_e32 v20, vcc_lo, v18, v7, vcc_lo
; GFX10-NEXT: v_ashrrev_i32_e32 v12, 31, v9 ; GFX10-NEXT: v_ashrrev_i32_e32 v12, 31, v9
; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[8:9], v[14:15] ; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[8:9], v[0:1]
; GFX10-NEXT: v_ashrrev_i32_e32 v0, 31, v20 ; GFX10-NEXT: v_ashrrev_i32_e32 v0, 31, v11
; GFX10-NEXT: v_add_co_u32 v1, s5, v12, 0 ; GFX10-NEXT: v_add_co_u32 v1, s5, v12, 0
; GFX10-NEXT: v_add_co_ci_u32_e64 v4, s5, 0x80000000, v12, s5 ; GFX10-NEXT: v_add_co_ci_u32_e64 v4, s5, 0x80000000, v12, s5
; GFX10-NEXT: v_cmp_lt_i64_e64 s5, v[19:20], v[17:18] ; GFX10-NEXT: v_cmp_lt_i64_e64 s5, v[10:11], v[2:3]
; GFX10-NEXT: v_add_co_u32 v2, s7, v0, 0 ; GFX10-NEXT: v_add_co_u32 v2, s7, v0, 0
; GFX10-NEXT: s_xor_b32 vcc_lo, s4, vcc_lo ; GFX10-NEXT: s_xor_b32 vcc_lo, s4, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e64 v3, s7, 0x80000000, v0, s7 ; GFX10-NEXT: v_add_co_ci_u32_e64 v3, s7, 0x80000000, v0, s7
; GFX10-NEXT: v_cndmask_b32_e32 v0, v8, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v8, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v9, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v9, v4, vcc_lo
; GFX10-NEXT: s_xor_b32 vcc_lo, s6, s5 ; GFX10-NEXT: s_xor_b32 vcc_lo, s6, s5
; GFX10-NEXT: v_cndmask_b32_e32 v2, v19, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v10, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v20, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v11, v3, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i64> @llvm.ssub.sat.v2i64(<2 x i64> %lhs, <2 x i64> %rhs) %result = call <2 x i64> @llvm.ssub.sat.v2i64(<2 x i64> %lhs, <2 x i64> %rhs)
ret <2 x i64> %result ret <2 x i64> %result
} }
define amdgpu_ps <2 x i64> @s_ssubsat_v2i64(<2 x i64> inreg %lhs, <2 x i64> inreg %rhs) { define amdgpu_ps <2 x i64> @s_ssubsat_v2i64(<2 x i64> inreg %lhs, <2 x i64> inreg %rhs) {
; GFX6-LABEL: s_ssubsat_v2i64: ; GFX6-LABEL: s_ssubsat_v2i64:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 744 Lines • ▼ Show 20 Lines
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: v_sub_co_u32 v4, vcc_lo, s0, v0 ; GFX10-NEXT: v_sub_co_u32 v4, vcc_lo, s0, v0
; GFX10-NEXT: v_sub_co_ci_u32_e32 v5, vcc_lo, s1, v1, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v5, vcc_lo, s1, v1, vcc_lo
; GFX10-NEXT: v_sub_co_ci_u32_e32 v6, vcc_lo, s2, v2, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v6, vcc_lo, s2, v2, vcc_lo
; GFX10-NEXT: v_sub_co_ci_u32_e32 v7, vcc_lo, s3, v3, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v7, vcc_lo, s3, v3, vcc_lo
; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[0:1], v[4:5] ; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[0:1], v[4:5]
; GFX10-NEXT: s_movk_i32 s0, 0x7f ; GFX10-NEXT: s_movk_i32 s0, 0x7f
; GFX10-NEXT: s_sub_i32 s1, 64, s0 ; GFX10-NEXT: s_sub_i32 s1, 64, s0
; GFX10-NEXT: v_lshrrev_b64 v[15:16], s0, v[4:5]
; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, s[2:3], v[6:7] ; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, s[2:3], v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, s[2:3], v[6:7] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, s[2:3], v[6:7]
; GFX10-NEXT: v_cndmask_b32_e32 v10, v9, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, v9, v8, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, 0, v[0:1] ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, 0, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[8:9], s1, v[6:7] ; GFX10-NEXT: v_lshlrev_b64 v[8:9], s1, v[6:7]
; GFX10-NEXT: v_lshrrev_b64 v[0:1], s0, v[4:5]
; GFX10-NEXT: s_sub_i32 s1, s0, 64 ; GFX10-NEXT: s_sub_i32 s1, s0, 64
; GFX10-NEXT: s_cmp_lt_u32 s0, 64 ; GFX10-NEXT: s_cmp_lt_u32 s0, 64
; GFX10-NEXT: v_ashrrev_i64 v[0:1], s0, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v11, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v11, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, 0, v[2:3] ; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, 0, v[2:3]
; GFX10-NEXT: v_or_b32_e32 v8, v15, v8 ; GFX10-NEXT: v_or_b32_e32 v8, v0, v8
; GFX10-NEXT: v_or_b32_e32 v9, v16, v9 ; GFX10-NEXT: v_or_b32_e32 v9, v1, v9
; GFX10-NEXT: v_ashrrev_i32_e32 v15, 31, v7 ; GFX10-NEXT: v_ashrrev_i64 v[0:1], s0, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v12, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v12, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, 0, v[2:3] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, 0, v[2:3]
; GFX10-NEXT: v_ashrrev_i64 v[2:3], s1, v[6:7] ; GFX10-NEXT: v_ashrrev_i64 v[2:3], s1, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e32 v11, v12, v11, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v11, v12, v11, vcc_lo
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s0, 0 ; GFX10-NEXT: s_cmp_eq_u32 s0, 0
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v9, vcc_lo
; GFX10-NEXT: s_cselect_b32 s1, 1, 0
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v8, vcc_lo
; GFX10-NEXT: s_cselect_b32 s1, 1, 0
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v9, vcc_lo
; GFX10-NEXT: s_and_b32 s0, 1, s1 ; GFX10-NEXT: s_and_b32 s0, 1, s1
; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX10-NEXT: v_ashrrev_i32_e32 v8, 31, v7
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1 ; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1
; GFX10-NEXT: v_xor_b32_e32 v9, v11, v10 ; GFX10-NEXT: v_xor_b32_e32 v9, v11, v10
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v5, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v15, v0, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v8, v0, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v8, v1, s0
; GFX10-NEXT: v_and_b32_e32 v8, 1, v9 ; GFX10-NEXT: v_and_b32_e32 v8, 1, v9
; GFX10-NEXT: v_cndmask_b32_e64 v1, v15, v1, s0
; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0 ; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0
; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, v8 ; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, v8
; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, 0, v0, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, 0, v0, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0x80000000, v1, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0x80000000, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v3, s0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v8, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v8, s0
▲ Show 20 Lines • Show All 191 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v2, v6, v2, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v2, v6, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, v7, v3, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v3, v7, v3, vcc
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: ssubsat_i128_vs: ; GFX10-LABEL: ssubsat_i128_vs:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; GFX10-NEXT: v_sub_co_u32 v4, vcc_lo, v0, s0
; GFX10-NEXT: v_mov_b32_e32 v6, v1
; GFX10-NEXT: v_mov_b32_e32 v9, v2
; GFX10-NEXT: v_mov_b32_e32 v10, v3
; GFX10-NEXT: s_cmp_eq_u64 s[2:3], 0
; GFX10-NEXT: v_sub_co_u32 v15, vcc_lo, v5, s0
; GFX10-NEXT: v_cmp_gt_u64_e64 s0, s[0:1], 0 ; GFX10-NEXT: v_cmp_gt_u64_e64 s0, s[0:1], 0
; GFX10-NEXT: v_subrev_co_ci_u32_e32 v16, vcc_lo, s1, v6, vcc_lo ; GFX10-NEXT: v_subrev_co_ci_u32_e32 v5, vcc_lo, s1, v1, vcc_lo
; GFX10-NEXT: s_cmp_eq_u64 s[2:3], 0
; GFX10-NEXT: v_subrev_co_ci_u32_e32 v6, vcc_lo, s2, v2, vcc_lo
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: v_subrev_co_ci_u32_e32 v19, vcc_lo, s2, v9, vcc_lo ; GFX10-NEXT: v_subrev_co_ci_u32_e32 v7, vcc_lo, s3, v3, vcc_lo
; GFX10-NEXT: s_and_b32 s1, 1, s4 ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[4:5], v[0:1]
; GFX10-NEXT: v_subrev_co_ci_u32_e32 v20, vcc_lo, s3, v10, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[15:16], v[5:6]
; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, s0
; GFX10-NEXT: v_cmp_gt_i64_e64 s0, s[2:3], 0 ; GFX10-NEXT: v_cmp_gt_i64_e64 s0, s[2:3], 0
; GFX10-NEXT: v_ashrrev_i32_e32 v7, 31, v20 ; GFX10-NEXT: s_and_b32 s1, 1, s4
; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[19:20], v[9:10] ; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[6:7], v[2:3]
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[19:20], v[9:10]
; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, s0
; GFX10-NEXT: s_movk_i32 s0, 0x7f ; GFX10-NEXT: s_movk_i32 s0, 0x7f
; GFX10-NEXT: s_sub_i32 s2, 64, s0 ; GFX10-NEXT: s_sub_i32 s2, 64, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[6:7], v[2:3]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], s2, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e32 v10, v1, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, v1, v0, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[0:1], s0, v[4:5]
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s1 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s1
; GFX10-NEXT: v_lshrrev_b64 v[0:1], s0, v[15:16]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], s2, v[19:20]
; GFX10-NEXT: s_sub_i32 s1, s0, 64 ; GFX10-NEXT: s_sub_i32 s1, s0, 64
; GFX10-NEXT: s_cmp_lt_u32 s0, 64 ; GFX10-NEXT: s_cmp_lt_u32 s0, 64
; GFX10-NEXT: v_cndmask_b32_e32 v11, v9, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v11, v9, v8, vcc_lo
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: v_ashrrev_i64 v[8:9], s1, v[19:20] ; GFX10-NEXT: v_ashrrev_i64 v[8:9], s1, v[6:7]
; GFX10-NEXT: s_cmp_eq_u32 s0, 0 ; GFX10-NEXT: s_cmp_eq_u32 s0, 0
; GFX10-NEXT: v_or_b32_e32 v2, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v2, v0, v2
; GFX10-NEXT: v_or_b32_e32 v3, v1, v3 ; GFX10-NEXT: v_or_b32_e32 v3, v1, v3
; GFX10-NEXT: s_cselect_b32 s1, 1, 0 ; GFX10-NEXT: s_cselect_b32 s1, 1, 0
; GFX10-NEXT: v_ashrrev_i64 v[0:1], s0, v[19:20] ; GFX10-NEXT: v_ashrrev_i64 v[0:1], s0, v[6:7]
; GFX10-NEXT: s_and_b32 s0, 1, s1 ; GFX10-NEXT: s_and_b32 s0, 1, s1
; GFX10-NEXT: v_cndmask_b32_e32 v2, v8, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v8, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v9, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v9, v3, vcc_lo
; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX10-NEXT: v_ashrrev_i32_e32 v8, 31, v7
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1 ; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1
; GFX10-NEXT: v_xor_b32_e32 v9, v11, v10 ; GFX10-NEXT: v_xor_b32_e32 v9, v11, v10
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v15, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v16, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v7, v0, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v8, v0, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v8, v1, s0
; GFX10-NEXT: v_and_b32_e32 v8, 1, v9 ; GFX10-NEXT: v_and_b32_e32 v8, 1, v9
; GFX10-NEXT: v_cndmask_b32_e64 v1, v7, v1, s0
; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0 ; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0
; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, v8 ; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, v8
; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, 0, v0, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, 0, v0, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0x80000000, v1, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, 0x80000000, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v15, v2, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v16, v3, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v3, s0
; GFX10-NEXT: v_cndmask_b32_e64 v2, v19, v8, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v8, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v20, v9, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v7, v9, s0
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.ssub.sat.i128(i128 %lhs, i128 %rhs) %result = call i128 @llvm.ssub.sat.i128(i128 %lhs, i128 %rhs)
%cast = bitcast i128 %result to <4 x float> %cast = bitcast i128 %result to <4 x float>
ret <4 x float> %cast ret <4 x float> %cast
} }
define <2 x i128> @v_ssubsat_v2i128(<2 x i128> %lhs, <2 x i128> %rhs) { define <2 x i128> @v_ssubsat_v2i128(<2 x i128> %lhs, <2 x i128> %rhs) {
; GFX6-LABEL: v_ssubsat_v2i128: ; GFX6-LABEL: v_ssubsat_v2i128:
▲ Show 20 Lines • Show All 316 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e32 v6, v10, v6, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v6, v10, v6, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v7, v11, v7, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v7, v11, v7, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_ssubsat_v2i128: ; GFX10-LABEL: v_ssubsat_v2i128:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v22, v0 ; GFX10-NEXT: v_sub_co_u32 v16, vcc_lo, v0, v8
; GFX10-NEXT: v_mov_b32_e32 v23, v1
; GFX10-NEXT: v_mov_b32_e32 v20, v2
; GFX10-NEXT: v_mov_b32_e32 v21, v3
; GFX10-NEXT: s_movk_i32 s5, 0x7f ; GFX10-NEXT: s_movk_i32 s5, 0x7f
; GFX10-NEXT: v_sub_co_u32 v16, vcc_lo, v22, v8 ; GFX10-NEXT: v_sub_co_ci_u32_e32 v17, vcc_lo, v1, v9, vcc_lo
; GFX10-NEXT: s_sub_i32 s6, 64, s5 ; GFX10-NEXT: s_sub_i32 s6, 64, s5
; GFX10-NEXT: v_sub_co_ci_u32_e32 v17, vcc_lo, v23, v9, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v18, vcc_lo, v2, v10, vcc_lo
; GFX10-NEXT: s_sub_i32 s7, s5, 64 ; GFX10-NEXT: s_sub_i32 s7, s5, 64
; GFX10-NEXT: v_sub_co_ci_u32_e32 v18, vcc_lo, v20, v10, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v19, vcc_lo, v3, v11, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[16:17], v[0:1]
; GFX10-NEXT: s_cmp_lt_u32 s5, 64 ; GFX10-NEXT: s_cmp_lt_u32 s5, 64
; GFX10-NEXT: v_sub_co_ci_u32_e32 v19, vcc_lo, v21, v11, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[16:17], v[22:23]
; GFX10-NEXT: v_mov_b32_e32 v26, v4
; GFX10-NEXT: v_mov_b32_e32 v27, v5
; GFX10-NEXT: v_mov_b32_e32 v24, v6
; GFX10-NEXT: v_lshlrev_b64 v[2:3], s6, v[18:19]
; GFX10-NEXT: v_mov_b32_e32 v25, v7
; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[18:19], v[20:21] ; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[18:19], v[2:3]
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[18:19], v[20:21] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[18:19], v[2:3]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], s6, v[18:19]
; GFX10-NEXT: v_cndmask_b32_e32 v20, v1, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v20, v1, v0, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, 0, v[8:9] ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, 0, v[8:9]
; GFX10-NEXT: v_lshrrev_b64 v[0:1], s5, v[16:17] ; GFX10-NEXT: v_lshrrev_b64 v[0:1], s5, v[16:17]
; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, 0, v[10:11] ; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, 0, v[10:11]
; GFX10-NEXT: v_or_b32_e32 v2, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v2, v0, v2
; GFX10-NEXT: v_or_b32_e32 v3, v1, v3 ; GFX10-NEXT: v_or_b32_e32 v3, v1, v3
; GFX10-NEXT: v_ashrrev_i64 v[0:1], s5, v[18:19] ; GFX10-NEXT: v_ashrrev_i64 v[0:1], s5, v[18:19]
; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, 0, v[10:11] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, 0, v[10:11]
; GFX10-NEXT: v_ashrrev_i32_e32 v11, 31, v19
; GFX10-NEXT: v_cndmask_b32_e32 v10, v9, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v10, v9, v8, vcc_lo
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0 ; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: v_ashrrev_i64 v[8:9], s7, v[18:19] ; GFX10-NEXT: v_ashrrev_i64 v[8:9], s7, v[18:19]
; GFX10-NEXT: s_cmp_eq_u32 s5, 0 ; GFX10-NEXT: s_cmp_eq_u32 s5, 0
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: s_and_b32 s8, 1, vcc_lo ; GFX10-NEXT: s_and_b32 s8, 1, vcc_lo
; GFX10-NEXT: s_and_b32 s4, 1, s4 ; GFX10-NEXT: s_and_b32 s4, 1, s4
; GFX10-NEXT: v_cndmask_b32_e32 v3, v9, v3, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, v8, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v8, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v9, v3, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s4 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s4
; GFX10-NEXT: v_ashrrev_i32_e32 v8, 31, v19
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s8 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s8
; GFX10-NEXT: v_xor_b32_e32 v9, v10, v20 ; GFX10-NEXT: v_xor_b32_e32 v9, v10, v20
; GFX10-NEXT: s_cmp_lt_u32 s5, 64 ; GFX10-NEXT: s_cmp_lt_u32 s5, 64
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v16, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v16, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v17, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v17, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v11, v0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v8, v0, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v11, v1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v8, v1, s4
; GFX10-NEXT: v_and_b32_e32 v8, 1, v9 ; GFX10-NEXT: v_and_b32_e32 v8, 1, v9
; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0 ; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, 0
; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v20, vcc_lo, 0, v0, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v20, vcc_lo, 0, v0, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v21, vcc_lo, 0x80000000, v1, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v21, vcc_lo, 0x80000000, v1, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v8 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v8
; GFX10-NEXT: v_sub_co_u32 v8, s4, v26, v12 ; GFX10-NEXT: v_sub_co_u32 v8, s4, v4, v12
; GFX10-NEXT: v_sub_co_ci_u32_e64 v9, s4, v27, v13, s4 ; GFX10-NEXT: v_sub_co_ci_u32_e64 v9, s4, v5, v13, s4
; GFX10-NEXT: v_sub_co_ci_u32_e64 v10, s4, v24, v14, s4 ; GFX10-NEXT: v_sub_co_ci_u32_e64 v10, s4, v6, v14, s4
; GFX10-NEXT: v_cndmask_b32_e32 v0, v16, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v16, v2, vcc_lo
; GFX10-NEXT: v_sub_co_ci_u32_e64 v11, s4, v25, v15, s4 ; GFX10-NEXT: v_sub_co_ci_u32_e64 v11, s4, v7, v15, s4
; GFX10-NEXT: v_cmp_lt_u64_e64 s4, v[8:9], v[26:27] ; GFX10-NEXT: v_cmp_lt_u64_e64 s4, v[8:9], v[4:5]
; GFX10-NEXT: v_cndmask_b32_e32 v1, v17, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v17, v3, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, v18, v20, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v2, v18, v20, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[3:4], s5, v[8:9] ; GFX10-NEXT: v_lshrrev_b64 v[3:4], s5, v[8:9]
; GFX10-NEXT: v_cndmask_b32_e64 v5, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v5, 0, 1, s4
; GFX10-NEXT: v_cmp_lt_i64_e64 s4, v[10:11], v[24:25] ; GFX10-NEXT: v_cmp_lt_i64_e64 s4, v[10:11], v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v16, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v16, 0, 1, s4
; GFX10-NEXT: v_cmp_lt_u64_e64 s4, 0, v[12:13] ; GFX10-NEXT: v_cmp_lt_u64_e64 s4, 0, v[12:13]
; GFX10-NEXT: v_lshlrev_b64 v[12:13], s6, v[10:11] ; GFX10-NEXT: v_lshlrev_b64 v[12:13], s6, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e64 v17, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v17, 0, 1, s4
; GFX10-NEXT: v_cmp_lt_i64_e64 s4, 0, v[14:15] ; GFX10-NEXT: v_cmp_lt_i64_e64 s4, 0, v[14:15]
; GFX10-NEXT: v_or_b32_e32 v12, v3, v12 ; GFX10-NEXT: v_or_b32_e32 v12, v3, v12
; GFX10-NEXT: v_or_b32_e32 v13, v4, v13 ; GFX10-NEXT: v_or_b32_e32 v13, v4, v13
; GFX10-NEXT: v_ashrrev_i64 v[3:4], s5, v[10:11] ; GFX10-NEXT: v_ashrrev_i64 v[3:4], s5, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e64 v18, 0, 1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v18, 0, 1, s4
; GFX10-NEXT: v_cmp_eq_u64_e64 s4, v[10:11], v[24:25] ; GFX10-NEXT: v_cmp_eq_u64_e64 s4, v[10:11], v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v7, v16, v5, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v7, v16, v5, s4
; GFX10-NEXT: v_cmp_eq_u64_e64 s4, 0, v[14:15] ; GFX10-NEXT: v_cmp_eq_u64_e64 s4, 0, v[14:15]
; GFX10-NEXT: v_ashrrev_i64 v[5:6], s7, v[10:11] ; GFX10-NEXT: v_ashrrev_i64 v[5:6], s7, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e64 v14, v18, v17, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v14, v18, v17, s4
; GFX10-NEXT: s_cselect_b32 s4, 1, 0 ; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s5, 0 ; GFX10-NEXT: s_cmp_eq_u32 s5, 0
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v12, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v12, s4
; GFX10-NEXT: s_cselect_b32 s6, 1, 0 ; GFX10-NEXT: s_cselect_b32 s6, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v13, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v13, s4
; GFX10-NEXT: s_and_b32 s5, 1, s6 ; GFX10-NEXT: s_and_b32 s5, 1, s6
; GFX10-NEXT: s_and_b32 s6, 1, s4 ; GFX10-NEXT: s_and_b32 s6, 1, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5
; GFX10-NEXT: v_xor_b32_e32 v7, v14, v7 ; GFX10-NEXT: v_xor_b32_e32 v7, v14, v7
; GFX10-NEXT: v_ashrrev_i32_e32 v18, 31, v11 ; GFX10-NEXT: v_ashrrev_i32_e32 v12, 31, v11
; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, s6 ; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, s6
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v8, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v8, s4
; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v9, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v9, s4
; GFX10-NEXT: v_and_b32_e32 v7, 1, v7 ; GFX10-NEXT: v_and_b32_e32 v7, 1, v7
; GFX10-NEXT: v_cndmask_b32_e64 v3, v18, v3, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v12, v3, s5
; GFX10-NEXT: v_cndmask_b32_e64 v4, v18, v4, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v12, v4, s5
; GFX10-NEXT: v_add_co_u32 v5, s4, v5, 0 ; GFX10-NEXT: v_add_co_u32 v5, s4, v5, 0
; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s4, 0, v6, s4 ; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s4, 0, v6, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, v7 ; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, v7
; GFX10-NEXT: v_add_co_ci_u32_e64 v7, s4, 0, v3, s4 ; GFX10-NEXT: v_add_co_ci_u32_e64 v7, s4, 0, v3, s4
; GFX10-NEXT: v_cndmask_b32_e32 v3, v19, v21, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v3, v19, v21, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e64 v12, s4, 0x80000000, v4, s4 ; GFX10-NEXT: v_add_co_ci_u32_e64 v12, s4, 0x80000000, v4, s4
; GFX10-NEXT: v_cndmask_b32_e64 v4, v8, v5, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v8, v5, s5
; GFX10-NEXT: v_cndmask_b32_e64 v5, v9, v6, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v9, v6, s5
▲ Show 20 Lines • Show All 520 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_readfirstlane_b32 s6, v2 ; GFX9-NEXT: v_readfirstlane_b32 s6, v2
; GFX9-NEXT: v_readfirstlane_b32 s7, v3 ; GFX9-NEXT: v_readfirstlane_b32 s7, v3
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: s_ssubsat_v2i128: ; GFX10-LABEL: s_ssubsat_v2i128:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_sub_u32 s16, s0, s8 ; GFX10-NEXT: s_sub_u32 s16, s0, s8
; GFX10-NEXT: s_cselect_b32 s17, 1, 0 ; GFX10-NEXT: s_cselect_b32 s17, 1, 0
; GFX10-NEXT: s_mov_b32 s46, s0
; GFX10-NEXT: s_and_b32 s17, s17, 1 ; GFX10-NEXT: s_and_b32 s17, s17, 1
; GFX10-NEXT: s_mov_b32 s47, s1
; GFX10-NEXT: s_cmp_lg_u32 s17, 0 ; GFX10-NEXT: s_cmp_lg_u32 s17, 0
; GFX10-NEXT: s_subb_u32 s17, s1, s9 ; GFX10-NEXT: s_subb_u32 s17, s1, s9
; GFX10-NEXT: s_cselect_b32 s18, 1, 0 ; GFX10-NEXT: s_cselect_b32 s18, 1, 0
; GFX10-NEXT: v_cmp_lt_u64_e64 s0, s[16:17], s[46:47] ; GFX10-NEXT: v_cmp_lt_u64_e64 s0, s[16:17], s[0:1]
; GFX10-NEXT: s_and_b32 s18, s18, 1 ; GFX10-NEXT: s_and_b32 s18, s18, 1
; GFX10-NEXT: s_cmp_lg_u32 s18, 0 ; GFX10-NEXT: s_cmp_lg_u32 s18, 0
; GFX10-NEXT: s_subb_u32 s30, s2, s10 ; GFX10-NEXT: s_subb_u32 s18, s2, s10
; GFX10-NEXT: s_cselect_b32 s19, 1, 0 ; GFX10-NEXT: s_cselect_b32 s19, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0
; GFX10-NEXT: s_and_b32 s19, s19, 1 ; GFX10-NEXT: s_and_b32 s19, s19, 1
; GFX10-NEXT: s_cmp_lg_u32 s19, 0 ; GFX10-NEXT: s_cmp_lg_u32 s19, 0
; GFX10-NEXT: s_subb_u32 s31, s3, s11 ; GFX10-NEXT: s_subb_u32 s19, s3, s11
; GFX10-NEXT: v_cmp_lt_i64_e64 s0, s[30:31], s[2:3] ; GFX10-NEXT: v_cmp_lt_i64_e64 s0, s[18:19], s[2:3]
; GFX10-NEXT: s_cmp_eq_u64 s[30:31], s[2:3] ; GFX10-NEXT: s_cmp_eq_u64 s[18:19], s[2:3]
; GFX10-NEXT: v_cmp_gt_u64_e64 s2, s[8:9], 0 ; GFX10-NEXT: v_cmp_gt_u64_e64 s2, s[8:9], 0
; GFX10-NEXT: s_cselect_b32 s20, 1, 0 ; GFX10-NEXT: s_cselect_b32 s20, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s0
; GFX10-NEXT: s_and_b32 s0, 1, s20 ; GFX10-NEXT: s_and_b32 s0, 1, s20
; GFX10-NEXT: s_cmp_eq_u64 s[10:11], 0 ; GFX10-NEXT: s_cmp_eq_u64 s[10:11], 0
; GFX10-NEXT: s_movk_i32 s20, 0x7f ; GFX10-NEXT: s_movk_i32 s20, 0x7f
; GFX10-NEXT: s_cselect_b32 s1, 1, 0 ; GFX10-NEXT: s_cselect_b32 s1, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s2
; GFX10-NEXT: v_cmp_gt_i64_e64 s2, s[10:11], 0 ; GFX10-NEXT: v_cmp_gt_i64_e64 s2, s[10:11], 0
; GFX10-NEXT: s_and_b32 s1, 1, s1 ; GFX10-NEXT: s_and_b32 s1, 1, s1
; GFX10-NEXT: s_sub_i32 s21, s20, 64 ; GFX10-NEXT: s_sub_i32 s21, s20, 64
; GFX10-NEXT: s_sub_i32 s22, 64, s20 ; GFX10-NEXT: s_sub_i32 s22, 64, s20
; GFX10-NEXT: s_cmp_lt_u32 s20, 64 ; GFX10-NEXT: s_cmp_lt_u32 s20, 64
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX10-NEXT: s_cselect_b32 s23, 1, 0 ; GFX10-NEXT: s_cselect_b32 s23, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s20, 0 ; GFX10-NEXT: s_cmp_eq_u32 s20, 0
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s2
; GFX10-NEXT: s_cselect_b32 s24, 1, 0 ; GFX10-NEXT: s_cselect_b32 s24, 1, 0
; GFX10-NEXT: s_lshr_b64 s[2:3], s[16:17], s20 ; GFX10-NEXT: s_lshr_b64 s[2:3], s[16:17], s20
; GFX10-NEXT: s_lshl_b64 s[8:9], s[30:31], s22 ; GFX10-NEXT: s_lshl_b64 s[8:9], s[18:19], s22
; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s1 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s1
; GFX10-NEXT: s_ashr_i32 s10, s31, 31 ; GFX10-NEXT: s_ashr_i32 s10, s19, 31
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[8:9] ; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[8:9]
; GFX10-NEXT: s_ashr_i64 s[0:1], s[30:31], s20 ; GFX10-NEXT: s_ashr_i64 s[0:1], s[18:19], s20
; GFX10-NEXT: s_ashr_i64 s[8:9], s[30:31], s21 ; GFX10-NEXT: s_ashr_i64 s[8:9], s[18:19], s21
; GFX10-NEXT: s_cmp_lg_u32 s23, 0 ; GFX10-NEXT: s_cmp_lg_u32 s23, 0
; GFX10-NEXT: s_mov_b32 s11, s10 ; GFX10-NEXT: s_mov_b32 s11, s10
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[8:9] ; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[8:9]
; GFX10-NEXT: s_cmp_lg_u32 s24, 0 ; GFX10-NEXT: s_cmp_lg_u32 s24, 0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v3, v2, vcc_lo
; GFX10-NEXT: s_cselect_b64 s[2:3], s[16:17], s[2:3] ; GFX10-NEXT: s_cselect_b64 s[2:3], s[16:17], s[2:3]
; GFX10-NEXT: s_cmp_lg_u32 s23, 0 ; GFX10-NEXT: s_cmp_lg_u32 s23, 0
; GFX10-NEXT: v_mov_b32_e32 v2, s17 ; GFX10-NEXT: v_mov_b32_e32 v2, s17
; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[10:11] ; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[10:11]
; GFX10-NEXT: s_add_u32 s2, s2, 0 ; GFX10-NEXT: s_add_u32 s2, s2, 0
; GFX10-NEXT: s_cselect_b32 s8, 1, 0 ; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: v_xor_b32_e32 v0, v1, v0 ; GFX10-NEXT: v_xor_b32_e32 v0, v1, v0
; GFX10-NEXT: s_and_b32 s8, s8, 1 ; GFX10-NEXT: s_and_b32 s8, s8, 1
; GFX10-NEXT: v_mov_b32_e32 v1, s16 ; GFX10-NEXT: v_mov_b32_e32 v1, s16
; GFX10-NEXT: s_cmp_lg_u32 s8, 0 ; GFX10-NEXT: s_cmp_lg_u32 s8, 0
; GFX10-NEXT: s_brev_b32 s23, 1 ; GFX10-NEXT: s_brev_b32 s23, 1
; GFX10-NEXT: s_addc_u32 s3, s3, 0 ; GFX10-NEXT: s_addc_u32 s3, s3, 0
; GFX10-NEXT: s_cselect_b32 s8, 1, 0 ; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: v_and_b32_e32 v0, 1, v0 ; GFX10-NEXT: v_and_b32_e32 v0, 1, v0
; GFX10-NEXT: s_and_b32 s8, s8, 1 ; GFX10-NEXT: s_and_b32 s8, s8, 1
; GFX10-NEXT: v_mov_b32_e32 v3, s31 ; GFX10-NEXT: v_mov_b32_e32 v3, s19
; GFX10-NEXT: s_cmp_lg_u32 s8, 0 ; GFX10-NEXT: s_cmp_lg_u32 s8, 0
; GFX10-NEXT: s_addc_u32 s0, s0, 0 ; GFX10-NEXT: s_addc_u32 s0, s0, 0
; GFX10-NEXT: s_cselect_b32 s8, 1, 0 ; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0
; GFX10-NEXT: s_and_b32 s8, s8, 1 ; GFX10-NEXT: s_and_b32 s8, s8, 1
; GFX10-NEXT: s_cmp_lg_u32 s8, 0 ; GFX10-NEXT: s_cmp_lg_u32 s8, 0
; GFX10-NEXT: s_addc_u32 s1, s1, s23 ; GFX10-NEXT: s_addc_u32 s1, s1, s23
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s2, vcc_lo
; GFX10-NEXT: s_sub_u32 s2, s4, s12 ; GFX10-NEXT: s_sub_u32 s2, s4, s12
; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s3, vcc_lo
; GFX10-NEXT: s_cselect_b32 s3, 1, 0 ; GFX10-NEXT: s_cselect_b32 s3, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, s1, vcc_lo
; GFX10-NEXT: s_and_b32 s3, s3, 1 ; GFX10-NEXT: s_and_b32 s3, s3, 1
; GFX10-NEXT: v_mov_b32_e32 v2, s30 ; GFX10-NEXT: v_mov_b32_e32 v2, s18
; GFX10-NEXT: s_cmp_lg_u32 s3, 0 ; GFX10-NEXT: s_cmp_lg_u32 s3, 0
; GFX10-NEXT: s_subb_u32 s3, s5, s13 ; GFX10-NEXT: s_subb_u32 s3, s5, s13
; GFX10-NEXT: s_cselect_b32 s8, 1, 0 ; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: v_cmp_lt_u64_e64 s1, s[2:3], s[4:5] ; GFX10-NEXT: v_cmp_lt_u64_e64 s1, s[2:3], s[4:5]
; GFX10-NEXT: s_and_b32 s8, s8, 1 ; GFX10-NEXT: s_and_b32 s8, s8, 1
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s0, vcc_lo
; GFX10-NEXT: s_cmp_lg_u32 s8, 0 ; GFX10-NEXT: s_cmp_lg_u32 s8, 0
; GFX10-NEXT: s_subb_u32 s8, s6, s14 ; GFX10-NEXT: s_subb_u32 s8, s6, s14
▲ Show 20 Lines • Show All 106 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.128.ll

Show First 20 Lines • Show All 170 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_lshr_b32 s2, s5, 8 ; GFX10-NEXT: s_lshr_b32 s2, s5, 8
; GFX10-NEXT: s_lshr_b32 s4, s5, 16 ; GFX10-NEXT: s_lshr_b32 s4, s5, 16
; GFX10-NEXT: s_lshr_b32 s8, s5, 24 ; GFX10-NEXT: s_lshr_b32 s8, s5, 24
; GFX10-NEXT: v_mov_b32_e32 v2, s5 ; GFX10-NEXT: v_mov_b32_e32 v2, s5
; GFX10-NEXT: s_lshr_b32 s5, s6, 8 ; GFX10-NEXT: s_lshr_b32 s5, s6, 8
; GFX10-NEXT: s_lshr_b32 s9, s6, 16 ; GFX10-NEXT: s_lshr_b32 s9, s6, 16
; GFX10-NEXT: v_mov_b32_e32 v4, s0 ; GFX10-NEXT: v_mov_b32_e32 v4, s0
; GFX10-NEXT: v_mov_b32_e32 v3, s6 ; GFX10-NEXT: v_mov_b32_e32 v3, s6
; GFX10-NEXT: v_mov_b32_e32 v15, s1 ; GFX10-NEXT: v_mov_b32_e32 v5, s1
; GFX10-NEXT: v_mov_b32_e32 v10, s5 ; GFX10-NEXT: v_mov_b32_e32 v10, s5
; GFX10-NEXT: s_lshr_b32 s0, s6, 24 ; GFX10-NEXT: s_lshr_b32 s0, s6, 24
; GFX10-NEXT: v_mov_b32_e32 v6, s3 ; GFX10-NEXT: v_mov_b32_e32 v6, s3
; GFX10-NEXT: v_mov_b32_e32 v11, s9 ; GFX10-NEXT: v_mov_b32_e32 v11, s9
; GFX10-NEXT: v_mov_b32_e32 v7, s2 ; GFX10-NEXT: v_mov_b32_e32 v7, s2
; GFX10-NEXT: v_mov_b32_e32 v8, s4 ; GFX10-NEXT: v_mov_b32_e32 v8, s4
; GFX10-NEXT: v_mov_b32_e32 v19, s8 ; GFX10-NEXT: v_mov_b32_e32 v9, s8
; GFX10-NEXT: ds_write_b8 v1, v0 ; GFX10-NEXT: ds_write_b8 v1, v0
; GFX10-NEXT: ds_write_b8 v1, v2 offset:4 ; GFX10-NEXT: ds_write_b8 v1, v2 offset:4
; GFX10-NEXT: ds_write_b8 v1, v4 offset:1 ; GFX10-NEXT: ds_write_b8 v1, v4 offset:1
; GFX10-NEXT: ds_write_b8 v1, v15 offset:2 ; GFX10-NEXT: ds_write_b8 v1, v5 offset:2
; GFX10-NEXT: ds_write_b8 v1, v6 offset:3 ; GFX10-NEXT: ds_write_b8 v1, v6 offset:3
; GFX10-NEXT: ds_write_b8 v1, v7 offset:5 ; GFX10-NEXT: ds_write_b8 v1, v7 offset:5
; GFX10-NEXT: ds_write_b8 v1, v8 offset:6 ; GFX10-NEXT: ds_write_b8 v1, v8 offset:6
; GFX10-NEXT: ds_write_b8 v1, v19 offset:7 ; GFX10-NEXT: ds_write_b8 v1, v9 offset:7
; GFX10-NEXT: v_mov_b32_e32 v0, s0 ; GFX10-NEXT: v_mov_b32_e32 v0, s0
; GFX10-NEXT: s_lshr_b32 s0, s7, 8 ; GFX10-NEXT: s_lshr_b32 s0, s7, 8
; GFX10-NEXT: s_lshr_b32 s1, s7, 16 ; GFX10-NEXT: s_lshr_b32 s1, s7, 16
; GFX10-NEXT: v_mov_b32_e32 v2, s7 ; GFX10-NEXT: v_mov_b32_e32 v2, s7
; GFX10-NEXT: ds_write_b8 v1, v3 offset:8 ; GFX10-NEXT: ds_write_b8 v1, v3 offset:8
; GFX10-NEXT: ds_write_b8 v1, v10 offset:9 ; GFX10-NEXT: ds_write_b8 v1, v10 offset:9
; GFX10-NEXT: ds_write_b8 v1, v11 offset:10 ; GFX10-NEXT: ds_write_b8 v1, v11 offset:10
; GFX10-NEXT: s_lshr_b32 s2, s7, 24 ; GFX10-NEXT: s_lshr_b32 s2, s7, 24
; GFX10-NEXT: v_mov_b32_e32 v3, s0 ; GFX10-NEXT: v_mov_b32_e32 v3, s0
; GFX10-NEXT: v_mov_b32_e32 v4, s1 ; GFX10-NEXT: v_mov_b32_e32 v4, s1
; GFX10-NEXT: v_mov_b32_e32 v7, s2 ; GFX10-NEXT: v_mov_b32_e32 v5, s2
; GFX10-NEXT: ds_write_b8 v1, v0 offset:11 ; GFX10-NEXT: ds_write_b8 v1, v0 offset:11
; GFX10-NEXT: ds_write_b8 v1, v2 offset:12 ; GFX10-NEXT: ds_write_b8 v1, v2 offset:12
; GFX10-NEXT: ds_write_b8 v1, v3 offset:13 ; GFX10-NEXT: ds_write_b8 v1, v3 offset:13
; GFX10-NEXT: ds_write_b8 v1, v4 offset:14 ; GFX10-NEXT: ds_write_b8 v1, v4 offset:14
; GFX10-NEXT: ds_write_b8 v1, v7 offset:15 ; GFX10-NEXT: ds_write_b8 v1, v5 offset:15
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
store <4 x i32> %x, <4 x i32> addrspace(3)* %out, align 1 store <4 x i32> %x, <4 x i32> addrspace(3)* %out, align 1
ret void ret void
} }
define amdgpu_kernel void @store_lds_v4i32_align2(<4 x i32> addrspace(3)* %out, <4 x i32> %x) { define amdgpu_kernel void @store_lds_v4i32_align2(<4 x i32> addrspace(3)* %out, <4 x i32> %x) {
; GFX9-LABEL: store_lds_v4i32_align2: ; GFX9-LABEL: store_lds_v4i32_align2:
; GFX9: ; %bb.0: ; GFX9: ; %bb.0:
▲ Show 20 Lines • Show All 62 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_mov_b32_e32 v1, s2 ; GFX10-NEXT: v_mov_b32_e32 v1, s2
; GFX10-NEXT: v_mov_b32_e32 v2, s5 ; GFX10-NEXT: v_mov_b32_e32 v2, s5
; GFX10-NEXT: s_lshr_b32 s0, s4, 16 ; GFX10-NEXT: s_lshr_b32 s0, s4, 16
; GFX10-NEXT: v_mov_b32_e32 v3, s6 ; GFX10-NEXT: v_mov_b32_e32 v3, s6
; GFX10-NEXT: s_lshr_b32 s1, s5, 16 ; GFX10-NEXT: s_lshr_b32 s1, s5, 16
; GFX10-NEXT: s_lshr_b32 s2, s6, 16 ; GFX10-NEXT: s_lshr_b32 s2, s6, 16
; GFX10-NEXT: s_lshr_b32 s3, s7, 16 ; GFX10-NEXT: s_lshr_b32 s3, s7, 16
; GFX10-NEXT: v_mov_b32_e32 v4, s7 ; GFX10-NEXT: v_mov_b32_e32 v4, s7
; GFX10-NEXT: v_mov_b32_e32 v11, s0 ; GFX10-NEXT: v_mov_b32_e32 v5, s0
; GFX10-NEXT: v_mov_b32_e32 v6, s1 ; GFX10-NEXT: v_mov_b32_e32 v6, s1
; GFX10-NEXT: v_mov_b32_e32 v7, s2 ; GFX10-NEXT: v_mov_b32_e32 v7, s2
; GFX10-NEXT: v_mov_b32_e32 v8, s3 ; GFX10-NEXT: v_mov_b32_e32 v8, s3
; GFX10-NEXT: ds_write_b16 v1, v0 ; GFX10-NEXT: ds_write_b16 v1, v0
; GFX10-NEXT: ds_write_b16 v1, v2 offset:4 ; GFX10-NEXT: ds_write_b16 v1, v2 offset:4
; GFX10-NEXT: ds_write_b16 v1, v3 offset:8 ; GFX10-NEXT: ds_write_b16 v1, v3 offset:8
; GFX10-NEXT: ds_write_b16 v1, v4 offset:12 ; GFX10-NEXT: ds_write_b16 v1, v4 offset:12
; GFX10-NEXT: ds_write_b16 v1, v11 offset:2 ; GFX10-NEXT: ds_write_b16 v1, v5 offset:2
; GFX10-NEXT: ds_write_b16 v1, v6 offset:6 ; GFX10-NEXT: ds_write_b16 v1, v6 offset:6
; GFX10-NEXT: ds_write_b16 v1, v7 offset:10 ; GFX10-NEXT: ds_write_b16 v1, v7 offset:10
; GFX10-NEXT: ds_write_b16 v1, v8 offset:14 ; GFX10-NEXT: ds_write_b16 v1, v8 offset:14
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
store <4 x i32> %x, <4 x i32> addrspace(3)* %out, align 2 store <4 x i32> %x, <4 x i32> addrspace(3)* %out, align 2
ret void ret void
} }
▲ Show 20 Lines • Show All 137 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.96.ll

Show First 20 Lines • Show All 141 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_mov_b32_e32 v1, s2 ; GFX10-NEXT: v_mov_b32_e32 v1, s2
; GFX10-NEXT: s_lshr_b32 s5, s13, 24 ; GFX10-NEXT: s_lshr_b32 s5, s13, 24
; GFX10-NEXT: v_mov_b32_e32 v0, s12 ; GFX10-NEXT: v_mov_b32_e32 v0, s12
; GFX10-NEXT: s_lshr_b32 s1, s12, 16 ; GFX10-NEXT: s_lshr_b32 s1, s12, 16
; GFX10-NEXT: v_mov_b32_e32 v2, s13 ; GFX10-NEXT: v_mov_b32_e32 v2, s13
; GFX10-NEXT: s_lshr_b32 s3, s12, 24 ; GFX10-NEXT: s_lshr_b32 s3, s12, 24
; GFX10-NEXT: s_lshr_b32 s6, s14, 8 ; GFX10-NEXT: s_lshr_b32 s6, s14, 8
; GFX10-NEXT: v_mov_b32_e32 v4, s0 ; GFX10-NEXT: v_mov_b32_e32 v4, s0
; GFX10-NEXT: v_mov_b32_e32 v15, s5 ; GFX10-NEXT: v_mov_b32_e32 v9, s5
; GFX10-NEXT: s_lshr_b32 s2, s13, 8 ; GFX10-NEXT: s_lshr_b32 s2, s13, 8
; GFX10-NEXT: s_lshr_b32 s4, s13, 16 ; GFX10-NEXT: s_lshr_b32 s4, s13, 16
; GFX10-NEXT: s_lshr_b32 s7, s14, 16 ; GFX10-NEXT: s_lshr_b32 s7, s14, 16
; GFX10-NEXT: v_mov_b32_e32 v3, s14 ; GFX10-NEXT: v_mov_b32_e32 v3, s14
; GFX10-NEXT: v_mov_b32_e32 v11, s1 ; GFX10-NEXT: v_mov_b32_e32 v5, s1
; GFX10-NEXT: s_lshr_b32 s8, s14, 24 ; GFX10-NEXT: s_lshr_b32 s8, s14, 24
; GFX10-NEXT: v_mov_b32_e32 v6, s3 ; GFX10-NEXT: v_mov_b32_e32 v6, s3
; GFX10-NEXT: v_mov_b32_e32 v10, s6 ; GFX10-NEXT: v_mov_b32_e32 v10, s6
; GFX10-NEXT: v_mov_b32_e32 v7, s2 ; GFX10-NEXT: v_mov_b32_e32 v7, s2
; GFX10-NEXT: v_mov_b32_e32 v8, s4 ; GFX10-NEXT: v_mov_b32_e32 v8, s4
; GFX10-NEXT: ds_write_b8 v1, v0 ; GFX10-NEXT: ds_write_b8 v1, v0
; GFX10-NEXT: ds_write_b8 v1, v2 offset:4 ; GFX10-NEXT: ds_write_b8 v1, v2 offset:4
; GFX10-NEXT: ds_write_b8 v1, v4 offset:1 ; GFX10-NEXT: ds_write_b8 v1, v4 offset:1
; GFX10-NEXT: ds_write_b8 v1, v11 offset:2 ; GFX10-NEXT: ds_write_b8 v1, v5 offset:2
; GFX10-NEXT: ds_write_b8 v1, v6 offset:3 ; GFX10-NEXT: ds_write_b8 v1, v6 offset:3
; GFX10-NEXT: ds_write_b8 v1, v7 offset:5 ; GFX10-NEXT: ds_write_b8 v1, v7 offset:5
; GFX10-NEXT: ds_write_b8 v1, v8 offset:6 ; GFX10-NEXT: ds_write_b8 v1, v8 offset:6
; GFX10-NEXT: v_mov_b32_e32 v0, s7 ; GFX10-NEXT: v_mov_b32_e32 v0, s7
; GFX10-NEXT: v_mov_b32_e32 v2, s8 ; GFX10-NEXT: v_mov_b32_e32 v2, s8
; GFX10-NEXT: ds_write_b8 v1, v15 offset:7 ; GFX10-NEXT: ds_write_b8 v1, v9 offset:7
; GFX10-NEXT: ds_write_b8 v1, v3 offset:8 ; GFX10-NEXT: ds_write_b8 v1, v3 offset:8
; GFX10-NEXT: ds_write_b8 v1, v10 offset:9 ; GFX10-NEXT: ds_write_b8 v1, v10 offset:9
; GFX10-NEXT: ds_write_b8 v1, v0 offset:10 ; GFX10-NEXT: ds_write_b8 v1, v0 offset:10
; GFX10-NEXT: ds_write_b8 v1, v2 offset:11 ; GFX10-NEXT: ds_write_b8 v1, v2 offset:11
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
store <3 x i32> %x, <3 x i32> addrspace(3)* %out, align 1 store <3 x i32> %x, <3 x i32> addrspace(3)* %out, align 1
ret void ret void
} }
▲ Show 20 Lines • Show All 55 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_mov_b32_e32 v0, s12 ; GFX10-NEXT: v_mov_b32_e32 v0, s12
; GFX10-NEXT: v_mov_b32_e32 v1, s2 ; GFX10-NEXT: v_mov_b32_e32 v1, s2
; GFX10-NEXT: s_lshr_b32 s0, s12, 16 ; GFX10-NEXT: s_lshr_b32 s0, s12, 16
; GFX10-NEXT: v_mov_b32_e32 v2, s13 ; GFX10-NEXT: v_mov_b32_e32 v2, s13
; GFX10-NEXT: s_lshr_b32 s1, s13, 16 ; GFX10-NEXT: s_lshr_b32 s1, s13, 16
; GFX10-NEXT: v_mov_b32_e32 v3, s14 ; GFX10-NEXT: v_mov_b32_e32 v3, s14
; GFX10-NEXT: s_lshr_b32 s2, s14, 16 ; GFX10-NEXT: s_lshr_b32 s2, s14, 16
; GFX10-NEXT: v_mov_b32_e32 v4, s0 ; GFX10-NEXT: v_mov_b32_e32 v4, s0
; GFX10-NEXT: v_mov_b32_e32 v7, s1 ; GFX10-NEXT: v_mov_b32_e32 v5, s1
; GFX10-NEXT: v_mov_b32_e32 v6, s2 ; GFX10-NEXT: v_mov_b32_e32 v6, s2
; GFX10-NEXT: ds_write_b16 v1, v0 ; GFX10-NEXT: ds_write_b16 v1, v0
; GFX10-NEXT: ds_write_b16 v1, v2 offset:4 ; GFX10-NEXT: ds_write_b16 v1, v2 offset:4
; GFX10-NEXT: ds_write_b16 v1, v3 offset:8 ; GFX10-NEXT: ds_write_b16 v1, v3 offset:8
; GFX10-NEXT: ds_write_b16 v1, v4 offset:2 ; GFX10-NEXT: ds_write_b16 v1, v4 offset:2
; GFX10-NEXT: ds_write_b16 v1, v7 offset:6 ; GFX10-NEXT: ds_write_b16 v1, v5 offset:6
; GFX10-NEXT: ds_write_b16 v1, v6 offset:10 ; GFX10-NEXT: ds_write_b16 v1, v6 offset:10
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
store <3 x i32> %x, <3 x i32> addrspace(3)* %out, align 2 store <3 x i32> %x, <3 x i32> addrspace(3)* %out, align 2
ret void ret void
} }
define amdgpu_kernel void @store_lds_v3i32_align4(<3 x i32> addrspace(3)* %out, <3 x i32> %x) { define amdgpu_kernel void @store_lds_v3i32_align4(<3 x i32> addrspace(3)* %out, <3 x i32> %x) {
; GFX9-LABEL: store_lds_v3i32_align4: ; GFX9-LABEL: store_lds_v3i32_align4:
▲ Show 20 Lines • Show All 129 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/uaddsat.ll

Show First 20 Lines • Show All 2,813 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e64 v2, v2, -1, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v2, v2, -1, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v3, v3, -1, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v3, v3, -1, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_uaddsat_v2i64: ; GFX10-LABEL: v_uaddsat_v2i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v10, v4 ; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT: v_mov_b32_e32 v11, v5 ; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v5, vcc_lo
; GFX10-NEXT: v_mov_b32_e32 v15, v6 ; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v2, v6
; GFX10-NEXT: v_mov_b32_e32 v16, v7 ; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, v3, v7, vcc_lo
; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v10 ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[0:1], v[4:5]
; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v11, vcc_lo ; GFX10-NEXT: v_cmp_lt_u64_e64 s4, v[2:3], v[6:7]
; GFX10-NEXT: v_add_co_u32 v5, vcc_lo, v2, v15
; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, v3, v16, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[0:1], v[10:11]
; GFX10-NEXT: v_cmp_lt_u64_e64 s4, v[5:6], v[15:16]
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, -1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, -1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, -1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, -1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v5, -1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, -1, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, v6, -1, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, -1, s4
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i64> @llvm.uadd.sat.v2i64(<2 x i64> %lhs, <2 x i64> %rhs) %result = call <2 x i64> @llvm.uadd.sat.v2i64(<2 x i64> %lhs, <2 x i64> %rhs)
ret <2 x i64> %result ret <2 x i64> %result
} }
define amdgpu_ps <2 x i64> @s_uaddsat_v2i64(<2 x i64> inreg %lhs, <2 x i64> inreg %rhs) { define amdgpu_ps <2 x i64> @s_uaddsat_v2i64(<2 x i64> inreg %lhs, <2 x i64> inreg %rhs) {
; GFX6-LABEL: s_uaddsat_v2i64: ; GFX6-LABEL: s_uaddsat_v2i64:
; GFX6: ; %bb.0: ; GFX6: ; %bb.0:
▲ Show 20 Lines • Show All 354 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e64 v0, v4, -1, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v0, v4, -1, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v1, v5, -1, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v1, v5, -1, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v2, v6, -1, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v2, v6, -1, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v3, v7, -1, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v3, v7, -1, vcc
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: uaddsat_i128_sv: ; GFX10-LABEL: uaddsat_i128_sv:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: v_add_co_u32 v10, vcc_lo, s0, v0 ; GFX10-NEXT: v_add_co_u32 v4, vcc_lo, s0, v0
; GFX10-NEXT: v_add_co_ci_u32_e32 v11, vcc_lo, s1, v1, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, s1, v1, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v4, vcc_lo, s2, v2, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, s2, v2, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, s3, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, s3, v3, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[10:11], v[0:1] ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[4:5], v[0:1]
; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[4:5], v[2:3] ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[6:7], v[2:3]
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[4:5], v[2:3] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[6:7], v[2:3]
; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GFX10-NEXT: v_and_b32_e32 v0, 1, v0 ; GFX10-NEXT: v_and_b32_e32 v0, 1, v0
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v10, -1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, -1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v11, -1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, -1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v4, -1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, -1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v3, v5, -1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v3, v7, -1, vcc_lo
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.uadd.sat.i128(i128 %lhs, i128 %rhs) %result = call i128 @llvm.uadd.sat.i128(i128 %lhs, i128 %rhs)
%cast = bitcast i128 %result to <4 x float> %cast = bitcast i128 %result to <4 x float>
ret <4 x float> %cast ret <4 x float> %cast
} }
define amdgpu_ps <4 x float> @uaddsat_i128_vs(i128 %lhs, i128 inreg %rhs) { define amdgpu_ps <4 x float> @uaddsat_i128_vs(i128 %lhs, i128 inreg %rhs) {
; GFX6-LABEL: uaddsat_i128_vs: ; GFX6-LABEL: uaddsat_i128_vs:
▲ Show 20 Lines • Show All 200 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e64 v6, v6, -1, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v6, v6, -1, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v7, v7, -1, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v7, v7, -1, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_uaddsat_v2i128: ; GFX10-LABEL: v_uaddsat_v2i128:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v18, v8 ; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v8
; GFX10-NEXT: v_mov_b32_e32 v19, v9 ; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v9, vcc_lo
; GFX10-NEXT: v_mov_b32_e32 v16, v10 ; GFX10-NEXT: v_add_co_ci_u32_e32 v2, vcc_lo, v2, v10, vcc_lo
; GFX10-NEXT: v_mov_b32_e32 v17, v11 ; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, v3, v11, vcc_lo
; GFX10-NEXT: v_mov_b32_e32 v10, v12 ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[0:1], v[8:9]
; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v18
; GFX10-NEXT: v_mov_b32_e32 v11, v13
; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v19, vcc_lo
; GFX10-NEXT: v_mov_b32_e32 v20, v14
; GFX10-NEXT: v_add_co_ci_u32_e32 v2, vcc_lo, v2, v16, vcc_lo
; GFX10-NEXT: v_mov_b32_e32 v21, v15
; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, v3, v17, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[0:1], v[18:19]
; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, 1, vcc_lo
; GFX10-NEXT: v_add_co_u32 v4, vcc_lo, v4, v10 ; GFX10-NEXT: v_add_co_u32 v4, vcc_lo, v4, v12
; GFX10-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, v5, v11, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, v5, v13, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, v6, v20, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, v6, v14, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, v7, v21, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, v7, v15, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[16:17] ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[10:11]
; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[4:5], v[10:11] ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[4:5], v[12:13]
; GFX10-NEXT: v_cndmask_b32_e64 v12, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v12, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[6:7], v[20:21] ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[6:7], v[14:15]
; GFX10-NEXT: v_cndmask_b32_e64 v13, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v13, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[16:17] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[10:11]
; GFX10-NEXT: v_cndmask_b32_e32 v8, v9, v8, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v8, v9, v8, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[6:7], v[20:21] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[6:7], v[14:15]
; GFX10-NEXT: v_and_b32_e32 v8, 1, v8 ; GFX10-NEXT: v_and_b32_e32 v8, 1, v8
; GFX10-NEXT: v_cndmask_b32_e32 v9, v13, v12, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v9, v13, v12, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v8 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v8
; GFX10-NEXT: v_and_b32_e32 v9, 1, v9 ; GFX10-NEXT: v_and_b32_e32 v9, 1, v9
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, -1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, -1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, -1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, -1, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, v9 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, v9
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, -1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, -1, vcc_lo
▲ Show 20 Lines • Show All 359 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/usubsat.ll

Show First 20 Lines • Show All 2,683 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e64 v2, v4, 0, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v2, v4, 0, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v3, v5, 0, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v3, v5, 0, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_usubsat_v2i64: ; GFX10-LABEL: v_usubsat_v2i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v10, v0 ; GFX10-NEXT: v_sub_co_u32 v8, vcc_lo, v0, v4
; GFX10-NEXT: v_mov_b32_e32 v11, v1 ; GFX10-NEXT: v_sub_co_ci_u32_e32 v9, vcc_lo, v1, v5, vcc_lo
; GFX10-NEXT: v_mov_b32_e32 v0, v2 ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[0:1], v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v1, v3 ; GFX10-NEXT: v_sub_co_u32 v4, s4, v2, v6
; GFX10-NEXT: v_sub_co_u32 v8, vcc_lo, v10, v4 ; GFX10-NEXT: v_sub_co_ci_u32_e64 v5, s4, v3, v7, s4
; GFX10-NEXT: v_sub_co_ci_u32_e32 v9, vcc_lo, v11, v5, vcc_lo ; GFX10-NEXT: v_cmp_lt_u64_e64 s4, v[2:3], v[6:7]
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[10:11], v[4:5]
; GFX10-NEXT: v_sub_co_u32 v4, s4, v0, v6
; GFX10-NEXT: v_sub_co_ci_u32_e64 v5, s4, v1, v7, s4
; GFX10-NEXT: v_cmp_lt_u64_e64 s4, v[0:1], v[6:7]
; GFX10-NEXT: v_cndmask_b32_e64 v0, v8, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, v8, 0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v9, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, v9, 0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v4, 0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v4, 0, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, v5, 0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v5, 0, s4
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i64> @llvm.usub.sat.v2i64(<2 x i64> %lhs, <2 x i64> %rhs) %result = call <2 x i64> @llvm.usub.sat.v2i64(<2 x i64> %lhs, <2 x i64> %rhs)
ret <2 x i64> %result ret <2 x i64> %result
} }
▲ Show 20 Lines • Show All 259 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_cmp_lt_u64_e64 s0, s[0:1], s[4:5] ; GFX10-NEXT: v_cmp_lt_u64_e64 s0, s[0:1], s[4:5]
; GFX10-NEXT: s_and_b32 s9, s9, 1 ; GFX10-NEXT: s_and_b32 s9, s9, 1
; GFX10-NEXT: s_cmp_lg_u32 s9, 0 ; GFX10-NEXT: s_cmp_lg_u32 s9, 0
; GFX10-NEXT: s_subb_u32 s9, s1, s5 ; GFX10-NEXT: s_subb_u32 s9, s1, s5
; GFX10-NEXT: s_cselect_b32 s10, 1, 0 ; GFX10-NEXT: s_cselect_b32 s10, 1, 0
; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0
; GFX10-NEXT: s_and_b32 s10, s10, 1 ; GFX10-NEXT: s_and_b32 s10, s10, 1
; GFX10-NEXT: s_cmp_lg_u32 s10, 0 ; GFX10-NEXT: s_cmp_lg_u32 s10, 0
; GFX10-NEXT: s_subb_u32 s14, s2, s6 ; GFX10-NEXT: s_subb_u32 s10, s2, s6
; GFX10-NEXT: s_cselect_b32 s11, 1, 0 ; GFX10-NEXT: s_cselect_b32 s11, 1, 0
; GFX10-NEXT: s_and_b32 s11, s11, 1 ; GFX10-NEXT: s_and_b32 s11, s11, 1
; GFX10-NEXT: s_cmp_lg_u32 s11, 0 ; GFX10-NEXT: s_cmp_lg_u32 s11, 0
; GFX10-NEXT: s_subb_u32 s1, s3, s7 ; GFX10-NEXT: s_subb_u32 s1, s3, s7
; GFX10-NEXT: s_cmp_eq_u64 s[2:3], s[6:7] ; GFX10-NEXT: s_cmp_eq_u64 s[2:3], s[6:7]
; GFX10-NEXT: v_cmp_lt_u64_e64 s2, s[2:3], s[6:7] ; GFX10-NEXT: v_cmp_lt_u64_e64 s2, s[2:3], s[6:7]
; GFX10-NEXT: s_cselect_b32 s0, 1, 0 ; GFX10-NEXT: s_cselect_b32 s0, 1, 0
; GFX10-NEXT: s_and_b32 s0, 1, s0 ; GFX10-NEXT: s_and_b32 s0, 1, s0
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0 ; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s2
; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GFX10-NEXT: v_and_b32_e32 v0, 1, v0 ; GFX10-NEXT: v_and_b32_e32 v0, 1, v0
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0
; GFX10-NEXT: v_cndmask_b32_e64 v0, s8, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, s8, 0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, s14, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, s10, 0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v3, s1, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v3, s1, 0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, s9, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, s9, 0, vcc_lo
; GFX10-NEXT: v_readfirstlane_b32 s0, v0 ; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: v_readfirstlane_b32 s2, v2 ; GFX10-NEXT: v_readfirstlane_b32 s2, v2
; GFX10-NEXT: v_readfirstlane_b32 s3, v3 ; GFX10-NEXT: v_readfirstlane_b32 s3, v3
; GFX10-NEXT: v_readfirstlane_b32 s1, v1 ; GFX10-NEXT: v_readfirstlane_b32 s1, v1
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.usub.sat.i128(i128 %lhs, i128 %rhs) %result = call i128 @llvm.usub.sat.i128(i128 %lhs, i128 %rhs)
▲ Show 20 Lines • Show All 299 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e64 v6, v10, 0, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v6, v10, 0, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v7, v11, 0, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v7, v11, 0, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_usubsat_v2i128: ; GFX10-LABEL: v_usubsat_v2i128:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v22, v0 ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[0:1], v[8:9]
; GFX10-NEXT: v_mov_b32_e32 v23, v1 ; GFX10-NEXT: v_cmp_eq_u64_e64 s5, v[6:7], v[14:15]
; GFX10-NEXT: v_mov_b32_e32 v20, v2
; GFX10-NEXT: v_mov_b32_e32 v21, v3
; GFX10-NEXT: v_mov_b32_e32 v26, v4
; GFX10-NEXT: v_mov_b32_e32 v27, v5
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[22:23], v[8:9]
; GFX10-NEXT: v_mov_b32_e32 v24, v6
; GFX10-NEXT: v_mov_b32_e32 v25, v7
; GFX10-NEXT: v_cndmask_b32_e64 v16, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v16, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[20:21], v[10:11] ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[10:11]
; GFX10-NEXT: v_cmp_eq_u64_e64 s5, v[24:25], v[14:15]
; GFX10-NEXT: v_cndmask_b32_e64 v17, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v17, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[20:21], v[10:11] ; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[10:11]
; GFX10-NEXT: v_cndmask_b32_e32 v16, v17, v16, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v16, v17, v16, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[26:27], v[12:13] ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[4:5], v[12:13]
; GFX10-NEXT: v_and_b32_e32 v16, 1, v16 ; GFX10-NEXT: v_and_b32_e32 v16, 1, v16
; GFX10-NEXT: v_cndmask_b32_e64 v17, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v17, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[24:25], v[14:15] ; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[6:7], v[14:15]
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, v16 ; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, v16
; GFX10-NEXT: v_cndmask_b32_e64 v18, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v18, 0, 1, vcc_lo
; GFX10-NEXT: v_sub_co_u32 v0, vcc_lo, v22, v8 ; GFX10-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v8
; GFX10-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v23, v9, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v9, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v8, v18, v17, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v8, v18, v17, s5
; GFX10-NEXT: v_sub_co_ci_u32_e32 v2, vcc_lo, v20, v10, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v2, vcc_lo, v2, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, 0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, 0, s4
; GFX10-NEXT: v_sub_co_ci_u32_e32 v3, vcc_lo, v21, v11, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v3, vcc_lo, v3, v11, vcc_lo
; GFX10-NEXT: v_sub_co_u32 v4, vcc_lo, v26, v12 ; GFX10-NEXT: v_sub_co_u32 v4, vcc_lo, v4, v12
; GFX10-NEXT: v_and_b32_e32 v8, 1, v8 ; GFX10-NEXT: v_and_b32_e32 v8, 1, v8
; GFX10-NEXT: v_sub_co_ci_u32_e32 v5, vcc_lo, v27, v13, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v5, vcc_lo, v5, v13, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, 0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, 0, s4
; GFX10-NEXT: v_sub_co_ci_u32_e32 v6, vcc_lo, v24, v14, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v6, vcc_lo, v6, v14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, 0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, 0, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, v8 ; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, v8
; GFX10-NEXT: v_sub_co_ci_u32_e32 v7, vcc_lo, v25, v15, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v7, vcc_lo, v7, v15, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, 0, s4 ; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, 0, s4
; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, 0, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v4, v4, 0, s5
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, 0, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, 0, s5
; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, 0, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, 0, s5
; GFX10-NEXT: v_cndmask_b32_e64 v7, v7, 0, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v7, v7, 0, s5
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i128> @llvm.usub.sat.v2i128(<2 x i128> %lhs, <2 x i128> %rhs) %result = call <2 x i128> @llvm.usub.sat.v2i128(<2 x i128> %lhs, <2 x i128> %rhs)
ret <2 x i128> %result ret <2 x i128> %result
▲ Show 20 Lines • Show All 274 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s2 ; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s2
; GFX10-NEXT: s_cmp_lg_u32 s1, 0 ; GFX10-NEXT: s_cmp_lg_u32 s1, 0
; GFX10-NEXT: s_subb_u32 s3, s5, s13 ; GFX10-NEXT: s_subb_u32 s3, s5, s13
; GFX10-NEXT: s_cselect_b32 s1, 1, 0 ; GFX10-NEXT: s_cselect_b32 s1, 1, 0
; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; GFX10-NEXT: s_and_b32 s1, s1, 1 ; GFX10-NEXT: s_and_b32 s1, s1, 1
; GFX10-NEXT: s_cmp_lg_u32 s1, 0 ; GFX10-NEXT: s_cmp_lg_u32 s1, 0
; GFX10-NEXT: v_cmp_lt_u64_e64 s1, s[4:5], s[12:13] ; GFX10-NEXT: v_cmp_lt_u64_e64 s1, s[4:5], s[12:13]
; GFX10-NEXT: s_subb_u32 s30, s6, s14 ; GFX10-NEXT: s_subb_u32 s10, s6, s14
; GFX10-NEXT: s_cselect_b32 s0, 1, 0 ; GFX10-NEXT: s_cselect_b32 s0, 1, 0
; GFX10-NEXT: v_and_b32_e32 v0, 1, v0 ; GFX10-NEXT: v_and_b32_e32 v0, 1, v0
; GFX10-NEXT: s_and_b32 s0, s0, 1 ; GFX10-NEXT: s_and_b32 s0, s0, 1
; GFX10-NEXT: s_cmp_lg_u32 s0, 0 ; GFX10-NEXT: s_cmp_lg_u32 s0, 0
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s1
; GFX10-NEXT: s_subb_u32 s9, s7, s15 ; GFX10-NEXT: s_subb_u32 s9, s7, s15
; GFX10-NEXT: s_cmp_eq_u64 s[6:7], s[14:15] ; GFX10-NEXT: s_cmp_eq_u64 s[6:7], s[14:15]
; GFX10-NEXT: v_cmp_lt_u64_e64 s1, s[6:7], s[14:15] ; GFX10-NEXT: v_cmp_lt_u64_e64 s1, s[6:7], s[14:15]
Show All 9 Lines
; GFX10-NEXT: v_cndmask_b32_e64 v3, s18, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v3, s18, 0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v4, s19, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v4, s19, 0, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0
; GFX10-NEXT: v_readfirstlane_b32 s0, v1 ; GFX10-NEXT: v_readfirstlane_b32 s0, v1
; GFX10-NEXT: v_readfirstlane_b32 s1, v2 ; GFX10-NEXT: v_readfirstlane_b32 s1, v2
; GFX10-NEXT: v_readfirstlane_b32 s2, v3 ; GFX10-NEXT: v_readfirstlane_b32 s2, v3
; GFX10-NEXT: v_cndmask_b32_e64 v1, s3, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v1, s3, 0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, s8, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v0, s8, 0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, s30, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, s10, 0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v3, s9, 0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v3, s9, 0, vcc_lo
; GFX10-NEXT: v_readfirstlane_b32 s3, v4 ; GFX10-NEXT: v_readfirstlane_b32 s3, v4
; GFX10-NEXT: v_readfirstlane_b32 s5, v1 ; GFX10-NEXT: v_readfirstlane_b32 s5, v1
; GFX10-NEXT: v_readfirstlane_b32 s4, v0 ; GFX10-NEXT: v_readfirstlane_b32 s4, v0
; GFX10-NEXT: v_readfirstlane_b32 s6, v2 ; GFX10-NEXT: v_readfirstlane_b32 s6, v2
; GFX10-NEXT: v_readfirstlane_b32 s7, v3 ; GFX10-NEXT: v_readfirstlane_b32 s7, v3
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
%result = call <2 x i128> @llvm.usub.sat.v2i128(<2 x i128> %lhs, <2 x i128> %rhs) %result = call <2 x i128> @llvm.usub.sat.v2i128(<2 x i128> %lhs, <2 x i128> %rhs)
Show All 34 Lines

llvm/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll

Show First 20 Lines • Show All 494 Lines • ▼ Show 20 Lines
; GFX1064-NEXT: v_writelane_b32 v3, s6, 48 ; GFX1064-NEXT: v_writelane_b32 v3, s6, 48
; GFX1064-NEXT: s_mov_b64 exec, s[4:5] ; GFX1064-NEXT: s_mov_b64 exec, s[4:5]
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 ; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT: s_mov_b32 s2, -1 ; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: ; implicit-def: $vgpr0 ; GFX1064-NEXT: ; implicit-def: $vgpr0
; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc ; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX1064-NEXT: s_cbranch_execz BB2_2 ; GFX1064-NEXT: s_cbranch_execz BB2_2
; GFX1064-NEXT: ; %bb.1: ; GFX1064-NEXT: ; %bb.1:
; GFX1064-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1064-NEXT: v_mov_b32_e32 v4, s7 ; GFX1064-NEXT: v_mov_b32_e32 v4, s7
; GFX1064-NEXT: s_mov_b32 s3, s7 ; GFX1064-NEXT: s_mov_b32 s3, s7
; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1064-NEXT: ds_add_rtn_u32 v0, v7, v4 ; GFX1064-NEXT: ds_add_rtn_u32 v0, v0, v4
; GFX1064-NEXT: s_waitcnt lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: buffer_gl0_inv ; GFX1064-NEXT: buffer_gl0_inv
; GFX1064-NEXT: BB2_2: ; GFX1064-NEXT: BB2_2:
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] ; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1064-NEXT: v_readfirstlane_b32 s3, v0
; GFX1064-NEXT: v_mov_b32_e32 v0, v3 ; GFX1064-NEXT: v_mov_b32_e32 v0, v3
; GFX1064-NEXT: v_add_nc_u32_e32 v0, s3, v0 ; GFX1064-NEXT: v_add_nc_u32_e32 v0, s3, v0
Show All 29 Lines
; GFX1032-NEXT: v_writelane_b32 v3, s3, 16 ; GFX1032-NEXT: v_writelane_b32 v3, s3, 16
; GFX1032-NEXT: s_mov_b32 exec_lo, s2 ; GFX1032-NEXT: s_mov_b32 exec_lo, s2
; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 ; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT: s_mov_b32 s2, -1 ; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: ; implicit-def: $vgpr0 ; GFX1032-NEXT: ; implicit-def: $vgpr0
; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo ; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo
; GFX1032-NEXT: s_cbranch_execz BB2_2 ; GFX1032-NEXT: s_cbranch_execz BB2_2
; GFX1032-NEXT: ; %bb.1: ; GFX1032-NEXT: ; %bb.1:
; GFX1032-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1032-NEXT: v_mov_b32_e32 v4, s4 ; GFX1032-NEXT: v_mov_b32_e32 v4, s4
; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1032-NEXT: ds_add_rtn_u32 v0, v7, v4 ; GFX1032-NEXT: ds_add_rtn_u32 v0, v0, v4
; GFX1032-NEXT: s_waitcnt lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032-NEXT: buffer_gl0_inv ; GFX1032-NEXT: buffer_gl0_inv
; GFX1032-NEXT: BB2_2: ; GFX1032-NEXT: BB2_2:
; GFX1032-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1032-NEXT: s_waitcnt_depctr 0xffe3
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3 ; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3
; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1032-NEXT: v_readfirstlane_b32 s3, v0
; GFX1032-NEXT: v_mov_b32_e32 v0, v3 ; GFX1032-NEXT: v_mov_b32_e32 v0, v3
; GFX1032-NEXT: v_add_nc_u32_e32 v0, s3, v0 ; GFX1032-NEXT: v_add_nc_u32_e32 v0, s3, v0
▲ Show 20 Lines • Show All 1,108 Lines • ▼ Show 20 Lines
; GFX1064-NEXT: v_writelane_b32 v3, s6, 48 ; GFX1064-NEXT: v_writelane_b32 v3, s6, 48
; GFX1064-NEXT: s_mov_b64 exec, s[4:5] ; GFX1064-NEXT: s_mov_b64 exec, s[4:5]
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 ; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT: s_mov_b32 s2, -1 ; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: ; implicit-def: $vgpr0 ; GFX1064-NEXT: ; implicit-def: $vgpr0
; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc ; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX1064-NEXT: s_cbranch_execz BB9_2 ; GFX1064-NEXT: s_cbranch_execz BB9_2
; GFX1064-NEXT: ; %bb.1: ; GFX1064-NEXT: ; %bb.1:
; GFX1064-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1064-NEXT: v_mov_b32_e32 v4, s7 ; GFX1064-NEXT: v_mov_b32_e32 v4, s7
; GFX1064-NEXT: s_mov_b32 s3, s7 ; GFX1064-NEXT: s_mov_b32 s3, s7
; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1064-NEXT: ds_sub_rtn_u32 v0, v7, v4 ; GFX1064-NEXT: ds_sub_rtn_u32 v0, v0, v4
; GFX1064-NEXT: s_waitcnt lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: buffer_gl0_inv ; GFX1064-NEXT: buffer_gl0_inv
; GFX1064-NEXT: BB9_2: ; GFX1064-NEXT: BB9_2:
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] ; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1064-NEXT: v_readfirstlane_b32 s3, v0
; GFX1064-NEXT: v_mov_b32_e32 v0, v3 ; GFX1064-NEXT: v_mov_b32_e32 v0, v3
; GFX1064-NEXT: v_sub_nc_u32_e32 v0, s3, v0 ; GFX1064-NEXT: v_sub_nc_u32_e32 v0, s3, v0
Show All 29 Lines
; GFX1032-NEXT: v_writelane_b32 v3, s3, 16 ; GFX1032-NEXT: v_writelane_b32 v3, s3, 16
; GFX1032-NEXT: s_mov_b32 exec_lo, s2 ; GFX1032-NEXT: s_mov_b32 exec_lo, s2
; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 ; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT: s_mov_b32 s2, -1 ; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: ; implicit-def: $vgpr0 ; GFX1032-NEXT: ; implicit-def: $vgpr0
; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo ; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo
; GFX1032-NEXT: s_cbranch_execz BB9_2 ; GFX1032-NEXT: s_cbranch_execz BB9_2
; GFX1032-NEXT: ; %bb.1: ; GFX1032-NEXT: ; %bb.1:
; GFX1032-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1032-NEXT: v_mov_b32_e32 v4, s4 ; GFX1032-NEXT: v_mov_b32_e32 v4, s4
; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1032-NEXT: ds_sub_rtn_u32 v0, v7, v4 ; GFX1032-NEXT: ds_sub_rtn_u32 v0, v0, v4
; GFX1032-NEXT: s_waitcnt lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032-NEXT: buffer_gl0_inv ; GFX1032-NEXT: buffer_gl0_inv
; GFX1032-NEXT: BB9_2: ; GFX1032-NEXT: BB9_2:
; GFX1032-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1032-NEXT: s_waitcnt_depctr 0xffe3
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3 ; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3
; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1032-NEXT: v_readfirstlane_b32 s3, v0
; GFX1032-NEXT: v_mov_b32_e32 v0, v3 ; GFX1032-NEXT: v_mov_b32_e32 v0, v3
; GFX1032-NEXT: v_sub_nc_u32_e32 v0, s3, v0 ; GFX1032-NEXT: v_sub_nc_u32_e32 v0, s3, v0
▲ Show 20 Lines • Show All 782 Lines • ▼ Show 20 Lines
; GFX1064-NEXT: v_writelane_b32 v3, s6, 48 ; GFX1064-NEXT: v_writelane_b32 v3, s6, 48
; GFX1064-NEXT: s_mov_b64 exec, s[4:5] ; GFX1064-NEXT: s_mov_b64 exec, s[4:5]
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 ; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT: s_mov_b32 s2, -1 ; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: ; implicit-def: $vgpr0 ; GFX1064-NEXT: ; implicit-def: $vgpr0
; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc ; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX1064-NEXT: s_cbranch_execz BB14_2 ; GFX1064-NEXT: s_cbranch_execz BB14_2
; GFX1064-NEXT: ; %bb.1: ; GFX1064-NEXT: ; %bb.1:
; GFX1064-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1064-NEXT: v_mov_b32_e32 v4, s7 ; GFX1064-NEXT: v_mov_b32_e32 v4, s7
; GFX1064-NEXT: s_mov_b32 s3, s7 ; GFX1064-NEXT: s_mov_b32 s3, s7
; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1064-NEXT: ds_and_rtn_b32 v0, v7, v4 ; GFX1064-NEXT: ds_and_rtn_b32 v0, v0, v4
; GFX1064-NEXT: s_waitcnt lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: buffer_gl0_inv ; GFX1064-NEXT: buffer_gl0_inv
; GFX1064-NEXT: BB14_2: ; GFX1064-NEXT: BB14_2:
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] ; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1064-NEXT: v_readfirstlane_b32 s3, v0
; GFX1064-NEXT: v_mov_b32_e32 v0, v3 ; GFX1064-NEXT: v_mov_b32_e32 v0, v3
; GFX1064-NEXT: v_and_b32_e32 v0, s3, v0 ; GFX1064-NEXT: v_and_b32_e32 v0, s3, v0
Show All 29 Lines
; GFX1032-NEXT: v_writelane_b32 v3, s3, 16 ; GFX1032-NEXT: v_writelane_b32 v3, s3, 16
; GFX1032-NEXT: s_mov_b32 exec_lo, s2 ; GFX1032-NEXT: s_mov_b32 exec_lo, s2
; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 ; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT: s_mov_b32 s2, -1 ; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: ; implicit-def: $vgpr0 ; GFX1032-NEXT: ; implicit-def: $vgpr0
; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo ; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo
; GFX1032-NEXT: s_cbranch_execz BB14_2 ; GFX1032-NEXT: s_cbranch_execz BB14_2
; GFX1032-NEXT: ; %bb.1: ; GFX1032-NEXT: ; %bb.1:
; GFX1032-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1032-NEXT: v_mov_b32_e32 v4, s4 ; GFX1032-NEXT: v_mov_b32_e32 v4, s4
; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1032-NEXT: ds_and_rtn_b32 v0, v7, v4 ; GFX1032-NEXT: ds_and_rtn_b32 v0, v0, v4
; GFX1032-NEXT: s_waitcnt lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032-NEXT: buffer_gl0_inv ; GFX1032-NEXT: buffer_gl0_inv
; GFX1032-NEXT: BB14_2: ; GFX1032-NEXT: BB14_2:
; GFX1032-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1032-NEXT: s_waitcnt_depctr 0xffe3
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3 ; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3
; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1032-NEXT: v_readfirstlane_b32 s3, v0
; GFX1032-NEXT: v_mov_b32_e32 v0, v3 ; GFX1032-NEXT: v_mov_b32_e32 v0, v3
; GFX1032-NEXT: v_and_b32_e32 v0, s3, v0 ; GFX1032-NEXT: v_and_b32_e32 v0, s3, v0
▲ Show 20 Lines • Show All 162 Lines • ▼ Show 20 Lines
; GFX1064-NEXT: v_writelane_b32 v3, s6, 48 ; GFX1064-NEXT: v_writelane_b32 v3, s6, 48
; GFX1064-NEXT: s_mov_b64 exec, s[4:5] ; GFX1064-NEXT: s_mov_b64 exec, s[4:5]
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 ; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT: s_mov_b32 s2, -1 ; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: ; implicit-def: $vgpr0 ; GFX1064-NEXT: ; implicit-def: $vgpr0
; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc ; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX1064-NEXT: s_cbranch_execz BB15_2 ; GFX1064-NEXT: s_cbranch_execz BB15_2
; GFX1064-NEXT: ; %bb.1: ; GFX1064-NEXT: ; %bb.1:
; GFX1064-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1064-NEXT: v_mov_b32_e32 v4, s7 ; GFX1064-NEXT: v_mov_b32_e32 v4, s7
; GFX1064-NEXT: s_mov_b32 s3, s7 ; GFX1064-NEXT: s_mov_b32 s3, s7
; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1064-NEXT: ds_or_rtn_b32 v0, v7, v4 ; GFX1064-NEXT: ds_or_rtn_b32 v0, v0, v4
; GFX1064-NEXT: s_waitcnt lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: buffer_gl0_inv ; GFX1064-NEXT: buffer_gl0_inv
; GFX1064-NEXT: BB15_2: ; GFX1064-NEXT: BB15_2:
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] ; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1064-NEXT: v_readfirstlane_b32 s3, v0
; GFX1064-NEXT: v_mov_b32_e32 v0, v3 ; GFX1064-NEXT: v_mov_b32_e32 v0, v3
; GFX1064-NEXT: v_or_b32_e32 v0, s3, v0 ; GFX1064-NEXT: v_or_b32_e32 v0, s3, v0
Show All 29 Lines
; GFX1032-NEXT: v_writelane_b32 v3, s3, 16 ; GFX1032-NEXT: v_writelane_b32 v3, s3, 16
; GFX1032-NEXT: s_mov_b32 exec_lo, s2 ; GFX1032-NEXT: s_mov_b32 exec_lo, s2
; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 ; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT: s_mov_b32 s2, -1 ; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: ; implicit-def: $vgpr0 ; GFX1032-NEXT: ; implicit-def: $vgpr0
; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo ; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo
; GFX1032-NEXT: s_cbranch_execz BB15_2 ; GFX1032-NEXT: s_cbranch_execz BB15_2
; GFX1032-NEXT: ; %bb.1: ; GFX1032-NEXT: ; %bb.1:
; GFX1032-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1032-NEXT: v_mov_b32_e32 v4, s4 ; GFX1032-NEXT: v_mov_b32_e32 v4, s4
; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1032-NEXT: ds_or_rtn_b32 v0, v7, v4 ; GFX1032-NEXT: ds_or_rtn_b32 v0, v0, v4
; GFX1032-NEXT: s_waitcnt lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032-NEXT: buffer_gl0_inv ; GFX1032-NEXT: buffer_gl0_inv
; GFX1032-NEXT: BB15_2: ; GFX1032-NEXT: BB15_2:
; GFX1032-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1032-NEXT: s_waitcnt_depctr 0xffe3
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3 ; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3
; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1032-NEXT: v_readfirstlane_b32 s3, v0
; GFX1032-NEXT: v_mov_b32_e32 v0, v3 ; GFX1032-NEXT: v_mov_b32_e32 v0, v3
; GFX1032-NEXT: v_or_b32_e32 v0, s3, v0 ; GFX1032-NEXT: v_or_b32_e32 v0, s3, v0
▲ Show 20 Lines • Show All 162 Lines • ▼ Show 20 Lines
; GFX1064-NEXT: v_writelane_b32 v3, s6, 48 ; GFX1064-NEXT: v_writelane_b32 v3, s6, 48
; GFX1064-NEXT: s_mov_b64 exec, s[4:5] ; GFX1064-NEXT: s_mov_b64 exec, s[4:5]
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 ; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT: s_mov_b32 s2, -1 ; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: ; implicit-def: $vgpr0 ; GFX1064-NEXT: ; implicit-def: $vgpr0
; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc ; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX1064-NEXT: s_cbranch_execz BB16_2 ; GFX1064-NEXT: s_cbranch_execz BB16_2
; GFX1064-NEXT: ; %bb.1: ; GFX1064-NEXT: ; %bb.1:
; GFX1064-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1064-NEXT: v_mov_b32_e32 v4, s7 ; GFX1064-NEXT: v_mov_b32_e32 v4, s7
; GFX1064-NEXT: s_mov_b32 s3, s7 ; GFX1064-NEXT: s_mov_b32 s3, s7
; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1064-NEXT: ds_xor_rtn_b32 v0, v7, v4 ; GFX1064-NEXT: ds_xor_rtn_b32 v0, v0, v4
; GFX1064-NEXT: s_waitcnt lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: buffer_gl0_inv ; GFX1064-NEXT: buffer_gl0_inv
; GFX1064-NEXT: BB16_2: ; GFX1064-NEXT: BB16_2:
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] ; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1064-NEXT: v_readfirstlane_b32 s3, v0
; GFX1064-NEXT: v_mov_b32_e32 v0, v3 ; GFX1064-NEXT: v_mov_b32_e32 v0, v3
; GFX1064-NEXT: v_xor_b32_e32 v0, s3, v0 ; GFX1064-NEXT: v_xor_b32_e32 v0, s3, v0
Show All 29 Lines
; GFX1032-NEXT: v_writelane_b32 v3, s3, 16 ; GFX1032-NEXT: v_writelane_b32 v3, s3, 16
; GFX1032-NEXT: s_mov_b32 exec_lo, s2 ; GFX1032-NEXT: s_mov_b32 exec_lo, s2
; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 ; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT: s_mov_b32 s2, -1 ; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: ; implicit-def: $vgpr0 ; GFX1032-NEXT: ; implicit-def: $vgpr0
; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo ; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo
; GFX1032-NEXT: s_cbranch_execz BB16_2 ; GFX1032-NEXT: s_cbranch_execz BB16_2
; GFX1032-NEXT: ; %bb.1: ; GFX1032-NEXT: ; %bb.1:
; GFX1032-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1032-NEXT: v_mov_b32_e32 v4, s4 ; GFX1032-NEXT: v_mov_b32_e32 v4, s4
; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1032-NEXT: ds_xor_rtn_b32 v0, v7, v4 ; GFX1032-NEXT: ds_xor_rtn_b32 v0, v0, v4
; GFX1032-NEXT: s_waitcnt lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032-NEXT: buffer_gl0_inv ; GFX1032-NEXT: buffer_gl0_inv
; GFX1032-NEXT: BB16_2: ; GFX1032-NEXT: BB16_2:
; GFX1032-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1032-NEXT: s_waitcnt_depctr 0xffe3
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3 ; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3
; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1032-NEXT: v_readfirstlane_b32 s3, v0
; GFX1032-NEXT: v_mov_b32_e32 v0, v3 ; GFX1032-NEXT: v_mov_b32_e32 v0, v3
; GFX1032-NEXT: v_xor_b32_e32 v0, s3, v0 ; GFX1032-NEXT: v_xor_b32_e32 v0, s3, v0
▲ Show 20 Lines • Show All 164 Lines • ▼ Show 20 Lines
; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 ; GFX1064-NEXT: v_writelane_b32 v1, s6, 48
; GFX1064-NEXT: s_mov_b64 exec, s[4:5] ; GFX1064-NEXT: s_mov_b64 exec, s[4:5]
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 ; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT: s_mov_b32 s2, -1 ; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: ; implicit-def: $vgpr0 ; GFX1064-NEXT: ; implicit-def: $vgpr0
; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc ; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX1064-NEXT: s_cbranch_execz BB17_2 ; GFX1064-NEXT: s_cbranch_execz BB17_2
; GFX1064-NEXT: ; %bb.1: ; GFX1064-NEXT: ; %bb.1:
; GFX1064-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1064-NEXT: v_mov_b32_e32 v4, s7 ; GFX1064-NEXT: v_mov_b32_e32 v4, s7
; GFX1064-NEXT: s_mov_b32 s3, s7 ; GFX1064-NEXT: s_mov_b32 s3, s7
; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1064-NEXT: ds_max_rtn_i32 v0, v7, v4 ; GFX1064-NEXT: ds_max_rtn_i32 v0, v0, v4
; GFX1064-NEXT: s_waitcnt lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: buffer_gl0_inv ; GFX1064-NEXT: buffer_gl0_inv
; GFX1064-NEXT: BB17_2: ; GFX1064-NEXT: BB17_2:
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] ; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1064-NEXT: v_readfirstlane_b32 s3, v0
; GFX1064-NEXT: v_mov_b32_e32 v0, v1 ; GFX1064-NEXT: v_mov_b32_e32 v0, v1
; GFX1064-NEXT: v_max_i32_e32 v0, s3, v0 ; GFX1064-NEXT: v_max_i32_e32 v0, s3, v0
Show All 31 Lines
; GFX1032-NEXT: v_writelane_b32 v1, s3, 16 ; GFX1032-NEXT: v_writelane_b32 v1, s3, 16
; GFX1032-NEXT: s_mov_b32 exec_lo, s2 ; GFX1032-NEXT: s_mov_b32 exec_lo, s2
; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 ; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT: s_mov_b32 s2, -1 ; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: ; implicit-def: $vgpr0 ; GFX1032-NEXT: ; implicit-def: $vgpr0
; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo ; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo
; GFX1032-NEXT: s_cbranch_execz BB17_2 ; GFX1032-NEXT: s_cbranch_execz BB17_2
; GFX1032-NEXT: ; %bb.1: ; GFX1032-NEXT: ; %bb.1:
; GFX1032-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1032-NEXT: v_mov_b32_e32 v4, s4 ; GFX1032-NEXT: v_mov_b32_e32 v4, s4
; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1032-NEXT: ds_max_rtn_i32 v0, v7, v4 ; GFX1032-NEXT: ds_max_rtn_i32 v0, v0, v4
; GFX1032-NEXT: s_waitcnt lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032-NEXT: buffer_gl0_inv ; GFX1032-NEXT: buffer_gl0_inv
; GFX1032-NEXT: BB17_2: ; GFX1032-NEXT: BB17_2:
; GFX1032-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1032-NEXT: s_waitcnt_depctr 0xffe3
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3 ; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3
; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1032-NEXT: v_readfirstlane_b32 s3, v0
; GFX1032-NEXT: v_mov_b32_e32 v0, v1 ; GFX1032-NEXT: v_mov_b32_e32 v0, v1
; GFX1032-NEXT: v_max_i32_e32 v0, s3, v0 ; GFX1032-NEXT: v_max_i32_e32 v0, s3, v0
▲ Show 20 Lines • Show All 343 Lines • ▼ Show 20 Lines
; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 ; GFX1064-NEXT: v_writelane_b32 v1, s6, 48
; GFX1064-NEXT: s_mov_b64 exec, s[4:5] ; GFX1064-NEXT: s_mov_b64 exec, s[4:5]
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 ; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT: s_mov_b32 s2, -1 ; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: ; implicit-def: $vgpr0 ; GFX1064-NEXT: ; implicit-def: $vgpr0
; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc ; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX1064-NEXT: s_cbranch_execz BB19_2 ; GFX1064-NEXT: s_cbranch_execz BB19_2
; GFX1064-NEXT: ; %bb.1: ; GFX1064-NEXT: ; %bb.1:
; GFX1064-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1064-NEXT: v_mov_b32_e32 v4, s7 ; GFX1064-NEXT: v_mov_b32_e32 v4, s7
; GFX1064-NEXT: s_mov_b32 s3, s7 ; GFX1064-NEXT: s_mov_b32 s3, s7
; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1064-NEXT: ds_min_rtn_i32 v0, v7, v4 ; GFX1064-NEXT: ds_min_rtn_i32 v0, v0, v4
; GFX1064-NEXT: s_waitcnt lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: buffer_gl0_inv ; GFX1064-NEXT: buffer_gl0_inv
; GFX1064-NEXT: BB19_2: ; GFX1064-NEXT: BB19_2:
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] ; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1064-NEXT: v_readfirstlane_b32 s3, v0
; GFX1064-NEXT: v_mov_b32_e32 v0, v1 ; GFX1064-NEXT: v_mov_b32_e32 v0, v1
; GFX1064-NEXT: v_min_i32_e32 v0, s3, v0 ; GFX1064-NEXT: v_min_i32_e32 v0, s3, v0
Show All 31 Lines
; GFX1032-NEXT: v_writelane_b32 v1, s3, 16 ; GFX1032-NEXT: v_writelane_b32 v1, s3, 16
; GFX1032-NEXT: s_mov_b32 exec_lo, s2 ; GFX1032-NEXT: s_mov_b32 exec_lo, s2
; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 ; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT: s_mov_b32 s2, -1 ; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: ; implicit-def: $vgpr0 ; GFX1032-NEXT: ; implicit-def: $vgpr0
; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo ; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo
; GFX1032-NEXT: s_cbranch_execz BB19_2 ; GFX1032-NEXT: s_cbranch_execz BB19_2
; GFX1032-NEXT: ; %bb.1: ; GFX1032-NEXT: ; %bb.1:
; GFX1032-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1032-NEXT: v_mov_b32_e32 v4, s4 ; GFX1032-NEXT: v_mov_b32_e32 v4, s4
; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1032-NEXT: ds_min_rtn_i32 v0, v7, v4 ; GFX1032-NEXT: ds_min_rtn_i32 v0, v0, v4
; GFX1032-NEXT: s_waitcnt lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032-NEXT: buffer_gl0_inv ; GFX1032-NEXT: buffer_gl0_inv
; GFX1032-NEXT: BB19_2: ; GFX1032-NEXT: BB19_2:
; GFX1032-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1032-NEXT: s_waitcnt_depctr 0xffe3
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3 ; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3
; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1032-NEXT: v_readfirstlane_b32 s3, v0
; GFX1032-NEXT: v_mov_b32_e32 v0, v1 ; GFX1032-NEXT: v_mov_b32_e32 v0, v1
; GFX1032-NEXT: v_min_i32_e32 v0, s3, v0 ; GFX1032-NEXT: v_min_i32_e32 v0, s3, v0
▲ Show 20 Lines • Show All 341 Lines • ▼ Show 20 Lines
; GFX1064-NEXT: v_writelane_b32 v3, s6, 48 ; GFX1064-NEXT: v_writelane_b32 v3, s6, 48
; GFX1064-NEXT: s_mov_b64 exec, s[4:5] ; GFX1064-NEXT: s_mov_b64 exec, s[4:5]
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 ; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT: s_mov_b32 s2, -1 ; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: ; implicit-def: $vgpr0 ; GFX1064-NEXT: ; implicit-def: $vgpr0
; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc ; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX1064-NEXT: s_cbranch_execz BB21_2 ; GFX1064-NEXT: s_cbranch_execz BB21_2
; GFX1064-NEXT: ; %bb.1: ; GFX1064-NEXT: ; %bb.1:
; GFX1064-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1064-NEXT: v_mov_b32_e32 v4, s7 ; GFX1064-NEXT: v_mov_b32_e32 v4, s7
; GFX1064-NEXT: s_mov_b32 s3, s7 ; GFX1064-NEXT: s_mov_b32 s3, s7
; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1064-NEXT: ds_max_rtn_u32 v0, v7, v4 ; GFX1064-NEXT: ds_max_rtn_u32 v0, v0, v4
; GFX1064-NEXT: s_waitcnt lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: buffer_gl0_inv ; GFX1064-NEXT: buffer_gl0_inv
; GFX1064-NEXT: BB21_2: ; GFX1064-NEXT: BB21_2:
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] ; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1064-NEXT: v_readfirstlane_b32 s3, v0
; GFX1064-NEXT: v_mov_b32_e32 v0, v3 ; GFX1064-NEXT: v_mov_b32_e32 v0, v3
; GFX1064-NEXT: v_max_u32_e32 v0, s3, v0 ; GFX1064-NEXT: v_max_u32_e32 v0, s3, v0
Show All 29 Lines
; GFX1032-NEXT: v_writelane_b32 v3, s3, 16 ; GFX1032-NEXT: v_writelane_b32 v3, s3, 16
; GFX1032-NEXT: s_mov_b32 exec_lo, s2 ; GFX1032-NEXT: s_mov_b32 exec_lo, s2
; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 ; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT: s_mov_b32 s2, -1 ; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: ; implicit-def: $vgpr0 ; GFX1032-NEXT: ; implicit-def: $vgpr0
; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo ; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo
; GFX1032-NEXT: s_cbranch_execz BB21_2 ; GFX1032-NEXT: s_cbranch_execz BB21_2
; GFX1032-NEXT: ; %bb.1: ; GFX1032-NEXT: ; %bb.1:
; GFX1032-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1032-NEXT: v_mov_b32_e32 v4, s4 ; GFX1032-NEXT: v_mov_b32_e32 v4, s4
; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1032-NEXT: ds_max_rtn_u32 v0, v7, v4 ; GFX1032-NEXT: ds_max_rtn_u32 v0, v0, v4
; GFX1032-NEXT: s_waitcnt lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032-NEXT: buffer_gl0_inv ; GFX1032-NEXT: buffer_gl0_inv
; GFX1032-NEXT: BB21_2: ; GFX1032-NEXT: BB21_2:
; GFX1032-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1032-NEXT: s_waitcnt_depctr 0xffe3
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3 ; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3
; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1032-NEXT: v_readfirstlane_b32 s3, v0
; GFX1032-NEXT: v_mov_b32_e32 v0, v3 ; GFX1032-NEXT: v_mov_b32_e32 v0, v3
; GFX1032-NEXT: v_max_u32_e32 v0, s3, v0 ; GFX1032-NEXT: v_max_u32_e32 v0, s3, v0
▲ Show 20 Lines • Show All 338 Lines • ▼ Show 20 Lines
; GFX1064-NEXT: v_writelane_b32 v3, s6, 48 ; GFX1064-NEXT: v_writelane_b32 v3, s6, 48
; GFX1064-NEXT: s_mov_b64 exec, s[4:5] ; GFX1064-NEXT: s_mov_b64 exec, s[4:5]
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 ; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT: s_mov_b32 s2, -1 ; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: ; implicit-def: $vgpr0 ; GFX1064-NEXT: ; implicit-def: $vgpr0
; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc ; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX1064-NEXT: s_cbranch_execz BB23_2 ; GFX1064-NEXT: s_cbranch_execz BB23_2
; GFX1064-NEXT: ; %bb.1: ; GFX1064-NEXT: ; %bb.1:
; GFX1064-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1064-NEXT: v_mov_b32_e32 v4, s7 ; GFX1064-NEXT: v_mov_b32_e32 v4, s7
; GFX1064-NEXT: s_mov_b32 s3, s7 ; GFX1064-NEXT: s_mov_b32 s3, s7
; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1064-NEXT: ds_min_rtn_u32 v0, v7, v4 ; GFX1064-NEXT: ds_min_rtn_u32 v0, v0, v4
; GFX1064-NEXT: s_waitcnt lgkmcnt(0) ; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: buffer_gl0_inv ; GFX1064-NEXT: buffer_gl0_inv
; GFX1064-NEXT: BB23_2: ; GFX1064-NEXT: BB23_2:
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] ; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1064-NEXT: v_readfirstlane_b32 s3, v0
; GFX1064-NEXT: v_mov_b32_e32 v0, v3 ; GFX1064-NEXT: v_mov_b32_e32 v0, v3
; GFX1064-NEXT: v_min_u32_e32 v0, s3, v0 ; GFX1064-NEXT: v_min_u32_e32 v0, s3, v0
Show All 29 Lines
; GFX1032-NEXT: v_writelane_b32 v3, s3, 16 ; GFX1032-NEXT: v_writelane_b32 v3, s3, 16
; GFX1032-NEXT: s_mov_b32 exec_lo, s2 ; GFX1032-NEXT: s_mov_b32 exec_lo, s2
; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 ; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT: s_mov_b32 s2, -1 ; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: ; implicit-def: $vgpr0 ; GFX1032-NEXT: ; implicit-def: $vgpr0
; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo ; GFX1032-NEXT: s_and_saveexec_b32 s3, vcc_lo
; GFX1032-NEXT: s_cbranch_execz BB23_2 ; GFX1032-NEXT: s_cbranch_execz BB23_2
; GFX1032-NEXT: ; %bb.1: ; GFX1032-NEXT: ; %bb.1:
; GFX1032-NEXT: v_mov_b32_e32 v7, local_var32@abs32@lo ; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo
; GFX1032-NEXT: v_mov_b32_e32 v4, s4 ; GFX1032-NEXT: v_mov_b32_e32 v4, s4
; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0
; GFX1032-NEXT: ds_min_rtn_u32 v0, v7, v4 ; GFX1032-NEXT: ds_min_rtn_u32 v0, v0, v4
; GFX1032-NEXT: s_waitcnt lgkmcnt(0) ; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032-NEXT: buffer_gl0_inv ; GFX1032-NEXT: buffer_gl0_inv
; GFX1032-NEXT: BB23_2: ; GFX1032-NEXT: BB23_2:
; GFX1032-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1032-NEXT: s_waitcnt_depctr 0xffe3
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3 ; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3
; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 ; GFX1032-NEXT: v_readfirstlane_b32 s3, v0
; GFX1032-NEXT: v_mov_b32_e32 v0, v3 ; GFX1032-NEXT: v_mov_b32_e32 v0, v3
; GFX1032-NEXT: v_min_u32_e32 v0, s3, v0 ; GFX1032-NEXT: v_min_u32_e32 v0, s3, v0
▲ Show 20 Lines • Show All 186 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/atomic_optimizations_pixelshader.ll

Show First 20 Lines • Show All 86 Lines • ▼ Show 20 Lines
; GFX1064-NEXT: s_and_saveexec_b64 s[8:9], s[10:11] ; GFX1064-NEXT: s_and_saveexec_b64 s[8:9], s[10:11]
; GFX1064-NEXT: s_cbranch_execz BB0_4 ; GFX1064-NEXT: s_cbranch_execz BB0_4
; GFX1064-NEXT: ; %bb.1: ; GFX1064-NEXT: ; %bb.1:
; GFX1064-NEXT: s_mov_b64 s[12:13], exec ; GFX1064-NEXT: s_mov_b64 s[12:13], exec
; GFX1064-NEXT: ; implicit-def: $vgpr1 ; GFX1064-NEXT: ; implicit-def: $vgpr1
; GFX1064-NEXT: v_mbcnt_lo_u32_b32 v0, s12, 0 ; GFX1064-NEXT: v_mbcnt_lo_u32_b32 v0, s12, 0
; GFX1064-NEXT: v_mbcnt_hi_u32_b32 v0, s13, v0 ; GFX1064-NEXT: v_mbcnt_hi_u32_b32 v0, s13, v0
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 ; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT: s_and_saveexec_b64 s[28:29], vcc ; GFX1064-NEXT: s_and_saveexec_b64 s[10:11], vcc
; GFX1064-NEXT: s_cbranch_execz BB0_3 ; GFX1064-NEXT: s_cbranch_execz BB0_3
; GFX1064-NEXT: ; %bb.2: ; GFX1064-NEXT: ; %bb.2:
; GFX1064-NEXT: s_bcnt1_i32_b64 s12, s[12:13] ; GFX1064-NEXT: s_bcnt1_i32_b64 s12, s[12:13]
; GFX1064-NEXT: s_mul_i32 s12, s12, 5 ; GFX1064-NEXT: s_mul_i32 s12, s12, 5
; GFX1064-NEXT: v_mov_b32_e32 v1, s12 ; GFX1064-NEXT: v_mov_b32_e32 v1, s12
; GFX1064-NEXT: buffer_atomic_add v1, off, s[4:7], 0 glc ; GFX1064-NEXT: buffer_atomic_add v1, off, s[4:7], 0 glc
; GFX1064-NEXT: BB0_3: ; GFX1064-NEXT: BB0_3:
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
; GFX1064-NEXT: s_or_b64 exec, exec, s[28:29] ; GFX1064-NEXT: s_or_b64 exec, exec, s[10:11]
; GFX1064-NEXT: s_waitcnt vmcnt(0) ; GFX1064-NEXT: s_waitcnt vmcnt(0)
; GFX1064-NEXT: v_readfirstlane_b32 s4, v1 ; GFX1064-NEXT: v_readfirstlane_b32 s4, v1
; GFX1064-NEXT: v_mad_u32_u24 v0, v0, 5, s4 ; GFX1064-NEXT: v_mad_u32_u24 v0, v0, 5, s4
; GFX1064-NEXT: BB0_4: ; %Flow ; GFX1064-NEXT: BB0_4: ; %Flow
; GFX1064-NEXT: s_or_b64 exec, exec, s[8:9] ; GFX1064-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1064-NEXT: s_wqm_b64 s[4:5], -1 ; GFX1064-NEXT: s_wqm_b64 s[4:5], -1
; GFX1064-NEXT: s_andn2_b64 vcc, exec, s[4:5] ; GFX1064-NEXT: s_andn2_b64 vcc, exec, s[4:5]
; GFX1064-NEXT: s_cbranch_vccnz BB0_6 ; GFX1064-NEXT: s_cbranch_vccnz BB0_6
▲ Show 20 Lines • Show All 210 Lines • ▼ Show 20 Lines
; GFX1064-NEXT: v_writelane_b32 v3, s13, 32 ; GFX1064-NEXT: v_writelane_b32 v3, s13, 32
; GFX1064-NEXT: s_mov_b64 exec, s[10:11] ; GFX1064-NEXT: s_mov_b64 exec, s[10:11]
; GFX1064-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0 ; GFX1064-NEXT: v_mbcnt_hi_u32_b32 v0, exec_hi, v0
; GFX1064-NEXT: s_or_saveexec_b64 s[10:11], -1 ; GFX1064-NEXT: s_or_saveexec_b64 s[10:11], -1
; GFX1064-NEXT: v_writelane_b32 v3, s14, 48 ; GFX1064-NEXT: v_writelane_b32 v3, s14, 48
; GFX1064-NEXT: s_mov_b64 exec, s[10:11] ; GFX1064-NEXT: s_mov_b64 exec, s[10:11]
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 ; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT: ; implicit-def: $vgpr0 ; GFX1064-NEXT: ; implicit-def: $vgpr0
; GFX1064-NEXT: s_and_saveexec_b64 s[28:29], vcc ; GFX1064-NEXT: s_and_saveexec_b64 s[10:11], vcc
; GFX1064-NEXT: s_cbranch_execz BB1_3 ; GFX1064-NEXT: s_cbranch_execz BB1_3
; GFX1064-NEXT: ; %bb.2: ; GFX1064-NEXT: ; %bb.2:
; GFX1064-NEXT: v_mov_b32_e32 v0, s12 ; GFX1064-NEXT: v_mov_b32_e32 v0, s12
; GFX1064-NEXT: buffer_atomic_add v0, off, s[4:7], 0 glc ; GFX1064-NEXT: buffer_atomic_add v0, off, s[4:7], 0 glc
; GFX1064-NEXT: BB1_3: ; GFX1064-NEXT: BB1_3:
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3 ; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
; GFX1064-NEXT: s_or_b64 exec, exec, s[28:29] ; GFX1064-NEXT: s_or_b64 exec, exec, s[10:11]
; GFX1064-NEXT: s_waitcnt vmcnt(0) ; GFX1064-NEXT: s_waitcnt vmcnt(0)
; GFX1064-NEXT: v_readfirstlane_b32 s4, v0 ; GFX1064-NEXT: v_readfirstlane_b32 s4, v0
; GFX1064-NEXT: v_mov_b32_e32 v0, v3 ; GFX1064-NEXT: v_mov_b32_e32 v0, v3
; GFX1064-NEXT: v_add_nc_u32_e32 v0, s4, v0 ; GFX1064-NEXT: v_add_nc_u32_e32 v0, s4, v0
; GFX1064-NEXT: BB1_4: ; %Flow ; GFX1064-NEXT: BB1_4: ; %Flow
; GFX1064-NEXT: s_or_b64 exec, exec, s[8:9] ; GFX1064-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1064-NEXT: s_wqm_b64 s[4:5], -1 ; GFX1064-NEXT: s_wqm_b64 s[4:5], -1
; GFX1064-NEXT: s_andn2_b64 vcc, exec, s[4:5] ; GFX1064-NEXT: s_andn2_b64 vcc, exec, s[4:5]
▲ Show 20 Lines • Show All 72 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/ctlz.ll

Show First 20 Lines • Show All 227 Lines • ▼ Show 20 Lines
; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.y, ; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.y,
; EG-NEXT: 32(4.484155e-44), 2(2.802597e-45) ; EG-NEXT: 32(4.484155e-44), 2(2.802597e-45)
; ;
; GFX10-LABEL: v_ctlz_v2i32: ; GFX10-LABEL: v_ctlz_v2i32:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x2c ; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x2c
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 ; GFX10-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX10-NEXT: v_mov_b32_e32 v7, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: global_load_dwordx2 v[0:1], v0, s[2:3] ; GFX10-NEXT: global_load_dwordx2 v[0:1], v0, s[2:3]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_ffbh_u32_e32 v2, v1 ; GFX10-NEXT: v_ffbh_u32_e32 v2, v1
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v1 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v1
; GFX10-NEXT: v_ffbh_u32_e32 v3, v0 ; GFX10-NEXT: v_ffbh_u32_e32 v3, v0
; GFX10-NEXT: v_cndmask_b32_e32 v1, 32, v2, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, 32, v2, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0
; GFX10-NEXT: v_cndmask_b32_e32 v0, 32, v3, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, 32, v3, vcc_lo
; GFX10-NEXT: global_store_dwordx2 v7, v[0:1], s[0:1] ; GFX10-NEXT: global_store_dwordx2 v4, v[0:1], s[0:1]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x() %tid = call i32 @llvm.amdgcn.workitem.id.x()
%in.gep = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %valptr, i32 %tid %in.gep = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %valptr, i32 %tid
%val = load <2 x i32>, <2 x i32> addrspace(1)* %in.gep, align 8 %val = load <2 x i32>, <2 x i32> addrspace(1)* %in.gep, align 8
%ctlz = call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> %val, i1 false) nounwind readnone %ctlz = call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> %val, i1 false) nounwind readnone
store <2 x i32> %ctlz, <2 x i32> addrspace(1)* %out, align 8 store <2 x i32> %ctlz, <2 x i32> addrspace(1)* %out, align 8
ret void ret void
} }
▲ Show 20 Lines • Show All 1,191 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/cvt_f32_ubyte.ll

Show First 20 Lines • Show All 1,115 Lines • ▼ Show 20 Lines
; VI-NEXT: s_endpgm ; VI-NEXT: s_endpgm
; ;
; GFX10-LABEL: load_v7i8_to_v7f32: ; GFX10-LABEL: load_v7i8_to_v7f32:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x2c ; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x2c
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX10-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-NEXT: v_mov_b32_e32 v2, 0
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 ; GFX10-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX10-NEXT: v_mov_b32_e32 v11, 0 ; GFX10-NEXT: v_mov_b32_e32 v8, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_clause 0x5 ; GFX10-NEXT: s_clause 0x5
; GFX10-NEXT: global_load_ubyte v1, v0, s[2:3] offset:2 ; GFX10-NEXT: global_load_ubyte v1, v0, s[2:3] offset:2
; GFX10-NEXT: global_load_ubyte v3, v0, s[2:3] offset:3 ; GFX10-NEXT: global_load_ubyte v3, v0, s[2:3] offset:3
; GFX10-NEXT: global_load_short_d16 v2, v0, s[2:3] offset:4 ; GFX10-NEXT: global_load_short_d16 v2, v0, s[2:3] offset:4
; GFX10-NEXT: global_load_ubyte v6, v0, s[2:3] offset:6 ; GFX10-NEXT: global_load_ubyte v6, v0, s[2:3] offset:6
; GFX10-NEXT: global_load_ubyte v4, v0, s[2:3] offset:1 ; GFX10-NEXT: global_load_ubyte v4, v0, s[2:3] offset:1
; GFX10-NEXT: global_load_ubyte v7, v0, s[2:3] ; GFX10-NEXT: global_load_ubyte v7, v0, s[2:3]
; GFX10-NEXT: s_waitcnt vmcnt(4) ; GFX10-NEXT: s_waitcnt vmcnt(4)
; GFX10-NEXT: v_lshl_or_b32 v0, v3, 8, v1 ; GFX10-NEXT: v_lshl_or_b32 v0, v3, 8, v1
; GFX10-NEXT: s_waitcnt vmcnt(3) ; GFX10-NEXT: s_waitcnt vmcnt(3)
; GFX10-NEXT: v_cvt_f32_ubyte1_e32 v5, v2 ; GFX10-NEXT: v_cvt_f32_ubyte1_e32 v5, v2
; GFX10-NEXT: s_waitcnt vmcnt(2) ; GFX10-NEXT: s_waitcnt vmcnt(2)
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v6, v6 ; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v6, v6
; GFX10-NEXT: s_waitcnt vmcnt(1) ; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_cvt_f32_ubyte2_e32 v1, v4 ; GFX10-NEXT: v_cvt_f32_ubyte2_e32 v1, v4
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v4, v2 ; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v4, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX10-NEXT: v_cvt_f32_ubyte3_e32 v3, v0 ; GFX10-NEXT: v_cvt_f32_ubyte3_e32 v3, v0
; GFX10-NEXT: v_cvt_f32_ubyte2_e32 v2, v0 ; GFX10-NEXT: v_cvt_f32_ubyte2_e32 v2, v0
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v0, v7 ; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v0, v7
; GFX10-NEXT: global_store_dword v11, v6, s[0:1] offset:24 ; GFX10-NEXT: global_store_dword v8, v6, s[0:1] offset:24
; GFX10-NEXT: global_store_dwordx2 v11, v[4:5], s[0:1] offset:16 ; GFX10-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
; GFX10-NEXT: global_store_dwordx4 v11, v[0:3], s[0:1] ; GFX10-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x() %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <7 x i8>, <7 x i8> addrspace(1)* %in, i32 %tid %gep = getelementptr <7 x i8>, <7 x i8> addrspace(1)* %in, i32 %tid
%load = load <7 x i8>, <7 x i8> addrspace(1)* %gep, align 1 %load = load <7 x i8>, <7 x i8> addrspace(1)* %gep, align 1
%cvt = uitofp <7 x i8> %load to <7 x float> %cvt = uitofp <7 x i8> %load to <7 x float>
store <7 x float> %cvt, <7 x float> addrspace(1)* %out, align 16 store <7 x float> %cvt, <7 x float> addrspace(1)* %out, align 16
ret void ret void
} }
▲ Show 20 Lines • Show All 637 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/ds-sub-offset.ll

Show First 20 Lines • Show All 84 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 2, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX10-NEXT: s_mov_b32 vcc_lo, 0 ; GFX10-NEXT: s_mov_b32 vcc_lo, 0
; GFX10-NEXT: v_mov_b32_e32 v2, 0x7b ; GFX10-NEXT: v_mov_b32_e32 v2, 0x7b
; GFX10-NEXT: v_sub_nc_u32_e32 v3, 0, v0 ; GFX10-NEXT: v_sub_nc_u32_e32 v3, 0, v0
; GFX10-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-NEXT: v_mov_b32_e32 v0, 0
; GFX10-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-NEXT: v_mov_b32_e32 v1, 0
; GFX10-NEXT: ds_write_b32 v3, v2 offset:12 ; GFX10-NEXT: ds_write_b32 v3, v2 offset:12
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_div_fmas_f32 v7, s0, s0, s0 ; GFX10-NEXT: v_div_fmas_f32 v4, s0, s0, s0
; GFX10-NEXT: global_store_dword v[0:1], v7, off ; GFX10-NEXT: global_store_dword v[0:1], v4, off
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
entry: entry:
%x.i = call i32 @llvm.amdgcn.workitem.id.x() #1 %x.i = call i32 @llvm.amdgcn.workitem.id.x() #1
%sub1 = sub i32 0, %x.i %sub1 = sub i32 0, %x.i
%tmp0 = getelementptr [256 x i32], [256 x i32] addrspace(3)* @lds.obj, i32 0, i32 %sub1 %tmp0 = getelementptr [256 x i32], [256 x i32] addrspace(3)* @lds.obj, i32 0, i32 %sub1
%arrayidx = getelementptr inbounds i32, i32 addrspace(3)* %tmp0, i32 3 %arrayidx = getelementptr inbounds i32, i32 addrspace(3)* %tmp0, i32 3
store i32 123, i32 addrspace(3)* %arrayidx store i32 123, i32 addrspace(3)* %arrayidx
▲ Show 20 Lines • Show All 232 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_mov_b32 vcc_lo, 0 ; GFX10-NEXT: s_mov_b32 vcc_lo, 0
; GFX10-NEXT: v_mov_b32_e32 v2, 0x7b ; GFX10-NEXT: v_mov_b32_e32 v2, 0x7b
; GFX10-NEXT: v_mov_b32_e32 v3, 0 ; GFX10-NEXT: v_mov_b32_e32 v3, 0
; GFX10-NEXT: v_sub_nc_u32_e32 v4, 0x3fb, v0 ; GFX10-NEXT: v_sub_nc_u32_e32 v4, 0x3fb, v0
; GFX10-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-NEXT: v_mov_b32_e32 v0, 0
; GFX10-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-NEXT: v_mov_b32_e32 v1, 0
; GFX10-NEXT: ds_write2_b32 v4, v2, v3 offset1:1 ; GFX10-NEXT: ds_write2_b32 v4, v2, v3 offset1:1
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_div_fmas_f32 v7, s0, s0, s0 ; GFX10-NEXT: v_div_fmas_f32 v5, s0, s0, s0
; GFX10-NEXT: global_store_dword v[0:1], v7, off ; GFX10-NEXT: global_store_dword v[0:1], v5, off
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%x.i = call i32 @llvm.amdgcn.workitem.id.x() #0 %x.i = call i32 @llvm.amdgcn.workitem.id.x() #0
%neg = sub i32 0, %x.i %neg = sub i32 0, %x.i
%shl = shl i32 %neg, 2 %shl = shl i32 %neg, 2
%add = add i32 1019, %shl %add = add i32 1019, %shl
%ptr = inttoptr i32 %add to i64 addrspace(3)* %ptr = inttoptr i32 %add to i64 addrspace(3)*
store i64 123, i64 addrspace(3)* %ptr, align 4 store i64 123, i64 addrspace(3)* %ptr, align 4
▲ Show 20 Lines • Show All 47 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/fneg-fold-legalize-dag-increase-insts.ll

Show All 14 Lines
; GFX9-NEXT: v_add_f64 v[0:1], v[4:5], -v[2:3] ; GFX9-NEXT: v_add_f64 v[0:1], v[4:5], -v[2:3]
; GFX9-NEXT: v_add_f64 v[2:3], -v[2:3], -v[4:5] ; GFX9-NEXT: v_add_f64 v[2:3], -v[2:3], -v[4:5]
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: testfn: ; GFX10-LABEL: testfn:
; GFX10: ; %bb.0: ; %bb ; GFX10: ; %bb.0: ; %bb
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v7, v5 ; GFX10-NEXT: v_add_f64 v[4:5], v[4:5], -v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v6, v4
; GFX10-NEXT: v_add_f64 v[4:5], v[6:7], -v[0:1]
; GFX10-NEXT: v_add_f64 v[0:1], v[4:5], -v[2:3] ; GFX10-NEXT: v_add_f64 v[0:1], v[4:5], -v[2:3]
; GFX10-NEXT: v_add_f64 v[2:3], -v[2:3], -v[4:5] ; GFX10-NEXT: v_add_f64 v[2:3], -v[2:3], -v[4:5]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
bb: bb:
%tmp = fsub fast double 0.000000e+00, %arg1 %tmp = fsub fast double 0.000000e+00, %arg1
%tmp3 = fsub fast double %arg2, %arg %tmp3 = fsub fast double %arg2, %arg
%tmp4 = fadd fast double %tmp3, %tmp %tmp4 = fadd fast double %tmp3, %tmp
%tmp5 = fsub fast double %tmp, %tmp3 %tmp5 = fsub fast double %tmp, %tmp3
%tmp6 = insertvalue { double, double } undef, double %tmp4, 0 %tmp6 = insertvalue { double, double } undef, double %tmp4, 0
%tmp7 = insertvalue { double, double } %tmp6, double %tmp5, 1 %tmp7 = insertvalue { double, double } %tmp6, double %tmp5, 1
ret { double, double } %tmp7 ret { double, double } %tmp7
} }

llvm/test/CodeGen/AMDGPU/frem.ll

Show First 20 Lines • Show All 964 Lines • ▼ Show 20 Lines
; GFX9-NEXT: global_store_dwordx2 v12, v[0:1], s[4:5] ; GFX9-NEXT: global_store_dwordx2 v12, v[0:1], s[4:5]
; GFX9-NEXT: s_endpgm ; GFX9-NEXT: s_endpgm
; ;
; GFX10-LABEL: frem_f64: ; GFX10-LABEL: frem_f64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24 ; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34 ; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
; GFX10-NEXT: v_mov_b32_e32 v15, 0 ; GFX10-NEXT: v_mov_b32_e32 v12, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx2 v[0:1], v15, s[6:7] ; GFX10-NEXT: global_load_dwordx2 v[0:1], v12, s[6:7]
; GFX10-NEXT: global_load_dwordx2 v[2:3], v15, s[2:3] ; GFX10-NEXT: global_load_dwordx2 v[2:3], v12, s[2:3]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_div_scale_f64 v[4:5], s0, v[2:3], v[2:3], v[0:1] ; GFX10-NEXT: v_div_scale_f64 v[4:5], s0, v[2:3], v[2:3], v[0:1]
; GFX10-NEXT: v_rcp_f64_e32 v[6:7], v[4:5] ; GFX10-NEXT: v_rcp_f64_e32 v[6:7], v[4:5]
; GFX10-NEXT: v_fma_f64 v[8:9], -v[4:5], v[6:7], 1.0 ; GFX10-NEXT: v_fma_f64 v[8:9], -v[4:5], v[6:7], 1.0
; GFX10-NEXT: v_fma_f64 v[6:7], v[6:7], v[8:9], v[6:7] ; GFX10-NEXT: v_fma_f64 v[6:7], v[6:7], v[8:9], v[6:7]
; GFX10-NEXT: v_fma_f64 v[8:9], -v[4:5], v[6:7], 1.0 ; GFX10-NEXT: v_fma_f64 v[8:9], -v[4:5], v[6:7], 1.0
; GFX10-NEXT: v_fma_f64 v[6:7], v[6:7], v[8:9], v[6:7] ; GFX10-NEXT: v_fma_f64 v[6:7], v[6:7], v[8:9], v[6:7]
; GFX10-NEXT: v_div_scale_f64 v[8:9], vcc_lo, v[0:1], v[2:3], v[0:1] ; GFX10-NEXT: v_div_scale_f64 v[8:9], vcc_lo, v[0:1], v[2:3], v[0:1]
; GFX10-NEXT: v_mul_f64 v[10:11], v[8:9], v[6:7] ; GFX10-NEXT: v_mul_f64 v[10:11], v[8:9], v[6:7]
; GFX10-NEXT: v_fma_f64 v[4:5], -v[4:5], v[10:11], v[8:9] ; GFX10-NEXT: v_fma_f64 v[4:5], -v[4:5], v[10:11], v[8:9]
; GFX10-NEXT: v_div_fmas_f64 v[4:5], v[4:5], v[6:7], v[10:11] ; GFX10-NEXT: v_div_fmas_f64 v[4:5], v[4:5], v[6:7], v[10:11]
; GFX10-NEXT: v_div_fixup_f64 v[4:5], v[4:5], v[2:3], v[0:1] ; GFX10-NEXT: v_div_fixup_f64 v[4:5], v[4:5], v[2:3], v[0:1]
; GFX10-NEXT: v_trunc_f64_e32 v[4:5], v[4:5] ; GFX10-NEXT: v_trunc_f64_e32 v[4:5], v[4:5]
; GFX10-NEXT: v_fma_f64 v[0:1], -v[4:5], v[2:3], v[0:1] ; GFX10-NEXT: v_fma_f64 v[0:1], -v[4:5], v[2:3], v[0:1]
; GFX10-NEXT: global_store_dwordx2 v15, v[0:1], s[4:5] ; GFX10-NEXT: global_store_dwordx2 v12, v[0:1], s[4:5]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
double addrspace(1)* %in2) #0 { double addrspace(1)* %in2) #0 {
%r0 = load double, double addrspace(1)* %in1, align 8 %r0 = load double, double addrspace(1)* %in1, align 8
%r1 = load double, double addrspace(1)* %in2, align 8 %r1 = load double, double addrspace(1)* %in2, align 8
%r2 = frem double %r0, %r1 %r2 = frem double %r0, %r1
store double %r2, double addrspace(1)* %out, align 8 store double %r2, double addrspace(1)* %out, align 8
ret void ret void
} }
▲ Show 20 Lines • Show All 135 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx2 v[0:1], v10, s[6:7] ; GFX10-NEXT: global_load_dwordx2 v[0:1], v10, s[6:7]
; GFX10-NEXT: global_load_dwordx2 v[2:3], v10, s[2:3] ; GFX10-NEXT: global_load_dwordx2 v[2:3], v10, s[2:3]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_rcp_f64_e32 v[4:5], v[2:3] ; GFX10-NEXT: v_rcp_f64_e32 v[4:5], v[2:3]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0 ; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5] ; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0 ; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0
; GFX10-NEXT: v_fma_f64 v[14:15], v[6:7], v[4:5], v[4:5] ; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5]
; GFX10-NEXT: v_mul_f64 v[6:7], v[0:1], v[14:15] ; GFX10-NEXT: v_mul_f64 v[6:7], v[0:1], v[4:5]
; GFX10-NEXT: v_fma_f64 v[8:9], -v[2:3], v[6:7], v[0:1] ; GFX10-NEXT: v_fma_f64 v[8:9], -v[2:3], v[6:7], v[0:1]
; GFX10-NEXT: v_fma_f64 v[4:5], v[8:9], v[14:15], v[6:7] ; GFX10-NEXT: v_fma_f64 v[4:5], v[8:9], v[4:5], v[6:7]
; GFX10-NEXT: v_trunc_f64_e32 v[4:5], v[4:5] ; GFX10-NEXT: v_trunc_f64_e32 v[4:5], v[4:5]
; GFX10-NEXT: v_fma_f64 v[0:1], -v[4:5], v[2:3], v[0:1] ; GFX10-NEXT: v_fma_f64 v[0:1], -v[4:5], v[2:3], v[0:1]
; GFX10-NEXT: global_store_dwordx2 v10, v[0:1], s[4:5] ; GFX10-NEXT: global_store_dwordx2 v10, v[0:1], s[4:5]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
double addrspace(1)* %in2) #0 { double addrspace(1)* %in2) #0 {
%r0 = load double, double addrspace(1)* %in1, align 8 %r0 = load double, double addrspace(1)* %in1, align 8
%r1 = load double, double addrspace(1)* %in2, align 8 %r1 = load double, double addrspace(1)* %in2, align 8
%r2 = frem fast double %r0, %r1 %r2 = frem fast double %r0, %r1
▲ Show 20 Lines • Show All 138 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx2 v[0:1], v10, s[6:7] ; GFX10-NEXT: global_load_dwordx2 v[0:1], v10, s[6:7]
; GFX10-NEXT: global_load_dwordx2 v[2:3], v10, s[2:3] ; GFX10-NEXT: global_load_dwordx2 v[2:3], v10, s[2:3]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_rcp_f64_e32 v[4:5], v[2:3] ; GFX10-NEXT: v_rcp_f64_e32 v[4:5], v[2:3]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0 ; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0
; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5] ; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0 ; GFX10-NEXT: v_fma_f64 v[6:7], -v[2:3], v[4:5], 1.0
; GFX10-NEXT: v_fma_f64 v[14:15], v[6:7], v[4:5], v[4:5] ; GFX10-NEXT: v_fma_f64 v[4:5], v[6:7], v[4:5], v[4:5]
; GFX10-NEXT: v_mul_f64 v[6:7], v[0:1], v[14:15] ; GFX10-NEXT: v_mul_f64 v[6:7], v[0:1], v[4:5]
; GFX10-NEXT: v_fma_f64 v[8:9], -v[2:3], v[6:7], v[0:1] ; GFX10-NEXT: v_fma_f64 v[8:9], -v[2:3], v[6:7], v[0:1]
; GFX10-NEXT: v_fma_f64 v[4:5], v[8:9], v[14:15], v[6:7] ; GFX10-NEXT: v_fma_f64 v[4:5], v[8:9], v[4:5], v[6:7]
; GFX10-NEXT: v_trunc_f64_e32 v[4:5], v[4:5] ; GFX10-NEXT: v_trunc_f64_e32 v[4:5], v[4:5]
; GFX10-NEXT: v_fma_f64 v[0:1], -v[4:5], v[2:3], v[0:1] ; GFX10-NEXT: v_fma_f64 v[0:1], -v[4:5], v[2:3], v[0:1]
; GFX10-NEXT: global_store_dwordx2 v10, v[0:1], s[4:5] ; GFX10-NEXT: global_store_dwordx2 v10, v[0:1], s[4:5]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
double addrspace(1)* %in2) #1 { double addrspace(1)* %in2) #1 {
%r0 = load double, double addrspace(1)* %in1, align 8 %r0 = load double, double addrspace(1)* %in1, align 8
%r1 = load double, double addrspace(1)* %in2, align 8 %r1 = load double, double addrspace(1)* %in2, align 8
%r2 = frem afn double %r0, %r1 %r2 = frem afn double %r0, %r1
▲ Show 20 Lines • Show All 574 Lines • ▼ Show 20 Lines
; GFX9-NEXT: global_store_dwordx2 v4, v[0:1], s[4:5] ; GFX9-NEXT: global_store_dwordx2 v4, v[0:1], s[4:5]
; GFX9-NEXT: s_endpgm ; GFX9-NEXT: s_endpgm
; ;
; GFX10-LABEL: frem_v4f16: ; GFX10-LABEL: frem_v4f16:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24 ; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34 ; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
; GFX10-NEXT: v_mov_b32_e32 v11, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx2 v[0:1], v11, s[6:7] ; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[6:7]
; GFX10-NEXT: global_load_dwordx2 v[15:16], v11, s[2:3] offset:32 ; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[2:3] offset:32
; GFX10-NEXT: s_waitcnt vmcnt(1) ; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_cvt_f32_f16_e32 v5, v1 ; GFX10-NEXT: v_cvt_f32_f16_e32 v5, v1
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cvt_f32_f16_e32 v6, v16 ; GFX10-NEXT: v_cvt_f32_f16_e32 v6, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 16, v16
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 16, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v15
; GFX10-NEXT: v_rcp_f32_e32 v6, v6 ; GFX10-NEXT: v_rcp_f32_e32 v6, v6
; GFX10-NEXT: v_cvt_f32_f16_e32 v7, v3
; GFX10-NEXT: v_rcp_f32_e32 v7, v7
; GFX10-NEXT: v_mul_f32_e32 v5, v5, v6 ; GFX10-NEXT: v_mul_f32_e32 v5, v5, v6
; GFX10-NEXT: v_cvt_f16_f32_e32 v6, v5
; GFX10-NEXT: v_div_fixup_f16 v5, v6, v16, v1
; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; GFX10-NEXT: v_mov_b32_e32 v6, v1
; GFX10-NEXT: v_cvt_f16_f32_e32 v5, v5
; GFX10-NEXT: v_div_fixup_f16 v5, v5, v3, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; GFX10-NEXT: v_trunc_f16_e32 v5, v5 ; GFX10-NEXT: v_trunc_f16_e32 v5, v5
; GFX10-NEXT: v_fmac_f16_e64 v6, -v5, v16 ; GFX10-NEXT: v_fmac_f16_e64 v6, -v5, v3
; GFX10-NEXT: v_cvt_f32_f16_e32 v5, v4 ; GFX10-NEXT: v_lshrrev_b32_e32 v3, 16, v3
; GFX10-NEXT: v_mul_f32_e32 v5, v5, v7 ; GFX10-NEXT: v_cvt_f32_f16_e32 v5, v1
; GFX10-NEXT: v_cvt_f32_f16_e32 v7, v2 ; GFX10-NEXT: v_cvt_f32_f16_e32 v7, v3
; GFX10-NEXT: v_cvt_f16_f32_e32 v10, v5
; GFX10-NEXT: v_rcp_f32_e32 v7, v7 ; GFX10-NEXT: v_rcp_f32_e32 v7, v7
; GFX10-NEXT: v_div_fixup_f16 v5, v10, v3, v4 ; GFX10-NEXT: v_mul_f32_e32 v5, v5, v7
; GFX10-NEXT: v_trunc_f16_e32 v10, v5 ; GFX10-NEXT: v_cvt_f16_f32_e32 v5, v5
; GFX10-NEXT: v_fmac_f16_e64 v4, -v10, v3 ; GFX10-NEXT: v_div_fixup_f16 v5, v5, v3, v1
; GFX10-NEXT: v_trunc_f16_e32 v5, v5
; GFX10-NEXT: v_fmac_f16_e64 v1, -v5, v3
; GFX10-NEXT: v_mov_b32_e32 v3, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v3, 0xffff
; GFX10-NEXT: v_and_b32_e32 v5, v3, v6 ; GFX10-NEXT: v_and_b32_e32 v5, v3, v6
; GFX10-NEXT: v_cvt_f32_f16_e32 v6, v15 ; GFX10-NEXT: v_cvt_f32_f16_e32 v6, v2
; GFX10-NEXT: v_lshl_or_b32 v1, v4, 16, v5 ; GFX10-NEXT: v_lshl_or_b32 v1, v1, 16, v5
; GFX10-NEXT: v_rcp_f32_e32 v6, v6 ; GFX10-NEXT: v_rcp_f32_e32 v6, v6
; GFX10-NEXT: v_cvt_f32_f16_e32 v5, v0 ; GFX10-NEXT: v_cvt_f32_f16_e32 v5, v0
; GFX10-NEXT: v_mul_f32_e32 v5, v5, v6 ; GFX10-NEXT: v_mul_f32_e32 v5, v5, v6
; GFX10-NEXT: v_mov_b32_e32 v6, v0 ; GFX10-NEXT: v_mov_b32_e32 v6, v0
; GFX10-NEXT: v_cvt_f16_f32_e32 v5, v5 ; GFX10-NEXT: v_cvt_f16_f32_e32 v5, v5
; GFX10-NEXT: v_div_fixup_f16 v5, v5, v15, v0 ; GFX10-NEXT: v_div_fixup_f16 v5, v5, v2, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v0, 16, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX10-NEXT: v_trunc_f16_e32 v5, v5 ; GFX10-NEXT: v_trunc_f16_e32 v5, v5
; GFX10-NEXT: v_fmac_f16_e64 v6, -v5, v15 ; GFX10-NEXT: v_fmac_f16_e64 v6, -v5, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v2
; GFX10-NEXT: v_cvt_f32_f16_e32 v5, v0 ; GFX10-NEXT: v_cvt_f32_f16_e32 v5, v0
; GFX10-NEXT: v_cvt_f32_f16_e32 v7, v2
; GFX10-NEXT: v_rcp_f32_e32 v7, v7
; GFX10-NEXT: v_mul_f32_e32 v5, v5, v7 ; GFX10-NEXT: v_mul_f32_e32 v5, v5, v7
; GFX10-NEXT: v_cvt_f16_f32_e32 v5, v5 ; GFX10-NEXT: v_cvt_f16_f32_e32 v5, v5
; GFX10-NEXT: v_div_fixup_f16 v5, v5, v2, v0 ; GFX10-NEXT: v_div_fixup_f16 v5, v5, v2, v0
; GFX10-NEXT: v_trunc_f16_e32 v5, v5 ; GFX10-NEXT: v_trunc_f16_e32 v5, v5
; GFX10-NEXT: v_fmac_f16_e64 v0, -v5, v2 ; GFX10-NEXT: v_fmac_f16_e64 v0, -v5, v2
; GFX10-NEXT: v_and_b32_e32 v2, v3, v6 ; GFX10-NEXT: v_and_b32_e32 v2, v3, v6
; GFX10-NEXT: v_lshl_or_b32 v0, v0, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v0, v0, 16, v2
; GFX10-NEXT: global_store_dwordx2 v11, v[0:1], s[4:5] ; GFX10-NEXT: global_store_dwordx2 v4, v[0:1], s[4:5]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
<4 x half> addrspace(1)* %in2) #0 { <4 x half> addrspace(1)* %in2) #0 {
%gep2 = getelementptr <4 x half>, <4 x half> addrspace(1)* %in2, i32 4 %gep2 = getelementptr <4 x half>, <4 x half> addrspace(1)* %in2, i32 4
%r0 = load <4 x half>, <4 x half> addrspace(1)* %in1, align 16 %r0 = load <4 x half>, <4 x half> addrspace(1)* %in1, align 16
%r1 = load <4 x half>, <4 x half> addrspace(1)* %gep2, align 16 %r1 = load <4 x half>, <4 x half> addrspace(1)* %gep2, align 16
%r2 = frem <4 x half> %r0, %r1 %r2 = frem <4 x half> %r0, %r1
store <4 x half> %r2, <4 x half> addrspace(1)* %out, align 16 store <4 x half> %r2, <4 x half> addrspace(1)* %out, align 16
ret void ret void
▲ Show 20 Lines • Show All 201 Lines • ▼ Show 20 Lines
; GFX9-NEXT: global_store_dwordx2 v4, v[0:1], s[4:5] ; GFX9-NEXT: global_store_dwordx2 v4, v[0:1], s[4:5]
; GFX9-NEXT: s_endpgm ; GFX9-NEXT: s_endpgm
; ;
; GFX10-LABEL: frem_v2f32: ; GFX10-LABEL: frem_v2f32:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24 ; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34 ; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
; GFX10-NEXT: v_mov_b32_e32 v11, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx2 v[0:1], v11, s[6:7] ; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[6:7]
; GFX10-NEXT: global_load_dwordx2 v[2:3], v11, s[2:3] offset:32 ; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[2:3] offset:32
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_div_scale_f32 v6, s0, v3, v3, v1 ; GFX10-NEXT: v_div_scale_f32 v6, s0, v3, v3, v1
; GFX10-NEXT: v_div_scale_f32 v5, vcc_lo, v1, v3, v1 ; GFX10-NEXT: v_div_scale_f32 v5, vcc_lo, v1, v3, v1
; GFX10-NEXT: v_rcp_f32_e32 v7, v6 ; GFX10-NEXT: v_rcp_f32_e32 v7, v6
; GFX10-NEXT: s_denorm_mode 15 ; GFX10-NEXT: s_denorm_mode 15
; GFX10-NEXT: v_fma_f32 v8, -v6, v7, 1.0 ; GFX10-NEXT: v_fma_f32 v8, -v6, v7, 1.0
; GFX10-NEXT: v_fma_f32 v7, v8, v7, v7 ; GFX10-NEXT: v_fma_f32 v7, v8, v7, v7
; GFX10-NEXT: v_mul_f32_e32 v8, v5, v7 ; GFX10-NEXT: v_mul_f32_e32 v8, v5, v7
; GFX10-NEXT: v_fma_f32 v9, -v6, v8, v5 ; GFX10-NEXT: v_fma_f32 v9, -v6, v8, v5
; GFX10-NEXT: v_fma_f32 v8, v9, v7, v8 ; GFX10-NEXT: v_fma_f32 v8, v9, v7, v8
; GFX10-NEXT: v_fma_f32 v5, -v6, v8, v5 ; GFX10-NEXT: v_fma_f32 v5, -v6, v8, v5
; GFX10-NEXT: s_denorm_mode 12 ; GFX10-NEXT: s_denorm_mode 12
; GFX10-NEXT: v_div_fmas_f32 v6, v5, v7, v8 ; GFX10-NEXT: v_div_fmas_f32 v5, v5, v7, v8
; GFX10-NEXT: v_div_scale_f32 v4, vcc_lo, v0, v2, v0 ; GFX10-NEXT: v_div_fixup_f32 v5, v5, v3, v1
; GFX10-NEXT: v_div_fixup_f32 v5, v6, v3, v1 ; GFX10-NEXT: v_trunc_f32_e32 v5, v5
; GFX10-NEXT: v_trunc_f32_e32 v6, v5 ; GFX10-NEXT: v_fma_f32 v1, v3, -v5, v1
; GFX10-NEXT: v_div_scale_f32 v5, s0, v2, v2, v0 ; GFX10-NEXT: v_div_scale_f32 v5, s0, v2, v2, v0
; GFX10-NEXT: v_fma_f32 v1, v3, -v6, v1 ; GFX10-NEXT: v_div_scale_f32 v3, vcc_lo, v0, v2, v0
; GFX10-NEXT: v_rcp_f32_e32 v6, v5 ; GFX10-NEXT: v_rcp_f32_e32 v6, v5
; GFX10-NEXT: s_denorm_mode 15 ; GFX10-NEXT: s_denorm_mode 15
; GFX10-NEXT: v_fma_f32 v7, -v5, v6, 1.0 ; GFX10-NEXT: v_fma_f32 v7, -v5, v6, 1.0
; GFX10-NEXT: v_fma_f32 v6, v7, v6, v6 ; GFX10-NEXT: v_fma_f32 v6, v7, v6, v6
; GFX10-NEXT: v_mul_f32_e32 v7, v4, v6 ; GFX10-NEXT: v_mul_f32_e32 v7, v3, v6
; GFX10-NEXT: v_fma_f32 v8, -v5, v7, v4 ; GFX10-NEXT: v_fma_f32 v8, -v5, v7, v3
; GFX10-NEXT: v_fma_f32 v7, v8, v6, v7 ; GFX10-NEXT: v_fma_f32 v7, v8, v6, v7
; GFX10-NEXT: v_fma_f32 v5, -v5, v7, v4 ; GFX10-NEXT: v_fma_f32 v3, -v5, v7, v3
; GFX10-NEXT: s_denorm_mode 12 ; GFX10-NEXT: s_denorm_mode 12
; GFX10-NEXT: v_div_fmas_f32 v3, v5, v6, v7 ; GFX10-NEXT: v_div_fmas_f32 v3, v3, v6, v7
; GFX10-NEXT: v_div_fixup_f32 v3, v3, v2, v0 ; GFX10-NEXT: v_div_fixup_f32 v3, v3, v2, v0
; GFX10-NEXT: v_trunc_f32_e32 v3, v3 ; GFX10-NEXT: v_trunc_f32_e32 v3, v3
; GFX10-NEXT: v_fmac_f32_e64 v0, -v3, v2 ; GFX10-NEXT: v_fmac_f32_e64 v0, -v3, v2
; GFX10-NEXT: global_store_dwordx2 v11, v[0:1], s[4:5] ; GFX10-NEXT: global_store_dwordx2 v4, v[0:1], s[4:5]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
<2 x float> addrspace(1)* %in2) #0 { <2 x float> addrspace(1)* %in2) #0 {
%gep2 = getelementptr <2 x float>, <2 x float> addrspace(1)* %in2, i32 4 %gep2 = getelementptr <2 x float>, <2 x float> addrspace(1)* %in2, i32 4
%r0 = load <2 x float>, <2 x float> addrspace(1)* %in1, align 8 %r0 = load <2 x float>, <2 x float> addrspace(1)* %in1, align 8
%r1 = load <2 x float>, <2 x float> addrspace(1)* %gep2, align 8 %r1 = load <2 x float>, <2 x float> addrspace(1)* %gep2, align 8
%r2 = frem <2 x float> %r0, %r1 %r2 = frem <2 x float> %r0, %r1
store <2 x float> %r2, <2 x float> addrspace(1)* %out, align 8 store <2 x float> %r2, <2 x float> addrspace(1)* %out, align 8
ret void ret void
▲ Show 20 Lines • Show All 324 Lines • ▼ Show 20 Lines
; GFX10-LABEL: frem_v4f32: ; GFX10-LABEL: frem_v4f32:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24 ; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34 ; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
; GFX10-NEXT: v_mov_b32_e32 v8, 0 ; GFX10-NEXT: v_mov_b32_e32 v8, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx4 v[15:18], v8, s[6:7] ; GFX10-NEXT: global_load_dwordx4 v[0:3], v8, s[6:7]
; GFX10-NEXT: global_load_dwordx4 v[4:7], v8, s[2:3] offset:64 ; GFX10-NEXT: global_load_dwordx4 v[4:7], v8, s[2:3] offset:64
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_div_scale_f32 v10, s0, v7, v7, v18 ; GFX10-NEXT: v_div_scale_f32 v10, s0, v7, v7, v3
; GFX10-NEXT: v_div_scale_f32 v9, vcc_lo, v18, v7, v18 ; GFX10-NEXT: v_div_scale_f32 v9, vcc_lo, v3, v7, v3
; GFX10-NEXT: v_rcp_f32_e32 v11, v10 ; GFX10-NEXT: v_rcp_f32_e32 v11, v10
; GFX10-NEXT: s_denorm_mode 15 ; GFX10-NEXT: s_denorm_mode 15
; GFX10-NEXT: v_fma_f32 v12, -v10, v11, 1.0 ; GFX10-NEXT: v_fma_f32 v12, -v10, v11, 1.0
; GFX10-NEXT: v_fma_f32 v11, v12, v11, v11 ; GFX10-NEXT: v_fma_f32 v11, v12, v11, v11
; GFX10-NEXT: v_mul_f32_e32 v12, v9, v11 ; GFX10-NEXT: v_mul_f32_e32 v12, v9, v11
; GFX10-NEXT: v_fma_f32 v13, -v10, v12, v9 ; GFX10-NEXT: v_fma_f32 v13, -v10, v12, v9
; GFX10-NEXT: v_fma_f32 v12, v13, v11, v12 ; GFX10-NEXT: v_fma_f32 v12, v13, v11, v12
; GFX10-NEXT: v_fma_f32 v9, -v10, v12, v9 ; GFX10-NEXT: v_fma_f32 v9, -v10, v12, v9
; GFX10-NEXT: s_denorm_mode 12 ; GFX10-NEXT: s_denorm_mode 12
; GFX10-NEXT: v_div_fmas_f32 v9, v9, v11, v12 ; GFX10-NEXT: v_div_fmas_f32 v9, v9, v11, v12
; GFX10-NEXT: v_div_scale_f32 v0, vcc_lo, v17, v6, v17 ; GFX10-NEXT: v_div_fixup_f32 v9, v9, v7, v3
; GFX10-NEXT: v_div_fixup_f32 v9, v9, v7, v18
; GFX10-NEXT: v_trunc_f32_e32 v9, v9 ; GFX10-NEXT: v_trunc_f32_e32 v9, v9
; GFX10-NEXT: v_fma_f32 v18, v7, -v9, v18 ; GFX10-NEXT: v_fma_f32 v3, v7, -v9, v3
; GFX10-NEXT: v_div_scale_f32 v9, s0, v6, v6, v17 ; GFX10-NEXT: v_div_scale_f32 v9, s0, v6, v6, v2
; GFX10-NEXT: v_div_scale_f32 v7, vcc_lo, v2, v6, v2
; GFX10-NEXT: v_rcp_f32_e32 v10, v9 ; GFX10-NEXT: v_rcp_f32_e32 v10, v9
; GFX10-NEXT: s_denorm_mode 15 ; GFX10-NEXT: s_denorm_mode 15
; GFX10-NEXT: v_fma_f32 v11, -v9, v10, 1.0 ; GFX10-NEXT: v_fma_f32 v11, -v9, v10, 1.0
; GFX10-NEXT: v_fma_f32 v10, v11, v10, v10 ; GFX10-NEXT: v_fma_f32 v10, v11, v10, v10
; GFX10-NEXT: v_mul_f32_e32 v11, v0, v10 ; GFX10-NEXT: v_mul_f32_e32 v11, v7, v10
; GFX10-NEXT: v_fma_f32 v12, -v9, v11, v0 ; GFX10-NEXT: v_fma_f32 v12, -v9, v11, v7
; GFX10-NEXT: v_fma_f32 v11, v12, v10, v11 ; GFX10-NEXT: v_fma_f32 v11, v12, v10, v11
; GFX10-NEXT: v_fma_f32 v1, -v9, v11, v0 ; GFX10-NEXT: v_fma_f32 v7, -v9, v11, v7
; GFX10-NEXT: s_denorm_mode 12 ; GFX10-NEXT: s_denorm_mode 12
; GFX10-NEXT: v_div_fmas_f32 v7, v1, v10, v11 ; GFX10-NEXT: v_div_fmas_f32 v7, v7, v10, v11
; GFX10-NEXT: v_div_fixup_f32 v7, v7, v6, v17 ; GFX10-NEXT: v_div_fixup_f32 v7, v7, v6, v2
; GFX10-NEXT: v_trunc_f32_e32 v7, v7 ; GFX10-NEXT: v_trunc_f32_e32 v7, v7
; GFX10-NEXT: v_fma_f32 v17, v6, -v7, v17 ; GFX10-NEXT: v_fma_f32 v2, v6, -v7, v2
; GFX10-NEXT: v_div_scale_f32 v7, s0, v5, v5, v16 ; GFX10-NEXT: v_div_scale_f32 v7, s0, v5, v5, v1
; GFX10-NEXT: v_div_scale_f32 v6, vcc_lo, v16, v5, v16 ; GFX10-NEXT: v_div_scale_f32 v6, vcc_lo, v1, v5, v1
; GFX10-NEXT: v_rcp_f32_e32 v9, v7 ; GFX10-NEXT: v_rcp_f32_e32 v9, v7
; GFX10-NEXT: s_denorm_mode 15 ; GFX10-NEXT: s_denorm_mode 15
; GFX10-NEXT: v_fma_f32 v10, -v7, v9, 1.0 ; GFX10-NEXT: v_fma_f32 v10, -v7, v9, 1.0
; GFX10-NEXT: v_fma_f32 v9, v10, v9, v9 ; GFX10-NEXT: v_fma_f32 v9, v10, v9, v9
; GFX10-NEXT: v_mul_f32_e32 v0, v6, v9 ; GFX10-NEXT: v_mul_f32_e32 v10, v6, v9
; GFX10-NEXT: v_fma_f32 v11, -v7, v0, v6 ; GFX10-NEXT: v_fma_f32 v11, -v7, v10, v6
; GFX10-NEXT: v_fma_f32 v0, v11, v9, v0 ; GFX10-NEXT: v_fma_f32 v10, v11, v9, v10
; GFX10-NEXT: v_fma_f32 v6, -v7, v0, v6 ; GFX10-NEXT: v_fma_f32 v6, -v7, v10, v6
; GFX10-NEXT: s_denorm_mode 12 ; GFX10-NEXT: s_denorm_mode 12
; GFX10-NEXT: v_div_fmas_f32 v6, v6, v9, v0 ; GFX10-NEXT: v_div_fmas_f32 v6, v6, v9, v10
; GFX10-NEXT: v_div_fixup_f32 v6, v6, v5, v16 ; GFX10-NEXT: v_div_fixup_f32 v6, v6, v5, v1
; GFX10-NEXT: v_trunc_f32_e32 v6, v6 ; GFX10-NEXT: v_trunc_f32_e32 v6, v6
; GFX10-NEXT: v_fma_f32 v16, v5, -v6, v16 ; GFX10-NEXT: v_fma_f32 v1, v5, -v6, v1
; GFX10-NEXT: v_div_scale_f32 v6, s0, v4, v4, v15 ; GFX10-NEXT: v_div_scale_f32 v6, s0, v4, v4, v0
; GFX10-NEXT: v_div_scale_f32 v5, vcc_lo, v15, v4, v15 ; GFX10-NEXT: v_div_scale_f32 v5, vcc_lo, v0, v4, v0
; GFX10-NEXT: v_rcp_f32_e32 v7, v6 ; GFX10-NEXT: v_rcp_f32_e32 v7, v6
; GFX10-NEXT: s_denorm_mode 15 ; GFX10-NEXT: s_denorm_mode 15
; GFX10-NEXT: v_fma_f32 v9, -v6, v7, 1.0 ; GFX10-NEXT: v_fma_f32 v9, -v6, v7, 1.0
; GFX10-NEXT: v_fma_f32 v7, v9, v7, v7 ; GFX10-NEXT: v_fma_f32 v7, v9, v7, v7
; GFX10-NEXT: v_mul_f32_e32 v0, v5, v7 ; GFX10-NEXT: v_mul_f32_e32 v9, v5, v7
; GFX10-NEXT: v_fma_f32 v10, -v6, v0, v5 ; GFX10-NEXT: v_fma_f32 v10, -v6, v9, v5
; GFX10-NEXT: v_fma_f32 v0, v10, v7, v0 ; GFX10-NEXT: v_fma_f32 v9, v10, v7, v9
; GFX10-NEXT: v_fma_f32 v5, -v6, v0, v5 ; GFX10-NEXT: v_fma_f32 v5, -v6, v9, v5
; GFX10-NEXT: s_denorm_mode 12 ; GFX10-NEXT: s_denorm_mode 12
; GFX10-NEXT: v_div_fmas_f32 v5, v5, v7, v0 ; GFX10-NEXT: v_div_fmas_f32 v5, v5, v7, v9
; GFX10-NEXT: v_div_fixup_f32 v5, v5, v4, v15 ; GFX10-NEXT: v_div_fixup_f32 v5, v5, v4, v0
; GFX10-NEXT: v_trunc_f32_e32 v5, v5 ; GFX10-NEXT: v_trunc_f32_e32 v5, v5
; GFX10-NEXT: v_fmac_f32_e64 v15, -v5, v4 ; GFX10-NEXT: v_fmac_f32_e64 v0, -v5, v4
; GFX10-NEXT: global_store_dwordx4 v8, v[15:18], s[4:5] ; GFX10-NEXT: global_store_dwordx4 v8, v[0:3], s[4:5]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
<4 x float> addrspace(1)* %in2) #0 { <4 x float> addrspace(1)* %in2) #0 {
%gep2 = getelementptr <4 x float>, <4 x float> addrspace(1)* %in2, i32 4 %gep2 = getelementptr <4 x float>, <4 x float> addrspace(1)* %in2, i32 4
%r0 = load <4 x float>, <4 x float> addrspace(1)* %in1, align 16 %r0 = load <4 x float>, <4 x float> addrspace(1)* %in1, align 16
%r1 = load <4 x float>, <4 x float> addrspace(1)* %gep2, align 16 %r1 = load <4 x float>, <4 x float> addrspace(1)* %gep2, align 16
%r2 = frem <4 x float> %r0, %r1 %r2 = frem <4 x float> %r0, %r1
store <4 x float> %r2, <4 x float> addrspace(1)* %out, align 16 store <4 x float> %r2, <4 x float> addrspace(1)* %out, align 16
ret void ret void
▲ Show 20 Lines • Show All 224 Lines • ▼ Show 20 Lines
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24 ; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34 ; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
; GFX10-NEXT: v_mov_b32_e32 v16, 0 ; GFX10-NEXT: v_mov_b32_e32 v16, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx4 v[0:3], v16, s[6:7] ; GFX10-NEXT: global_load_dwordx4 v[0:3], v16, s[6:7]
; GFX10-NEXT: global_load_dwordx4 v[18:21], v16, s[2:3] offset:64 ; GFX10-NEXT: global_load_dwordx4 v[4:7], v16, s[2:3] offset:64
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_div_scale_f64 v[8:9], s0, v[20:21], v[20:21], v[2:3] ; GFX10-NEXT: v_div_scale_f64 v[8:9], s0, v[6:7], v[6:7], v[2:3]
; GFX10-NEXT: v_div_scale_f64 v[6:7], s0, v[18:19], v[18:19], v[0:1]
; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[8:9] ; GFX10-NEXT: v_rcp_f64_e32 v[10:11], v[8:9]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[8:9], v[10:11], 1.0 ; GFX10-NEXT: v_fma_f64 v[12:13], -v[8:9], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[12:13], v[10:11] ; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[12:13], v[10:11]
; GFX10-NEXT: v_fma_f64 v[12:13], -v[8:9], v[10:11], 1.0 ; GFX10-NEXT: v_fma_f64 v[12:13], -v[8:9], v[10:11], 1.0
; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[12:13], v[10:11] ; GFX10-NEXT: v_fma_f64 v[10:11], v[10:11], v[12:13], v[10:11]
; GFX10-NEXT: v_div_scale_f64 v[12:13], vcc_lo, v[2:3], v[20:21], v[2:3] ; GFX10-NEXT: v_div_scale_f64 v[12:13], vcc_lo, v[2:3], v[6:7], v[2:3]
; GFX10-NEXT: v_mul_f64 v[14:15], v[12:13], v[10:11] ; GFX10-NEXT: v_mul_f64 v[14:15], v[12:13], v[10:11]
; GFX10-NEXT: v_fma_f64 v[8:9], -v[8:9], v[14:15], v[12:13] ; GFX10-NEXT: v_fma_f64 v[8:9], -v[8:9], v[14:15], v[12:13]
; GFX10-NEXT: v_div_fmas_f64 v[8:9], v[8:9], v[10:11], v[14:15] ; GFX10-NEXT: v_div_fmas_f64 v[8:9], v[8:9], v[10:11], v[14:15]
; GFX10-NEXT: v_div_fixup_f64 v[8:9], v[8:9], v[20:21], v[2:3] ; GFX10-NEXT: v_div_fixup_f64 v[8:9], v[8:9], v[6:7], v[2:3]
; GFX10-NEXT: v_trunc_f64_e32 v[8:9], v[8:9] ; GFX10-NEXT: v_trunc_f64_e32 v[8:9], v[8:9]
; GFX10-NEXT: v_fma_f64 v[2:3], -v[8:9], v[20:21], v[2:3] ; GFX10-NEXT: v_fma_f64 v[2:3], -v[8:9], v[6:7], v[2:3]
; GFX10-NEXT: v_div_scale_f64 v[6:7], s0, v[4:5], v[4:5], v[0:1]
; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[6:7] ; GFX10-NEXT: v_rcp_f64_e32 v[8:9], v[6:7]
; GFX10-NEXT: v_fma_f64 v[10:11], -v[6:7], v[8:9], 1.0 ; GFX10-NEXT: v_fma_f64 v[10:11], -v[6:7], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[10:11], v[8:9] ; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[10:11], v[8:9]
; GFX10-NEXT: v_fma_f64 v[10:11], -v[6:7], v[8:9], 1.0 ; GFX10-NEXT: v_fma_f64 v[10:11], -v[6:7], v[8:9], 1.0
; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[10:11], v[8:9] ; GFX10-NEXT: v_fma_f64 v[8:9], v[8:9], v[10:11], v[8:9]
; GFX10-NEXT: v_div_scale_f64 v[10:11], vcc_lo, v[0:1], v[18:19], v[0:1] ; GFX10-NEXT: v_div_scale_f64 v[10:11], vcc_lo, v[0:1], v[4:5], v[0:1]
; GFX10-NEXT: v_mul_f64 v[12:13], v[10:11], v[8:9] ; GFX10-NEXT: v_mul_f64 v[12:13], v[10:11], v[8:9]
; GFX10-NEXT: v_fma_f64 v[6:7], -v[6:7], v[12:13], v[10:11] ; GFX10-NEXT: v_fma_f64 v[6:7], -v[6:7], v[12:13], v[10:11]
; GFX10-NEXT: v_div_fmas_f64 v[6:7], v[6:7], v[8:9], v[12:13] ; GFX10-NEXT: v_div_fmas_f64 v[6:7], v[6:7], v[8:9], v[12:13]
; GFX10-NEXT: v_div_fixup_f64 v[6:7], v[6:7], v[18:19], v[0:1] ; GFX10-NEXT: v_div_fixup_f64 v[6:7], v[6:7], v[4:5], v[0:1]
; GFX10-NEXT: v_trunc_f64_e32 v[6:7], v[6:7] ; GFX10-NEXT: v_trunc_f64_e32 v[6:7], v[6:7]
; GFX10-NEXT: v_fma_f64 v[0:1], -v[6:7], v[18:19], v[0:1] ; GFX10-NEXT: v_fma_f64 v[0:1], -v[6:7], v[4:5], v[0:1]
; GFX10-NEXT: global_store_dwordx4 v16, v[0:3], s[4:5] ; GFX10-NEXT: global_store_dwordx4 v16, v[0:3], s[4:5]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
<2 x double> addrspace(1)* %in2) #0 { <2 x double> addrspace(1)* %in2) #0 {
%gep2 = getelementptr <2 x double>, <2 x double> addrspace(1)* %in2, i32 4 %gep2 = getelementptr <2 x double>, <2 x double> addrspace(1)* %in2, i32 4
%r0 = load <2 x double>, <2 x double> addrspace(1)* %in1, align 16 %r0 = load <2 x double>, <2 x double> addrspace(1)* %in1, align 16
%r1 = load <2 x double>, <2 x double> addrspace(1)* %gep2, align 16 %r1 = load <2 x double>, <2 x double> addrspace(1)* %gep2, align 16
%r2 = frem <2 x double> %r0, %r1 %r2 = frem <2 x double> %r0, %r1
store <2 x double> %r2, <2 x double> addrspace(1)* %out, align 16 store <2 x double> %r2, <2 x double> addrspace(1)* %out, align 16
ret void ret void
} }
attributes #0 = { nounwind "unsafe-fp-math"="false" "denormal-fp-math-f32"="preserve-sign,preserve-sign" } attributes #0 = { nounwind "unsafe-fp-math"="false" "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
attributes #1 = { nounwind "unsafe-fp-math"="true" "denormal-fp-math-f32"="preserve-sign,preserve-sign" } attributes #1 = { nounwind "unsafe-fp-math"="true" "denormal-fp-math-f32"="preserve-sign,preserve-sign" }

llvm/test/CodeGen/AMDGPU/fshr.ll

Show First 20 Lines • Show All 837 Lines • ▼ Show 20 Lines
; R600: ; %bb.0: ; R600: ; %bb.0:
; R600-NEXT: CF_END ; R600-NEXT: CF_END
; R600-NEXT: PAD ; R600-NEXT: PAD
; ;
; GFX10-LABEL: v_fshr_v3i16: ; GFX10-LABEL: v_fshr_v3i16:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 16, v4
; GFX10-NEXT: v_xor_b32_e32 v8, -1, v4 ; GFX10-NEXT: v_xor_b32_e32 v8, -1, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 16, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 16, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 16, v0
; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0 ; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0
; GFX10-NEXT: v_and_b32_e32 v4, 15, v4 ; GFX10-NEXT: v_and_b32_e32 v4, 15, v4
; GFX10-NEXT: v_and_b32_e32 v8, 15, v8
; GFX10-NEXT: v_and_b32_e32 v9, 15, v6 ; GFX10-NEXT: v_and_b32_e32 v9, 15, v6
; GFX10-NEXT: v_xor_b32_e32 v6, -1, v6 ; GFX10-NEXT: v_xor_b32_e32 v6, -1, v6
; GFX10-NEXT: v_and_b32_e32 v15, 15, v8
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 16, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 16, v2
; GFX10-NEXT: v_lshrrev_b16 v2, v4, v2 ; GFX10-NEXT: v_lshrrev_b16 v2, v4, v2
; GFX10-NEXT: v_lshlrev_b16 v0, v8, v0
; GFX10-NEXT: v_lshlrev_b16 v10, 1, v10 ; GFX10-NEXT: v_lshlrev_b16 v10, 1, v10
; GFX10-NEXT: v_and_b32_e32 v19, 15, v6 ; GFX10-NEXT: v_and_b32_e32 v6, 15, v6
; GFX10-NEXT: v_lshlrev_b16 v0, v15, v0
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v5 ; GFX10-NEXT: v_xor_b32_e32 v11, -1, v5
; GFX10-NEXT: v_lshrrev_b16 v4, v9, v7 ; GFX10-NEXT: v_lshrrev_b16 v4, v9, v7
; GFX10-NEXT: v_lshlrev_b16 v1, 1, v1
; GFX10-NEXT: v_lshlrev_b16 v6, v19, v10
; GFX10-NEXT: v_or_b32_e32 v0, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v0, v0, v2
; GFX10-NEXT: v_and_b32_e32 v7, 15, v11
; GFX10-NEXT: v_and_b32_e32 v2, 15, v5 ; GFX10-NEXT: v_and_b32_e32 v2, 15, v5
; GFX10-NEXT: v_or_b32_e32 v11, v6, v4 ; GFX10-NEXT: v_lshlrev_b16 v6, v6, v10
; GFX10-NEXT: v_lshlrev_b16 v1, 1, v1
; GFX10-NEXT: v_and_b32_e32 v7, 15, v11
; GFX10-NEXT: v_and_b32_e32 v0, 0xffff, v0 ; GFX10-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX10-NEXT: v_lshlrev_b16 v1, v7, v1
; GFX10-NEXT: v_lshrrev_b16 v2, v2, v3 ; GFX10-NEXT: v_lshrrev_b16 v2, v2, v3
; GFX10-NEXT: v_lshl_or_b32 v0, v11, 16, v0 ; GFX10-NEXT: v_or_b32_e32 v4, v6, v4
; GFX10-NEXT: v_lshlrev_b16 v1, v7, v1
; GFX10-NEXT: v_lshl_or_b32 v0, v4, 16, v0
; GFX10-NEXT: v_or_b32_e32 v1, v1, v2 ; GFX10-NEXT: v_or_b32_e32 v1, v1, v2
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%ret = call <3 x i16> @llvm.fshr.v3i16(<3 x i16> %src0, <3 x i16> %src1, <3 x i16> %src2) %ret = call <3 x i16> @llvm.fshr.v3i16(<3 x i16> %src0, <3 x i16> %src1, <3 x i16> %src2)
ret <3 x i16> %ret ret <3 x i16> %ret
} }
define <4 x i16> @v_fshr_v4i16(<4 x i16> %src0, <4 x i16> %src1, <4 x i16> %src2) { define <4 x i16> @v_fshr_v4i16(<4 x i16> %src0, <4 x i16> %src1, <4 x i16> %src2) {
; SI-LABEL: v_fshr_v4i16: ; SI-LABEL: v_fshr_v4i16:
▲ Show 20 Lines • Show All 121 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_and_b32_e32 v13, 15, v10 ; GFX10-NEXT: v_and_b32_e32 v13, 15, v10
; GFX10-NEXT: v_lshlrev_b16 v1, 1, v1 ; GFX10-NEXT: v_lshlrev_b16 v1, 1, v1
; GFX10-NEXT: v_and_b32_e32 v9, 15, v9 ; GFX10-NEXT: v_and_b32_e32 v9, 15, v9
; GFX10-NEXT: v_lshrrev_b16 v6, v6, v7 ; GFX10-NEXT: v_lshrrev_b16 v6, v6, v7
; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0 ; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 16, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v12, 16, v2
; GFX10-NEXT: v_lshlrev_b16 v11, 1, v11 ; GFX10-NEXT: v_lshlrev_b16 v11, 1, v11
; GFX10-NEXT: v_lshlrev_b16 v7, v9, v8 ; GFX10-NEXT: v_lshlrev_b16 v7, v9, v8
; GFX10-NEXT: v_xor_b32_e32 v8, -1, v4
; GFX10-NEXT: v_xor_b32_e32 v9, -1, v10 ; GFX10-NEXT: v_xor_b32_e32 v9, -1, v10
; GFX10-NEXT: v_xor_b32_e32 v10, -1, v5 ; GFX10-NEXT: v_xor_b32_e32 v10, -1, v5
; GFX10-NEXT: v_xor_b32_e32 v8, -1, v4
; GFX10-NEXT: v_and_b32_e32 v4, 15, v4 ; GFX10-NEXT: v_and_b32_e32 v4, 15, v4
; GFX10-NEXT: v_and_b32_e32 v5, 15, v5 ; GFX10-NEXT: v_and_b32_e32 v5, 15, v5
; GFX10-NEXT: v_and_b32_e32 v8, 15, v8
; GFX10-NEXT: v_and_b32_e32 v9, 15, v9 ; GFX10-NEXT: v_and_b32_e32 v9, 15, v9
; GFX10-NEXT: v_and_b32_e32 v10, 15, v10 ; GFX10-NEXT: v_and_b32_e32 v10, 15, v10
; GFX10-NEXT: v_and_b32_e32 v15, 15, v8
; GFX10-NEXT: v_lshrrev_b16 v2, v4, v2 ; GFX10-NEXT: v_lshrrev_b16 v2, v4, v2
; GFX10-NEXT: v_lshrrev_b16 v3, v5, v3 ; GFX10-NEXT: v_lshrrev_b16 v3, v5, v3
; GFX10-NEXT: v_lshlrev_b16 v0, v8, v0
; GFX10-NEXT: v_lshrrev_b16 v4, v13, v12 ; GFX10-NEXT: v_lshrrev_b16 v4, v13, v12
; GFX10-NEXT: v_lshlrev_b16 v1, v10, v1 ; GFX10-NEXT: v_lshlrev_b16 v1, v10, v1
; GFX10-NEXT: v_lshlrev_b16 v0, v15, v0
; GFX10-NEXT: v_lshlrev_b16 v5, v9, v11 ; GFX10-NEXT: v_lshlrev_b16 v5, v9, v11
; GFX10-NEXT: v_or_b32_e32 v1, v1, v3
; GFX10-NEXT: v_or_b32_e32 v0, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v0, v0, v2
; GFX10-NEXT: v_mov_b32_e32 v2, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v2, 0xffff
; GFX10-NEXT: v_or_b32_e32 v1, v1, v3
; GFX10-NEXT: v_or_b32_e32 v3, v7, v6 ; GFX10-NEXT: v_or_b32_e32 v3, v7, v6
; GFX10-NEXT: v_or_b32_e32 v7, v5, v4 ; GFX10-NEXT: v_or_b32_e32 v4, v5, v4
; GFX10-NEXT: v_and_b32_e32 v0, v2, v0 ; GFX10-NEXT: v_and_b32_e32 v0, v2, v0
; GFX10-NEXT: v_and_b32_e32 v1, v2, v1 ; GFX10-NEXT: v_and_b32_e32 v1, v2, v1
; GFX10-NEXT: v_lshl_or_b32 v0, v7, 16, v0 ; GFX10-NEXT: v_lshl_or_b32 v0, v4, 16, v0
; GFX10-NEXT: v_lshl_or_b32 v1, v3, 16, v1 ; GFX10-NEXT: v_lshl_or_b32 v1, v3, 16, v1
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%ret = call <4 x i16> @llvm.fshr.v4i16(<4 x i16> %src0, <4 x i16> %src1, <4 x i16> %src2) %ret = call <4 x i16> @llvm.fshr.v4i16(<4 x i16> %src0, <4 x i16> %src1, <4 x i16> %src2)
ret <4 x i16> %ret ret <4 x i16> %ret
} }
define i64 @v_fshr_i64(i64 %src0, i64 %src1, i64 %src2) { define i64 @v_fshr_i64(i64 %src0, i64 %src1, i64 %src2) {
; SI-LABEL: v_fshr_i64: ; SI-LABEL: v_fshr_i64:
▲ Show 20 Lines • Show All 42 Lines • ▼ Show 20 Lines
; ;
; GFX10-LABEL: v_fshr_i64: ; GFX10-LABEL: v_fshr_i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_not_b32_e32 v5, v4 ; GFX10-NEXT: v_not_b32_e32 v5, v4
; GFX10-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX10-NEXT: v_and_b32_e32 v4, 63, v4 ; GFX10-NEXT: v_and_b32_e32 v4, 63, v4
; GFX10-NEXT: v_and_b32_e32 v7, 63, v5 ; GFX10-NEXT: v_and_b32_e32 v5, 63, v5
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v4, v[2:3] ; GFX10-NEXT: v_lshrrev_b64 v[2:3], v4, v[2:3]
; GFX10-NEXT: v_lshlrev_b64 v[0:1], v7, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], v5, v[0:1]
; GFX10-NEXT: v_or_b32_e32 v0, v0, v2 ; GFX10-NEXT: v_or_b32_e32 v0, v0, v2
; GFX10-NEXT: v_or_b32_e32 v1, v1, v3 ; GFX10-NEXT: v_or_b32_e32 v1, v1, v3
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%ret = call i64 @llvm.fshr.i64(i64 %src0, i64 %src1, i64 %src2) %ret = call i64 @llvm.fshr.i64(i64 %src0, i64 %src1, i64 %src2)
ret i64 %ret ret i64 %ret
} }
define <2 x i64> @v_fshr_v2i64(<2 x i64> %src0, <2 x i64> %src1, <2 x i64> %src2) { define <2 x i64> @v_fshr_v2i64(<2 x i64> %src0, <2 x i64> %src1, <2 x i64> %src2) {
▲ Show 20 Lines • Show All 68 Lines • ▼ Show 20 Lines
; GFX10-LABEL: v_fshr_v2i64: ; GFX10-LABEL: v_fshr_v2i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_not_b32_e32 v9, v8 ; GFX10-NEXT: v_not_b32_e32 v9, v8
; GFX10-NEXT: v_not_b32_e32 v11, v10 ; GFX10-NEXT: v_not_b32_e32 v11, v10
; GFX10-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], 1, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[2:3], 1, v[2:3]
; GFX10-NEXT: v_and_b32_e32 v15, 63, v8 ; GFX10-NEXT: v_and_b32_e32 v8, 63, v8
; GFX10-NEXT: v_and_b32_e32 v19, 63, v9 ; GFX10-NEXT: v_and_b32_e32 v9, 63, v9
; GFX10-NEXT: v_and_b32_e32 v9, 63, v10 ; GFX10-NEXT: v_and_b32_e32 v10, 63, v10
; GFX10-NEXT: v_and_b32_e32 v13, 63, v11 ; GFX10-NEXT: v_and_b32_e32 v11, 63, v11
; GFX10-NEXT: v_lshrrev_b64 v[4:5], v15, v[4:5] ; GFX10-NEXT: v_lshrrev_b64 v[4:5], v8, v[4:5]
; GFX10-NEXT: v_lshlrev_b64 v[11:12], v19, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[0:1], v9, v[0:1]
; GFX10-NEXT: v_lshrrev_b64 v[6:7], v9, v[6:7] ; GFX10-NEXT: v_lshrrev_b64 v[6:7], v10, v[6:7]
; GFX10-NEXT: v_lshlrev_b64 v[15:16], v13, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[2:3], v11, v[2:3]
; GFX10-NEXT: v_or_b32_e32 v0, v11, v4 ; GFX10-NEXT: v_or_b32_e32 v0, v0, v4
; GFX10-NEXT: v_or_b32_e32 v1, v12, v5 ; GFX10-NEXT: v_or_b32_e32 v1, v1, v5
; GFX10-NEXT: v_or_b32_e32 v2, v15, v6 ; GFX10-NEXT: v_or_b32_e32 v2, v2, v6
; GFX10-NEXT: v_or_b32_e32 v3, v16, v7 ; GFX10-NEXT: v_or_b32_e32 v3, v3, v7
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%ret = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %src0, <2 x i64> %src1, <2 x i64> %src2) %ret = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %src0, <2 x i64> %src1, <2 x i64> %src2)
ret <2 x i64> %ret ret <2 x i64> %ret
} }
define i24 @v_fshr_i24(i24 %src0, i24 %src1, i24 %src2) { define i24 @v_fshr_i24(i24 %src0, i24 %src1, i24 %src2) {
; SI-LABEL: v_fshr_i24: ; SI-LABEL: v_fshr_i24:
; SI: ; %bb.0: ; SI: ; %bb.0:
▲ Show 20 Lines • Show All 131 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_mul_hi_u32 v7, v5, s4 ; GFX10-NEXT: v_mul_hi_u32 v7, v5, s4
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 8, v3 ; GFX10-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 4, v6 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 4, v6
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 4, v7 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 4, v7
; GFX10-NEXT: v_mul_lo_u32 v6, v6, 24 ; GFX10-NEXT: v_mul_lo_u32 v6, v6, 24
; GFX10-NEXT: v_mul_lo_u32 v7, v7, 24 ; GFX10-NEXT: v_mul_lo_u32 v7, v7, 24
; GFX10-NEXT: v_sub_nc_u32_e32 v4, v4, v6 ; GFX10-NEXT: v_sub_nc_u32_e32 v4, v4, v6
; GFX10-NEXT: v_sub_nc_u32_e32 v5, v5, v7 ; GFX10-NEXT: v_sub_nc_u32_e32 v5, v5, v7
; GFX10-NEXT: v_add_nc_u32_e32 v7, 8, v4 ; GFX10-NEXT: v_add_nc_u32_e32 v4, 8, v4
; GFX10-NEXT: v_add_nc_u32_e32 v6, 8, v5 ; GFX10-NEXT: v_add_nc_u32_e32 v5, 8, v5
; GFX10-NEXT: v_alignbit_b32 v0, v0, v2, v7 ; GFX10-NEXT: v_alignbit_b32 v0, v0, v2, v4
; GFX10-NEXT: v_alignbit_b32 v1, v1, v3, v6 ; GFX10-NEXT: v_alignbit_b32 v1, v1, v3, v5
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%ret = call <2 x i24> @llvm.fshr.v2i24(<2 x i24> %src0, <2 x i24> %src1, <2 x i24> %src2) %ret = call <2 x i24> @llvm.fshr.v2i24(<2 x i24> %src0, <2 x i24> %src1, <2 x i24> %src2)
ret <2 x i24> %ret ret <2 x i24> %ret
} }

llvm/test/CodeGen/AMDGPU/idiv-licm.ll

Show First 20 Lines • Show All 460 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_mul_f32_e32 v2, v7, v1 ; GFX10-NEXT: v_mul_f32_e32 v2, v7, v1
; GFX10-NEXT: v_add_co_u32 v5, s0, s2, v5 ; GFX10-NEXT: v_add_co_u32 v5, s0, s2, v5
; GFX10-NEXT: s_and_b32 vcc_lo, exec_lo, vcc_lo ; GFX10-NEXT: s_and_b32 vcc_lo, exec_lo, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s0, s3, v6, s0 ; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s0, s3, v6, s0
; GFX10-NEXT: v_trunc_f32_e32 v2, v2 ; GFX10-NEXT: v_trunc_f32_e32 v2, v2
; GFX10-NEXT: v_mad_f32 v7, -v2, v0, v7 ; GFX10-NEXT: v_mad_f32 v7, -v2, v0, v7
; GFX10-NEXT: v_cvt_u32_f32_e32 v2, v2 ; GFX10-NEXT: v_cvt_u32_f32_e32 v2, v2
; GFX10-NEXT: v_cmp_ge_f32_e64 s0, |v7|, v0 ; GFX10-NEXT: v_cmp_ge_f32_e64 s0, |v7|, v0
; GFX10-NEXT: v_add_co_ci_u32_e64 v7, s0, 0, v2, s0 ; GFX10-NEXT: v_add_co_ci_u32_e64 v2, s0, 0, v2, s0
; GFX10-NEXT: global_store_short v[5:6], v7, off ; GFX10-NEXT: global_store_short v[5:6], v2, off
; GFX10-NEXT: s_cbranch_vccz BB4_1 ; GFX10-NEXT: s_cbranch_vccz BB4_1
; GFX10-NEXT: ; %bb.2: ; %bb2 ; GFX10-NEXT: ; %bb.2: ; %bb2
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
bb: bb:
br label %bb3 br label %bb3
bb2: ; preds = %bb3 bb2: ; preds = %bb3
ret void ret void
▲ Show 20 Lines • Show All 62 Lines • ▼ Show 20 Lines
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1 ; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: v_and_b32_e32 v2, s1, v4 ; GFX10-NEXT: v_and_b32_e32 v2, s1, v4
; GFX10-NEXT: v_add_nc_u16 v4, v4, 1 ; GFX10-NEXT: v_add_nc_u16 v4, v4, 1
; GFX10-NEXT: v_cvt_f32_u32_e32 v7, v2 ; GFX10-NEXT: v_cvt_f32_u32_e32 v7, v2
; GFX10-NEXT: v_lshlrev_b64 v[5:6], 1, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[5:6], 1, v[2:3]
; GFX10-NEXT: v_mul_f32_e32 v8, v7, v1 ; GFX10-NEXT: v_mul_f32_e32 v8, v7, v1
; GFX10-NEXT: v_add_co_u32 v5, s0, s2, v5 ; GFX10-NEXT: v_add_co_u32 v5, s0, s2, v5
; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s0, s3, v6, s0 ; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s0, s3, v6, s0
; GFX10-NEXT: v_trunc_f32_e32 v10, v8 ; GFX10-NEXT: v_trunc_f32_e32 v8, v8
; GFX10-NEXT: v_mad_f32 v7, -v10, v0, v7 ; GFX10-NEXT: v_mad_f32 v7, -v8, v0, v7
; GFX10-NEXT: v_cvt_u32_f32_e32 v8, v10 ; GFX10-NEXT: v_cvt_u32_f32_e32 v8, v8
; GFX10-NEXT: v_cmp_ge_f32_e64 vcc_lo, |v7|, v0 ; GFX10-NEXT: v_cmp_ge_f32_e64 vcc_lo, |v7|, v0
; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, 0, v8, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, 0, v8, vcc_lo
; GFX10-NEXT: v_cmp_eq_u16_e32 vcc_lo, 0x400, v4 ; GFX10-NEXT: v_cmp_eq_u16_e32 vcc_lo, 0x400, v4
; GFX10-NEXT: v_mul_lo_u32 v7, v7, s4 ; GFX10-NEXT: v_mul_lo_u32 v7, v7, s4
; GFX10-NEXT: s_and_b32 vcc_lo, exec_lo, vcc_lo ; GFX10-NEXT: s_and_b32 vcc_lo, exec_lo, vcc_lo
; GFX10-NEXT: v_sub_nc_u32_e32 v7, v2, v7 ; GFX10-NEXT: v_sub_nc_u32_e32 v2, v2, v7
; GFX10-NEXT: global_store_short v[5:6], v7, off ; GFX10-NEXT: global_store_short v[5:6], v2, off
; GFX10-NEXT: s_cbranch_vccz BB5_1 ; GFX10-NEXT: s_cbranch_vccz BB5_1
; GFX10-NEXT: ; %bb.2: ; %bb2 ; GFX10-NEXT: ; %bb.2: ; %bb2
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
bb: bb:
br label %bb3 br label %bb3
bb2: ; preds = %bb3 bb2: ; preds = %bb3
ret void ret void
▲ Show 20 Lines • Show All 74 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_and_b32 vcc_lo, exec_lo, vcc_lo ; GFX10-NEXT: s_and_b32 vcc_lo, exec_lo, vcc_lo
; GFX10-NEXT: v_trunc_f32_e32 v2, v2 ; GFX10-NEXT: v_trunc_f32_e32 v2, v2
; GFX10-NEXT: v_or_b32_e32 v8, 1, v8 ; GFX10-NEXT: v_or_b32_e32 v8, 1, v8
; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s0, s3, v6, s0 ; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s0, s3, v6, s0
; GFX10-NEXT: v_mad_f32 v7, -v2, v0, v7 ; GFX10-NEXT: v_mad_f32 v7, -v2, v0, v7
; GFX10-NEXT: v_cvt_i32_f32_e32 v2, v2 ; GFX10-NEXT: v_cvt_i32_f32_e32 v2, v2
; GFX10-NEXT: v_cmp_ge_f32_e64 s1, |v7|, |v0| ; GFX10-NEXT: v_cmp_ge_f32_e64 s1, |v7|, |v0|
; GFX10-NEXT: v_cndmask_b32_e64 v7, 0, v8, s1 ; GFX10-NEXT: v_cndmask_b32_e64 v7, 0, v8, s1
; GFX10-NEXT: v_add_nc_u32_e32 v7, v2, v7 ; GFX10-NEXT: v_add_nc_u32_e32 v2, v2, v7
; GFX10-NEXT: global_store_short v[5:6], v7, off ; GFX10-NEXT: global_store_short v[5:6], v2, off
; GFX10-NEXT: s_cbranch_vccz BB6_1 ; GFX10-NEXT: s_cbranch_vccz BB6_1
; GFX10-NEXT: ; %bb.2: ; %bb2 ; GFX10-NEXT: ; %bb.2: ; %bb2
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
bb: bb:
br label %bb3 br label %bb3
bb2: ; preds = %bb3 bb2: ; preds = %bb3
ret void ret void
▲ Show 20 Lines • Show All 61 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_sext_i32_i16 s1, s4 ; GFX10-NEXT: s_sext_i32_i16 s1, s4
; GFX10-NEXT: v_cvt_f32_i32_e32 v0, s1 ; GFX10-NEXT: v_cvt_f32_i32_e32 v0, s1
; GFX10-NEXT: v_rcp_iflag_f32_e32 v1, v0 ; GFX10-NEXT: v_rcp_iflag_f32_e32 v1, v0
; GFX10-NEXT: BB7_1: ; %bb3 ; GFX10-NEXT: BB7_1: ; %bb3
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1 ; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: v_bfe_i32 v7, v4, 0, 16 ; GFX10-NEXT: v_bfe_i32 v7, v4, 0, 16
; GFX10-NEXT: v_and_b32_e32 v2, 0xffff, v4 ; GFX10-NEXT: v_and_b32_e32 v2, 0xffff, v4
; GFX10-NEXT: v_add_nc_u16 v4, v4, 1 ; GFX10-NEXT: v_add_nc_u16 v4, v4, 1
; GFX10-NEXT: v_cvt_f32_i32_e32 v11, v7 ; GFX10-NEXT: v_cvt_f32_i32_e32 v5, v7
; GFX10-NEXT: v_xor_b32_e32 v6, s1, v7 ; GFX10-NEXT: v_xor_b32_e32 v6, s1, v7
; GFX10-NEXT: v_mul_f32_e32 v8, v11, v1 ; GFX10-NEXT: v_mul_f32_e32 v8, v5, v1
; GFX10-NEXT: v_ashrrev_i32_e32 v6, 30, v6 ; GFX10-NEXT: v_ashrrev_i32_e32 v6, 30, v6
; GFX10-NEXT: v_trunc_f32_e32 v10, v8 ; GFX10-NEXT: v_trunc_f32_e32 v8, v8
; GFX10-NEXT: v_or_b32_e32 v6, 1, v6 ; GFX10-NEXT: v_or_b32_e32 v6, 1, v6
; GFX10-NEXT: v_mad_f32 v5, -v10, v0, v11 ; GFX10-NEXT: v_mad_f32 v5, -v8, v0, v5
; GFX10-NEXT: v_cvt_i32_f32_e32 v8, v10 ; GFX10-NEXT: v_cvt_i32_f32_e32 v8, v8
; GFX10-NEXT: v_cmp_ge_f32_e64 vcc_lo, |v5|, |v0| ; GFX10-NEXT: v_cmp_ge_f32_e64 vcc_lo, |v5|, |v0|
; GFX10-NEXT: v_cndmask_b32_e32 v9, 0, v6, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v9, 0, v6, vcc_lo
; GFX10-NEXT: v_lshlrev_b64 v[5:6], 1, v[2:3] ; GFX10-NEXT: v_lshlrev_b64 v[5:6], 1, v[2:3]
; GFX10-NEXT: v_cmp_eq_u16_e32 vcc_lo, 0x400, v4 ; GFX10-NEXT: v_cmp_eq_u16_e32 vcc_lo, 0x400, v4
; GFX10-NEXT: v_add_nc_u32_e32 v2, v8, v9 ; GFX10-NEXT: v_add_nc_u32_e32 v2, v8, v9
; GFX10-NEXT: v_add_co_u32 v5, s0, s2, v5 ; GFX10-NEXT: v_add_co_u32 v5, s0, s2, v5
; GFX10-NEXT: s_and_b32 vcc_lo, exec_lo, vcc_lo ; GFX10-NEXT: s_and_b32 vcc_lo, exec_lo, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s0, s3, v6, s0 ; GFX10-NEXT: v_add_co_ci_u32_e64 v6, s0, s3, v6, s0
; GFX10-NEXT: v_mul_lo_u32 v2, v2, s1 ; GFX10-NEXT: v_mul_lo_u32 v2, v2, s1
; GFX10-NEXT: v_sub_nc_u32_e32 v7, v7, v2 ; GFX10-NEXT: v_sub_nc_u32_e32 v2, v7, v2
; GFX10-NEXT: global_store_short v[5:6], v7, off ; GFX10-NEXT: global_store_short v[5:6], v2, off
; GFX10-NEXT: s_cbranch_vccz BB7_1 ; GFX10-NEXT: s_cbranch_vccz BB7_1
; GFX10-NEXT: ; %bb.2: ; %bb2 ; GFX10-NEXT: ; %bb.2: ; %bb2
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
bb: bb:
br label %bb3 br label %bb3
bb2: ; preds = %bb3 bb2: ; preds = %bb3
ret void ret void
Show All 11 Lines

llvm/test/CodeGen/AMDGPU/idot2.ll

Show First 20 Lines • Show All 2,726 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0 ; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7]
; GFX10-DL-NEXT: global_load_ushort v4, v1, s[2:3] ; GFX10-DL-NEXT: global_load_ushort v4, v1, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v0, 16, v2
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 16, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 16, v3
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v0, v7, v5, v4 ; GFX10-DL-NEXT: v_mad_u16 v0, v0, v5, v4
; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0
; GFX10-DL-NEXT: global_store_short v1, v0, s[2:3] ; GFX10-DL-NEXT: global_store_short v1, v0, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<2 x i16> addrspace(1)* %src2, <2 x i16> addrspace(1)* %src2,
i16 addrspace(1)* nocapture %dst) { i16 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
%gep1 = getelementptr <2 x i16>, <2 x i16> addrspace(1)* %src1, i32 %idx %gep1 = getelementptr <2 x i16>, <2 x i16> addrspace(1)* %src1, i32 %idx
▲ Show 20 Lines • Show All 169 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/idot4s.ll

Show First 20 Lines • Show All 335 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0
; GFX10-DL-NEXT: global_load_ushort v3, v0, s[2:3] ; GFX10-DL-NEXT: global_load_ushort v3, v0, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v1
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 8, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 8, v2
; GFX10-DL-NEXT: v_bfe_i32 v4, v1, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v4, v1, 0, 8
; GFX10-DL-NEXT: v_bfe_i32 v10, v2, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v7, v2, 0, 8
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v8, 16, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v8, 16, v1
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v9, 16, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v9, 16, v2
; GFX10-DL-NEXT: v_bfe_i32 v5, v5, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v5, v5, 0, 8
; GFX10-DL-NEXT: v_bfe_i32 v6, v6, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v6, v6, 0, 8
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v3, v4, v10, v3 ; GFX10-DL-NEXT: v_mad_u16 v3, v4, v7, v3
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 24, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 24, v1
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2
; GFX10-DL-NEXT: v_bfe_i32 v4, v8, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v4, v8, 0, 8
; GFX10-DL-NEXT: v_bfe_i32 v10, v9, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v7, v9, 0, 8
; GFX10-DL-NEXT: v_mad_u16 v3, v5, v6, v3 ; GFX10-DL-NEXT: v_mad_u16 v3, v5, v6, v3
; GFX10-DL-NEXT: v_bfe_i32 v1, v1, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v1, v1, 0, 8
; GFX10-DL-NEXT: v_bfe_i32 v2, v2, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v2, v2, 0, 8
; GFX10-DL-NEXT: v_mad_u16 v3, v4, v10, v3 ; GFX10-DL-NEXT: v_mad_u16 v3, v4, v7, v3
; GFX10-DL-NEXT: v_mad_u16 v1, v1, v2, v3 ; GFX10-DL-NEXT: v_mad_u16 v1, v1, v2, v3
; GFX10-DL-NEXT: global_store_short v0, v1, s[2:3] ; GFX10-DL-NEXT: global_store_short v0, v1, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i16 addrspace(1)* nocapture %dst) { i16 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
%gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx %gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx
▲ Show 20 Lines • Show All 162 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0 ; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7]
; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3] ; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v11, 8, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v0, 8, v2
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v3
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v4, v2, v3, v4 ; GFX10-DL-NEXT: v_mad_u16 v4, v2, v3, v4
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 16, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 16, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 24, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 24, v3
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v4 ; GFX10-DL-NEXT: v_mad_u16 v0, v0, v5, v4
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0
; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3] ; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i8 addrspace(1)* nocapture %dst) { i8 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
▲ Show 20 Lines • Show All 158 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v1, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v1, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: s_load_dword s2, s[0:1], 0x0 ; GFX10-DL-NEXT: s_load_dword s2, s[0:1], 0x0
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_bfe_i32 v0, v1, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v0, v1, 0, 8
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_bfe_i32 v3, v2, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v3, v2, 0, 8
; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v7, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1 ; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v4, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
; GFX10-DL-NEXT: v_mul_i32_i24_e32 v5, v0, v3 ; GFX10-DL-NEXT: v_mul_i32_i24_e32 v5, v0, v3
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: v_mad_i32_i24 v0, v0, v3, s2 ; GFX10-DL-NEXT: v_mad_i32_i24 v0, v0, v3, s2
; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v3, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2 ; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v3, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2
; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v1, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3 ; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v1, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0
; GFX10-DL-NEXT: v_add3_u32 v0, v7, v0, v5 ; GFX10-DL-NEXT: v_add3_u32 v0, v4, v0, v5
; GFX10-DL-NEXT: v_add3_u32 v0, v0, v3, v1 ; GFX10-DL-NEXT: v_add3_u32 v0, v0, v3, v1
; GFX10-DL-NEXT: global_store_dword v2, v0, s[0:1] ; GFX10-DL-NEXT: global_store_dword v2, v0, s[0:1]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i32 addrspace(1)* nocapture %dst) { i32 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
%gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx %gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx
▲ Show 20 Lines • Show All 166 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v1, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v1, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: s_load_dword s2, s[0:1], 0x0 ; GFX10-DL-NEXT: s_load_dword s2, s[0:1], 0x0
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_lshrrev_b16 v0, 8, v1 ; GFX10-DL-NEXT: v_lshrrev_b16 v0, 8, v1
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_lshrrev_b16 v3, 8, v2 ; GFX10-DL-NEXT: v_lshrrev_b16 v3, 8, v2
; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v7, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:BYTE_0 ; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v4, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:BYTE_0
; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v0, sext(v0), sext(v3) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:BYTE_0 ; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v0, sext(v0), sext(v3) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:BYTE_0
; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v3, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2 ; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v3, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2
; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v1, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3 ; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v1, sext(v1), sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: v_add3_u32 v0, v7, s2, v0 ; GFX10-DL-NEXT: v_add3_u32 v0, v4, s2, v0
; GFX10-DL-NEXT: v_add3_u32 v0, v0, v3, v1 ; GFX10-DL-NEXT: v_add3_u32 v0, v0, v3, v1
; GFX10-DL-NEXT: global_store_dword v2, v0, s[0:1] ; GFX10-DL-NEXT: global_store_dword v2, v0, s[0:1]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i32 addrspace(1)* nocapture %dst) { i32 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
%gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx %gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx
▲ Show 20 Lines • Show All 252 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/idot4u.ll

Show First 20 Lines • Show All 321 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0
; GFX10-DL-NEXT: global_load_ushort v3, v0, s[2:3] ; GFX10-DL-NEXT: global_load_ushort v3, v0, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v1
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 8, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 8, v2
; GFX10-DL-NEXT: v_and_b32_e32 v4, s0, v1 ; GFX10-DL-NEXT: v_and_b32_e32 v4, s0, v1
; GFX10-DL-NEXT: v_and_b32_e32 v10, s0, v2 ; GFX10-DL-NEXT: v_and_b32_e32 v7, s0, v2
; GFX10-DL-NEXT: v_and_b32_e32 v5, s0, v5 ; GFX10-DL-NEXT: v_and_b32_e32 v5, s0, v5
; GFX10-DL-NEXT: v_and_b32_e32 v6, s0, v6 ; GFX10-DL-NEXT: v_and_b32_e32 v6, s0, v6
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v3, v4, v10, v3 ; GFX10-DL-NEXT: v_mad_u16 v3, v4, v7, v3
; GFX10-DL-NEXT: v_and_b32_sdwa v4, v1, s0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-DL-NEXT: v_and_b32_sdwa v4, v1, s0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-DL-NEXT: v_and_b32_sdwa v10, v2, s0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-DL-NEXT: v_and_b32_sdwa v7, v2, s0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 24, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 24, v1
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2
; GFX10-DL-NEXT: v_mad_u16 v3, v5, v6, v3 ; GFX10-DL-NEXT: v_mad_u16 v3, v5, v6, v3
; GFX10-DL-NEXT: v_mad_u16 v3, v4, v10, v3 ; GFX10-DL-NEXT: v_mad_u16 v3, v4, v7, v3
; GFX10-DL-NEXT: v_mad_u16 v1, v1, v2, v3 ; GFX10-DL-NEXT: v_mad_u16 v1, v1, v2, v3
; GFX10-DL-NEXT: global_store_short v0, v1, s[2:3] ; GFX10-DL-NEXT: global_store_short v0, v1, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i16 addrspace(1)* nocapture %dst) { i16 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
%gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx %gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx
▲ Show 20 Lines • Show All 163 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0 ; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7]
; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3] ; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v11, 8, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v0, 8, v2
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v3
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v4, v2, v3, v4 ; GFX10-DL-NEXT: v_mad_u16 v4, v2, v3, v4
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 16, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 16, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 24, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 24, v3
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v4 ; GFX10-DL-NEXT: v_mad_u16 v0, v0, v5, v4
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0
; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3] ; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i8 addrspace(1)* nocapture %dst) { i8 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
▲ Show 20 Lines • Show All 298 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0 ; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7]
; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3] ; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v11, 8, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v0, 8, v2
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v3
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v4, v3, v2, v4 ; GFX10-DL-NEXT: v_mad_u16 v4, v3, v2, v4
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 16, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 16, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 24, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 24, v3
; GFX10-DL-NEXT: v_mad_u16 v0, v5, v11, v4 ; GFX10-DL-NEXT: v_mad_u16 v0, v5, v0, v4
; GFX10-DL-NEXT: v_mad_u16 v0, v7, v6, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v7, v6, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v3, v2, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v3, v2, v0
; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3] ; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i8 addrspace(1)* nocapture %dst) { i8 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
▲ Show 20 Lines • Show All 158 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0 ; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7]
; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3] ; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 8, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v0, 8, v2
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v3
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v0, v5, v7, v4 ; GFX10-DL-NEXT: v_mad_u16 v0, v5, v0, v4
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 16, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 16, v3
; GFX10-DL-NEXT: v_mad_u16 v0, v3, v2, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v3, v2, v0
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 24, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 24, v3
; GFX10-DL-NEXT: v_mad_u16 v0, v5, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v5, v4, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v3, v2, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v3, v2, v0
; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3] ; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i8 addrspace(1)* nocapture %dst) { i8 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
%gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx %gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx
▲ Show 20 Lines • Show All 163 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v1, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v1, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: s_load_dword s2, s[0:1], 0x0 ; GFX10-DL-NEXT: s_load_dword s2, s[0:1], 0x0
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_and_b32_e32 v0, s3, v1 ; GFX10-DL-NEXT: v_and_b32_e32 v0, s3, v1
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_and_b32_e32 v3, s3, v2 ; GFX10-DL-NEXT: v_and_b32_e32 v3, s3, v2
; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v7, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1 ; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v4, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
; GFX10-DL-NEXT: v_mul_u32_u24_e32 v5, v0, v3 ; GFX10-DL-NEXT: v_mul_u32_u24_e32 v5, v0, v3
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: v_mad_u32_u24 v0, v0, v3, s2 ; GFX10-DL-NEXT: v_mad_u32_u24 v0, v0, v3, s2
; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v3, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2 ; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v3, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2
; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3 ; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0
; GFX10-DL-NEXT: v_add3_u32 v0, v7, v0, v5 ; GFX10-DL-NEXT: v_add3_u32 v0, v4, v0, v5
; GFX10-DL-NEXT: v_add3_u32 v0, v0, v3, v1 ; GFX10-DL-NEXT: v_add3_u32 v0, v0, v3, v1
; GFX10-DL-NEXT: global_store_dword v2, v0, s[0:1] ; GFX10-DL-NEXT: global_store_dword v2, v0, s[0:1]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i32 addrspace(1)* nocapture %dst) { i32 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
%gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx %gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx
▲ Show 20 Lines • Show All 173 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: s_load_dword s2, s[0:1], 0x0 ; GFX10-DL-NEXT: s_load_dword s2, s[0:1], 0x0
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_bfe_u32 v0, v1, 8, 8 ; GFX10-DL-NEXT: v_bfe_u32 v0, v1, 8, 8
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_bfe_u32 v3, v2, 8, 8 ; GFX10-DL-NEXT: v_bfe_u32 v3, v2, 8, 8
; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v4, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:BYTE_0 ; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v4, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:BYTE_0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: v_mad_u32_u24 v6, v0, v3, s2 ; GFX10-DL-NEXT: v_mad_u32_u24 v0, v0, v3, s2
; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v3, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2 ; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v3, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2
; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3 ; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
; GFX10-DL-NEXT: v_add_nc_u32_e32 v2, s2, v6 ; GFX10-DL-NEXT: v_add_nc_u32_e32 v2, s2, v0
; GFX10-DL-NEXT: v_add3_u32 v0, v6, v4, v3 ; GFX10-DL-NEXT: v_add3_u32 v0, v0, v4, v3
; GFX10-DL-NEXT: v_mov_b32_e32 v3, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v3, 0
; GFX10-DL-NEXT: v_add3_u32 v0, v0, v1, v2 ; GFX10-DL-NEXT: v_add3_u32 v0, v0, v1, v2
; GFX10-DL-NEXT: global_store_dword v3, v0, s[0:1] ; GFX10-DL-NEXT: global_store_dword v3, v0, s[0:1]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i32 addrspace(1)* nocapture %dst) { i32 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
▲ Show 20 Lines • Show All 189 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0
; GFX10-DL-NEXT: global_load_ushort v3, v0, s[2:3] ; GFX10-DL-NEXT: global_load_ushort v3, v0, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v4, 8, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v4, 8, v1
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v2
; GFX10-DL-NEXT: v_bfe_i32 v6, v1, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v6, v1, 0, 8
; GFX10-DL-NEXT: v_bfe_i32 v9, v2, 0, 8 ; GFX10-DL-NEXT: v_bfe_i32 v7, v2, 0, 8
; GFX10-DL-NEXT: v_and_b32_e32 v4, s0, v4 ; GFX10-DL-NEXT: v_and_b32_e32 v4, s0, v4
; GFX10-DL-NEXT: v_and_b32_e32 v5, s0, v5 ; GFX10-DL-NEXT: v_and_b32_e32 v5, s0, v5
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v3, v4, v5, v3 ; GFX10-DL-NEXT: v_mad_u16 v3, v4, v5, v3
; GFX10-DL-NEXT: v_and_b32_sdwa v4, v1, s0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-DL-NEXT: v_and_b32_sdwa v4, v1, s0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-DL-NEXT: v_and_b32_sdwa v5, v2, s0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX10-DL-NEXT: v_and_b32_sdwa v5, v2, s0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 24, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 24, v1
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 24, v2
; GFX10-DL-NEXT: v_mad_u16 v3, v6, v9, v3 ; GFX10-DL-NEXT: v_mad_u16 v3, v6, v7, v3
; GFX10-DL-NEXT: v_mad_u16 v3, v4, v5, v3 ; GFX10-DL-NEXT: v_mad_u16 v3, v4, v5, v3
; GFX10-DL-NEXT: v_mad_u16 v1, v1, v2, v3 ; GFX10-DL-NEXT: v_mad_u16 v1, v1, v2, v3
; GFX10-DL-NEXT: global_store_short v0, v1, s[2:3] ; GFX10-DL-NEXT: global_store_short v0, v1, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i16 addrspace(1)* nocapture %dst) { i16 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
▲ Show 20 Lines • Show All 161 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v1, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v1, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: s_load_dword s2, s[0:1], 0x0 ; GFX10-DL-NEXT: s_load_dword s2, s[0:1], 0x0
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_and_b32_sdwa v0, s3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_1 ; GFX10-DL-NEXT: v_and_b32_sdwa v0, s3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_1
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_and_b32_sdwa v3, s3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_1 ; GFX10-DL-NEXT: v_and_b32_sdwa v3, s3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_1
; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v7, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:BYTE_0 ; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v4, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:BYTE_0
; GFX10-DL-NEXT: v_mul_u32_u24_e32 v0, v0, v3 ; GFX10-DL-NEXT: v_mul_u32_u24_e32 v0, v0, v3
; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v3, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2 ; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v3, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2
; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3 ; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: v_add3_u32 v0, v7, s2, v0 ; GFX10-DL-NEXT: v_add3_u32 v0, v4, s2, v0
; GFX10-DL-NEXT: v_add3_u32 v0, v0, v3, v1 ; GFX10-DL-NEXT: v_add3_u32 v0, v0, v3, v1
; GFX10-DL-NEXT: global_store_dword v2, v0, s[0:1] ; GFX10-DL-NEXT: global_store_dword v2, v0, s[0:1]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i32 addrspace(1)* nocapture %dst) { i32 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
%gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx %gep1 = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %src1, i32 %idx
▲ Show 20 Lines • Show All 398 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 24, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 24, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 16, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v6, 16, v1
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v2
; GFX10-DL-NEXT: v_lshrrev_b16 v8, 8, v2 ; GFX10-DL-NEXT: v_lshrrev_b16 v8, 8, v2
; GFX10-DL-NEXT: v_mul_lo_u16 v4, v4, v5 ; GFX10-DL-NEXT: v_mul_lo_u16 v4, v4, v5
; GFX10-DL-NEXT: v_lshrrev_b16 v5, 8, v1 ; GFX10-DL-NEXT: v_lshrrev_b16 v5, 8, v1
; GFX10-DL-NEXT: v_mul_lo_u16 v9, v6, v7 ; GFX10-DL-NEXT: v_mul_lo_u16 v9, v6, v7
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v3, v1, v2, v3 ; GFX10-DL-NEXT: v_mad_u16 v1, v1, v2, v3
; GFX10-DL-NEXT: v_lshlrev_b16 v4, 8, v4 ; GFX10-DL-NEXT: v_lshlrev_b16 v4, 8, v4
; GFX10-DL-NEXT: v_mul_lo_u16 v5, v5, v8 ; GFX10-DL-NEXT: v_mul_lo_u16 v5, v5, v8
; GFX10-DL-NEXT: v_or_b32_sdwa v4, v9, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; GFX10-DL-NEXT: v_or_b32_sdwa v4, v9, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-DL-NEXT: v_lshlrev_b16 v5, 8, v5 ; GFX10-DL-NEXT: v_lshlrev_b16 v5, 8, v5
; GFX10-DL-NEXT: v_lshlrev_b32_e32 v8, 16, v4 ; GFX10-DL-NEXT: v_lshlrev_b32_e32 v8, 16, v4
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 8, v4 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 8, v4
; GFX10-DL-NEXT: v_or_b32_sdwa v5, v5, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD ; GFX10-DL-NEXT: v_or_b32_sdwa v5, v5, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v5 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v5
; GFX10-DL-NEXT: v_add_nc_u16 v1, v3, v5 ; GFX10-DL-NEXT: v_add_nc_u16 v1, v1, v5
; GFX10-DL-NEXT: v_mad_u16 v1, v6, v7, v1 ; GFX10-DL-NEXT: v_mad_u16 v1, v6, v7, v1
; GFX10-DL-NEXT: v_add_nc_u16 v1, v1, v2 ; GFX10-DL-NEXT: v_add_nc_u16 v1, v1, v2
; GFX10-DL-NEXT: global_store_byte v0, v1, s[0:1] ; GFX10-DL-NEXT: global_store_byte v0, v1, s[0:1]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<4 x i8> addrspace(1)* %src2, <4 x i8> addrspace(1)* %src2,
i8 addrspace(1)* nocapture %dst) { i8 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
Show All 22 Lines

llvm/test/CodeGen/AMDGPU/idot8s.ll

Show First 20 Lines • Show All 638 Lines • ▼ Show 20 Lines
; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v4, 28, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v4, 28, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v5, 24, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v5, 24, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v6, 20, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v6, 20, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v7, 16, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v7, 16, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v10, 4, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v10, 4, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v1, 12, v1
; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v16, 4, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v16, 4, v2
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v17, 12, v2 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v17, 12, v2
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v1, 12, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v11, 28, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v11, 28, v2
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v10, 12, v10 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v10, 12, v10
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v16, 12, v16
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v12, 24, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v12, 24, v2
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v16, 12, v16
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v13, 20, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v13, 20, v2
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v14, 16, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v14, 16, v2
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v2
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v2, 8, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v2, 8, v2
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v18, 12, v1 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v1, 12, v1
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v17, 12, v17 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v17, 12, v17
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v9, 12, v9 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v9, 12, v9
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v10, 12, v10 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v10, 12, v10
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v2, 12, v2 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v2, 12, v2
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v16, 12, v16 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v16, 12, v16
; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v18, v17, v3 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v1, v17, v3
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 12, v8 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 12, v8
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v9 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v9
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v9, 12, v15 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v9, 12, v15
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v2 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v2
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v10, v16, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v10, v16, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v7, 12, v7 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v7, 12, v7
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v10, 12, v14 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v10, 12, v14
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v8, 12, v8 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v8, 12, v8
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v15, 12, v9 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v9, 12, v9
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v3, v2, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v3, v2, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v6, 12, v6 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v6, 12, v6
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v7 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v7
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v7, 12, v13 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v7, 12, v13
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v10 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v10
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v8, v15, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v8, v9, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v5, 12, v5 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v5, 12, v5
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 12, v12 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 12, v12
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v6, 12, v6 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v6, 12, v6
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v7, 12, v7 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v7, 12, v7
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v2, v3, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v2, v3, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v4, 12, v4 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v4, 12, v4
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v5 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v5
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v8
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v5, 12, v11 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v5, 12, v11
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v8
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v6, v7, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v6, v7, v1
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v4, 12, v4 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v4, 12, v4
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v7, 12, v5 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v5, 12, v5
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v2, v3, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v2, v3, v1
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v4, v7, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v4, v5, v1
; GFX10-DL-XNACK-NEXT: global_store_short v0, v1, s[0:1] ; GFX10-DL-XNACK-NEXT: global_store_short v0, v1, s[0:1]
; GFX10-DL-XNACK-NEXT: s_endpgm ; GFX10-DL-XNACK-NEXT: s_endpgm
; ;
; GFX10-DL-NOXNACK-LABEL: idot8_acc16: ; GFX10-DL-NOXNACK-LABEL: idot8_acc16:
; GFX10-DL-NOXNACK: ; %bb.0: ; %entry ; GFX10-DL-NOXNACK: ; %bb.0: ; %entry
; GFX10-DL-NOXNACK-NEXT: s_clause 0x1 ; GFX10-DL-NOXNACK-NEXT: s_clause 0x1
; GFX10-DL-NOXNACK-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24 ; GFX10-DL-NOXNACK-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX10-DL-NOXNACK-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34 ; GFX10-DL-NOXNACK-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
Show All 13 Lines
; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v4, 28, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v4, 28, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v5, 24, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v5, 24, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v6, 20, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v6, 20, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v7, 16, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v7, 16, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v10, 4, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v10, 4, v1
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v1, 12, v1
; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v16, 4, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v16, 4, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v17, 12, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v17, 12, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v1, 12, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v11, 28, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v11, 28, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v10, 12, v10 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v10, 12, v10
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v16, 12, v16
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v12, 24, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v12, 24, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v16, 12, v16
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v13, 20, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v13, 20, v0
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v14, 16, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v14, 16, v0
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v0
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v0, 8, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v18, 12, v1 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v1
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v17, 12, v17 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v17, 12, v17
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v9, 12, v9 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v9, 12, v9
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v10, 12, v10 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v10, 12, v10
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v0, 12, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v0, 12, v0
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v16, 12, v16 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v16, 12, v16
; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v1, v18, v17, v3 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v1, v1, v17, v3
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v8, 12, v8
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v9 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v9
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v9, 12, v15 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v9, 12, v15
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v8, 12, v8
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v0, 12, v0 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v0, 12, v0
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v1, v10, v16, v1 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v1, v10, v16, v1
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v7, 12, v7 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v7, 12, v7
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v10, 12, v14 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v10, 12, v14
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v15, 12, v8 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v8, 12, v8
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v9, 12, v9 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v9, 12, v9
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v3, v0, v1 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v3, v0, v1
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v6, 12, v6 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v6, 12, v6
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v7 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v7
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v7, 12, v13 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v7, 12, v13
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v10 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v10
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v15, v9, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v8, v9, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v5, 12, v5 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v5, 12, v5
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v8, 12, v12 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v8, 12, v12
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v6, 12, v6 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v6, 12, v6
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v7, 12, v7 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v7, 12, v7
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v1, v3, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v1, v3, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v4, 12, v4
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v5 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v5
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v5, 12, v11 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v5, 12, v11
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v8 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v8
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v4, 12, v4
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v4, 12, v4
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v5, 12, v5 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v5, 12, v5
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v7, 12, v4
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v1, v3, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v1, v3, v0
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v7, v5, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NOXNACK-NEXT: global_store_short v2, v0, s[0:1] ; GFX10-DL-NOXNACK-NEXT: global_store_short v2, v0, s[0:1]
; GFX10-DL-NOXNACK-NEXT: s_endpgm ; GFX10-DL-NOXNACK-NEXT: s_endpgm
; GFX10-DL-LABEL: idot8_acc16: ; GFX10-DL-LABEL: idot8_acc16:
; GFX10-DL: ; %bb.0: ; %entry ; GFX10-DL: ; %bb.0: ; %entry
; GFX10-DL-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x34 ; GFX10-DL-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x34
; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0
; GFX10-DL-NEXT: s_mov_b32 s12, SCRATCH_RSRC_DWORD0 ; GFX10-DL-NEXT: s_mov_b32 s12, SCRATCH_RSRC_DWORD0
; GFX10-DL-NEXT: s_mov_b32 s13, SCRATCH_RSRC_DWORD1 ; GFX10-DL-NEXT: s_mov_b32 s13, SCRATCH_RSRC_DWORD1
▲ Show 20 Lines • Show All 431 Lines • ▼ Show 20 Lines
; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v4, 28, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v4, 28, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v5, 24, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v5, 24, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v6, 20, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v6, 20, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v7, 16, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v7, 16, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v10, 4, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v10, 4, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v1, 12, v1
; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v16, 4, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v16, 4, v2
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v17, 12, v2 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v17, 12, v2
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v1, 12, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v11, 28, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v11, 28, v2
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v10, 12, v10 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v10, 12, v10
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v16, 12, v16
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v12, 24, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v12, 24, v2
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v16, 12, v16
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v13, 20, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v13, 20, v2
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v14, 16, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v14, 16, v2
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v2
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v2, 8, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v2, 8, v2
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v18, 12, v1 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v1, 12, v1
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v17, 12, v17 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v17, 12, v17
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v9, 12, v9 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v9, 12, v9
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v10, 12, v10 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v10, 12, v10
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v2, 12, v2 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v2, 12, v2
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v16, 12, v16 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v16, 12, v16
; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v18, v17, v3 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v1, v17, v3
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 12, v8 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 12, v8
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v9 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v9
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v9, 12, v15 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v9, 12, v15
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v2 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v2
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v10, v16, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v10, v16, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v7, 12, v7 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v7, 12, v7
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v10, 12, v14 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v10, 12, v14
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v8, 12, v8 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v8, 12, v8
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v15, 12, v9 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v9, 12, v9
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v3, v2, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v3, v2, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v6, 12, v6 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v6, 12, v6
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v7 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v7
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v7, 12, v13 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v7, 12, v13
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v10 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v10
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v8, v15, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v8, v9, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v5, 12, v5 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v5, 12, v5
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 12, v12 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 12, v12
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v6, 12, v6 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v6, 12, v6
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v7, 12, v7 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v7, 12, v7
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v2, v3, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v2, v3, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v4, 12, v4 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v4, 12, v4
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v5 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v5
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v8
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v5, 12, v11 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v5, 12, v11
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v3, 12, v8
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v6, v7, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v6, v7, v1
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v4, 12, v4 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v4, 12, v4
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v7, 12, v5 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v5, 12, v5
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v2, v3, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v2, v3, v1
; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v4, v7, v1 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v1, v4, v5, v1
; GFX10-DL-XNACK-NEXT: global_store_byte v0, v1, s[0:1] ; GFX10-DL-XNACK-NEXT: global_store_byte v0, v1, s[0:1]
; GFX10-DL-XNACK-NEXT: s_endpgm ; GFX10-DL-XNACK-NEXT: s_endpgm
; ;
; GFX10-DL-NOXNACK-LABEL: idot8_acc8: ; GFX10-DL-NOXNACK-LABEL: idot8_acc8:
; GFX10-DL-NOXNACK: ; %bb.0: ; %entry ; GFX10-DL-NOXNACK: ; %bb.0: ; %entry
; GFX10-DL-NOXNACK-NEXT: s_clause 0x1 ; GFX10-DL-NOXNACK-NEXT: s_clause 0x1
; GFX10-DL-NOXNACK-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24 ; GFX10-DL-NOXNACK-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX10-DL-NOXNACK-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34 ; GFX10-DL-NOXNACK-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
Show All 13 Lines
; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v4, 28, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v4, 28, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v5, 24, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v5, 24, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v6, 20, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v6, 20, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v7, 16, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v7, 16, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v10, 4, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v10, 4, v1
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v1, 12, v1
; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v16, 4, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v16, 4, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v17, 12, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v17, 12, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v1, 12, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v11, 28, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v11, 28, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v10, 12, v10 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v10, 12, v10
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v16, 12, v16
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v12, 24, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v12, 24, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v16, 12, v16
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v13, 20, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v13, 20, v0
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v14, 16, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v14, 16, v0
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v0
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v0, 8, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v0, 8, v0
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v18, 12, v1 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v1
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v17, 12, v17 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v17, 12, v17
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v9, 12, v9 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v9, 12, v9
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v10, 12, v10 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v10, 12, v10
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v0, 12, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v0, 12, v0
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v16, 12, v16 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v16, 12, v16
; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v1, v18, v17, v3 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v1, v1, v17, v3
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v8, 12, v8
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v9 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v9
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v9, 12, v15 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v9, 12, v15
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v8, 12, v8
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v0, 12, v0 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v0, 12, v0
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v1, v10, v16, v1 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v1, v10, v16, v1
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v7, 12, v7 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v7, 12, v7
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v10, 12, v14 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v10, 12, v14
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v15, 12, v8 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v8, 12, v8
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v9, 12, v9 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v9, 12, v9
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v3, v0, v1 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v3, v0, v1
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v6, 12, v6 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v6, 12, v6
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v7 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v7
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v7, 12, v13 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v7, 12, v13
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v10 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v10
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v15, v9, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v8, v9, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v5, 12, v5 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v5, 12, v5
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v8, 12, v12 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v8, 12, v12
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v6, 12, v6 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v6, 12, v6
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v7, 12, v7 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v7, 12, v7
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v1, v3, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v1, v3, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v4, 12, v4
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v5 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v5
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v5, 12, v11 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v5, 12, v11
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v8 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v8
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v4, 12, v4
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v4, 12, v4
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v5, 12, v5 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v5, 12, v5
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v7, 12, v4
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v1, v3, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v1, v3, v0
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v7, v5, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NOXNACK-NEXT: global_store_byte v2, v0, s[0:1] ; GFX10-DL-NOXNACK-NEXT: global_store_byte v2, v0, s[0:1]
; GFX10-DL-NOXNACK-NEXT: s_endpgm ; GFX10-DL-NOXNACK-NEXT: s_endpgm
; GFX10-DL-LABEL: idot8_acc8: ; GFX10-DL-LABEL: idot8_acc8:
; GFX10-DL: ; %bb.0: ; %entry ; GFX10-DL: ; %bb.0: ; %entry
; GFX10-DL-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x34 ; GFX10-DL-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x34
; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0
; GFX10-DL-NEXT: s_mov_b32 s12, SCRATCH_RSRC_DWORD0 ; GFX10-DL-NEXT: s_mov_b32 s12, SCRATCH_RSRC_DWORD0
; GFX10-DL-NEXT: s_mov_b32 s13, SCRATCH_RSRC_DWORD1 ; GFX10-DL-NEXT: s_mov_b32 s13, SCRATCH_RSRC_DWORD1
▲ Show 20 Lines • Show All 352 Lines • ▼ Show 20 Lines
; GFX10-DL-XNACK-NEXT: v_bfe_i32 v6, v2, 8, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_i32 v6, v2, 8, 4
; GFX10-DL-XNACK-NEXT: v_bfe_i32 v7, v2, 0, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_i32 v7, v2, 0, 4
; GFX10-DL-XNACK-NEXT: v_bfe_i32 v8, v1, 12, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_i32 v8, v1, 12, 4
; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v3, v3, v4 ; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v3, v3, v4
; GFX10-DL-XNACK-NEXT: v_bfe_i32 v9, v2, 12, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_i32 v9, v2, 12, 4
; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v4, v5, v6 ; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v4, v5, v6
; GFX10-DL-XNACK-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-XNACK-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-XNACK-NEXT: v_mad_i32_i24 v5, v0, v7, s2 ; GFX10-DL-XNACK-NEXT: v_mad_i32_i24 v5, v0, v7, s2
; GFX10-DL-XNACK-NEXT: v_bfe_i32 v15, v1, 16, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_i32 v6, v1, 16, 4
; GFX10-DL-XNACK-NEXT: v_bfe_i32 v10, v2, 16, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_i32 v10, v2, 16, 4
; GFX10-DL-XNACK-NEXT: v_bfe_i32 v11, v1, 20, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_i32 v11, v1, 20, 4
; GFX10-DL-XNACK-NEXT: v_bfe_i32 v12, v2, 20, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_i32 v12, v2, 20, 4
; GFX10-DL-XNACK-NEXT: v_mad_i32_i24 v14, v0, v7, v5 ; GFX10-DL-XNACK-NEXT: v_mad_i32_i24 v0, v0, v7, v5
; GFX10-DL-XNACK-NEXT: v_bfe_i32 v7, v1, 24, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_i32 v7, v1, 24, 4
; GFX10-DL-XNACK-NEXT: v_bfe_i32 v13, v2, 24, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_i32 v13, v2, 24, 4
; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v6, v15, v10
; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v8, v8, v9 ; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v8, v8, v9
; GFX10-DL-XNACK-NEXT: v_add3_u32 v15, v14, v3, v4 ; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v6, v6, v10
; GFX10-DL-XNACK-NEXT: v_add3_u32 v0, v0, v3, v4
; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v3, v11, v12 ; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v3, v11, v12
; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v4, v7, v13 ; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v4, v7, v13
; GFX10-DL-XNACK-NEXT: v_ashrrev_i32_e32 v1, 28, v1 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i32_e32 v1, 28, v1
; GFX10-DL-XNACK-NEXT: v_ashrrev_i32_e32 v2, 28, v2 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i32_e32 v2, 28, v2
; GFX10-DL-XNACK-NEXT: v_add3_u32 v6, v15, v8, v6 ; GFX10-DL-XNACK-NEXT: v_add3_u32 v0, v0, v8, v6
; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v7, v1, v2 ; GFX10-DL-XNACK-NEXT: v_mul_i32_i24_e32 v1, v1, v2
; GFX10-DL-XNACK-NEXT: v_add3_u32 v0, v6, v3, v4 ; GFX10-DL-XNACK-NEXT: v_add3_u32 v0, v0, v3, v4
; GFX10-DL-XNACK-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-DL-XNACK-NEXT: v_mov_b32_e32 v2, 0
; GFX10-DL-XNACK-NEXT: v_add3_u32 v0, v0, v7, v5 ; GFX10-DL-XNACK-NEXT: v_add3_u32 v0, v0, v1, v5
; GFX10-DL-XNACK-NEXT: global_store_dword v2, v0, s[0:1] ; GFX10-DL-XNACK-NEXT: global_store_dword v2, v0, s[0:1]
; GFX10-DL-XNACK-NEXT: s_endpgm ; GFX10-DL-XNACK-NEXT: s_endpgm
; ;
; GFX10-DL-NOXNACK-LABEL: idot8_multiuses_mul1: ; GFX10-DL-NOXNACK-LABEL: idot8_multiuses_mul1:
; GFX10-DL-NOXNACK: ; %bb.0: ; %entry ; GFX10-DL-NOXNACK: ; %bb.0: ; %entry
; GFX10-DL-NOXNACK-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24 ; GFX10-DL-NOXNACK-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b32_e32 v0, 2, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX10-DL-NOXNACK-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34 ; GFX10-DL-NOXNACK-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
Show All 17 Lines
; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v6, v0, 8, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v6, v0, 8, 4
; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v7, v0, 0, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v7, v0, 0, 4
; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v8, v1, 12, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v8, v1, 12, 4
; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v3, v3, v4 ; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v3, v3, v4
; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v9, v0, 12, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v9, v0, 12, 4
; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v4, v5, v6 ; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v4, v5, v6
; GFX10-DL-NOXNACK-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NOXNACK-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NOXNACK-NEXT: v_mad_i32_i24 v5, v2, v7, s2 ; GFX10-DL-NOXNACK-NEXT: v_mad_i32_i24 v5, v2, v7, s2
; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v15, v1, 16, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v6, v1, 16, 4
; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v10, v0, 16, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v10, v0, 16, 4
; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v11, v1, 20, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v11, v1, 20, 4
; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v12, v0, 20, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v12, v0, 20, 4
; GFX10-DL-NOXNACK-NEXT: v_mad_i32_i24 v2, v2, v7, v5 ; GFX10-DL-NOXNACK-NEXT: v_mad_i32_i24 v2, v2, v7, v5
; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v7, v1, 24, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v7, v1, 24, 4
; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v13, v0, 24, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_i32 v13, v0, 24, 4
; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v6, v15, v10
; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v8, v8, v9 ; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v8, v8, v9
; GFX10-DL-NOXNACK-NEXT: v_add3_u32 v15, v2, v3, v4 ; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v6, v6, v10
; GFX10-DL-NOXNACK-NEXT: v_add3_u32 v2, v2, v3, v4
; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v3, v11, v12 ; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v3, v11, v12
; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v4, v7, v13 ; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v4, v7, v13
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i32_e32 v1, 28, v1 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i32_e32 v1, 28, v1
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i32_e32 v0, 28, v0 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i32_e32 v0, 28, v0
; GFX10-DL-NOXNACK-NEXT: v_add3_u32 v2, v15, v8, v6 ; GFX10-DL-NOXNACK-NEXT: v_add3_u32 v2, v2, v8, v6
; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v0, v1, v0 ; GFX10-DL-NOXNACK-NEXT: v_mul_i32_i24_e32 v0, v1, v0
; GFX10-DL-NOXNACK-NEXT: v_add3_u32 v3, v2, v3, v4 ; GFX10-DL-NOXNACK-NEXT: v_add3_u32 v1, v2, v3, v4
; GFX10-DL-NOXNACK-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-DL-NOXNACK-NEXT: v_mov_b32_e32 v2, 0
; GFX10-DL-NOXNACK-NEXT: v_add3_u32 v0, v3, v0, v5 ; GFX10-DL-NOXNACK-NEXT: v_add3_u32 v0, v1, v0, v5
; GFX10-DL-NOXNACK-NEXT: global_store_dword v2, v0, s[0:1] ; GFX10-DL-NOXNACK-NEXT: global_store_dword v2, v0, s[0:1]
; GFX10-DL-NOXNACK-NEXT: s_endpgm ; GFX10-DL-NOXNACK-NEXT: s_endpgm
; GFX10-DL-LABEL: idot8_multiuses_mul1: ; GFX10-DL-LABEL: idot8_multiuses_mul1:
; GFX10-DL: ; %bb.0: ; %entry ; GFX10-DL: ; %bb.0: ; %entry
; GFX10-DL-NEXT: s_mov_b32 s8, SCRATCH_RSRC_DWORD0 ; GFX10-DL-NEXT: s_mov_b32 s8, SCRATCH_RSRC_DWORD0
; GFX10-DL-NEXT: s_mov_b32 s9, SCRATCH_RSRC_DWORD1 ; GFX10-DL-NEXT: s_mov_b32 s9, SCRATCH_RSRC_DWORD1
; GFX10-DL-NEXT: s_mov_b32 s10, -1 ; GFX10-DL-NEXT: s_mov_b32 s10, -1
; GFX10-DL-NEXT: s_mov_b32 s11, 0x31c16000 ; GFX10-DL-NEXT: s_mov_b32 s11, 0x31c16000
▲ Show 20 Lines • Show All 750 Lines • ▼ Show 20 Lines
; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v11, v4, v11 ; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v11, v4, v11
; GFX10-DL-XNACK-NEXT: v_bfe_u32 v16, v2, 4, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_u32 v16, v2, 4, 4
; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v13, v4, v13 ; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v13, v4, v13
; GFX10-DL-XNACK-NEXT: v_bfe_u32 v18, v2, 8, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_u32 v18, v2, 8, 4
; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v9, v4, v9 ; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v9, v4, v9
; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v1, v1, 16, v11 ; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v1, v1, 16, v11
; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v7, v4, v7 ; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v7, v4, v7
; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v11, v16, 16, v13 ; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v11, v16, 16, v13
; GFX10-DL-XNACK-NEXT: v_bfe_u32 v19, v2, 24, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_u32 v12, v2, 24, 4
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v14, 28, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v14, 28, v2
; GFX10-DL-XNACK-NEXT: v_pk_lshlrev_b16 v1, 12, v1 op_sel_hi:[0,1] ; GFX10-DL-XNACK-NEXT: v_pk_lshlrev_b16 v1, 12, v1 op_sel_hi:[0,1]
; GFX10-DL-XNACK-NEXT: v_bfe_u32 v15, v2, 16, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_u32 v15, v2, 16, 4
; GFX10-DL-XNACK-NEXT: v_bfe_u32 v17, v2, 20, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_u32 v17, v2, 20, 4
; GFX10-DL-XNACK-NEXT: v_bfe_u32 v2, v2, 12, 4 ; GFX10-DL-XNACK-NEXT: v_bfe_u32 v2, v2, 12, 4
; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v13, v4, v18 ; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v13, v4, v18
; GFX10-DL-XNACK-NEXT: v_pk_lshlrev_b16 v11, 12, v11 op_sel_hi:[0,1] ; GFX10-DL-XNACK-NEXT: v_pk_lshlrev_b16 v11, 12, v11 op_sel_hi:[0,1]
; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v9, v10, 16, v9 ; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v9, v10, 16, v9
Show All 10 Lines
; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v8, v17, 16, v10 ; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v8, v17, 16, v10
; GFX10-DL-XNACK-NEXT: v_pk_ashrrev_i16 v7, 12, v7 op_sel_hi:[0,1] ; GFX10-DL-XNACK-NEXT: v_pk_ashrrev_i16 v7, 12, v7 op_sel_hi:[0,1]
; GFX10-DL-XNACK-NEXT: v_pk_ashrrev_i16 v2, 12, v2 op_sel_hi:[0,1] ; GFX10-DL-XNACK-NEXT: v_pk_ashrrev_i16 v2, 12, v2 op_sel_hi:[0,1]
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v10, 16, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v10, 16, v1
; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v3 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v3
; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v3, v4, v5 ; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v3, v4, v5
; GFX10-DL-XNACK-NEXT: v_pk_lshlrev_b16 v5, 12, v8 op_sel_hi:[0,1] ; GFX10-DL-XNACK-NEXT: v_pk_lshlrev_b16 v5, 12, v8 op_sel_hi:[0,1]
; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v4, v4, v19 ; GFX10-DL-XNACK-NEXT: v_and_b32_e32 v4, v4, v12
; GFX10-DL-XNACK-NEXT: v_pk_mul_lo_u16 v2, v9, v2 ; GFX10-DL-XNACK-NEXT: v_pk_mul_lo_u16 v2, v9, v2
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v10 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v10
; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v3, v6, 16, v3 ; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v3, v6, 16, v3
; GFX10-DL-XNACK-NEXT: v_pk_ashrrev_i16 v5, 12, v5 op_sel_hi:[0,1] ; GFX10-DL-XNACK-NEXT: v_pk_ashrrev_i16 v5, 12, v5 op_sel_hi:[0,1]
; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v4, v14, 16, v4 ; GFX10-DL-XNACK-NEXT: v_lshl_or_b32 v4, v14, 16, v4
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v6, 16, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v6, 16, v2
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v2 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v2
; GFX10-DL-XNACK-NEXT: v_pk_lshlrev_b16 v2, 12, v3 op_sel_hi:[0,1] ; GFX10-DL-XNACK-NEXT: v_pk_lshlrev_b16 v2, 12, v3 op_sel_hi:[0,1]
; GFX10-DL-XNACK-NEXT: v_pk_lshlrev_b16 v3, 12, v4 op_sel_hi:[0,1] ; GFX10-DL-XNACK-NEXT: v_pk_lshlrev_b16 v3, 12, v4 op_sel_hi:[0,1]
; GFX10-DL-XNACK-NEXT: v_pk_mul_lo_u16 v4, v7, v5 ; GFX10-DL-XNACK-NEXT: v_pk_mul_lo_u16 v4, v7, v5
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v6 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v6
; GFX10-DL-XNACK-NEXT: v_pk_ashrrev_i16 v2, 12, v2 op_sel_hi:[0,1] ; GFX10-DL-XNACK-NEXT: v_pk_ashrrev_i16 v2, 12, v2 op_sel_hi:[0,1]
; GFX10-DL-XNACK-NEXT: v_pk_ashrrev_i16 v3, 12, v3 op_sel_hi:[0,1] ; GFX10-DL-XNACK-NEXT: v_pk_ashrrev_i16 v3, 12, v3 op_sel_hi:[0,1]
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v5, 16, v4 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v5, 16, v4
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v7, v1, v4 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v4
; GFX10-DL-XNACK-NEXT: v_pk_mul_lo_u16 v2, v2, v3 ; GFX10-DL-XNACK-NEXT: v_pk_mul_lo_u16 v2, v2, v3
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v7, v5 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v5
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v3, 16, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v2 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v2
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v3 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v1, v1, v3
; GFX10-DL-XNACK-NEXT: global_store_short v0, v1, s[0:1] ; GFX10-DL-XNACK-NEXT: global_store_short v0, v1, s[0:1]
; GFX10-DL-XNACK-NEXT: s_endpgm ; GFX10-DL-XNACK-NEXT: s_endpgm
; ;
; GFX10-DL-NOXNACK-LABEL: idot8_acc16_vecMul: ; GFX10-DL-NOXNACK-LABEL: idot8_acc16_vecMul:
; GFX10-DL-NOXNACK: ; %bb.0: ; %entry ; GFX10-DL-NOXNACK: ; %bb.0: ; %entry
Show All 27 Lines
; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v11, v4, v11 ; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v11, v4, v11
; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v16, v0, 4, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v16, v0, 4, 4
; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v13, v4, v13 ; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v13, v4, v13
; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v18, v0, 8, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v18, v0, 8, 4
; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v9, v4, v9 ; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v9, v4, v9
; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v1, v1, 16, v11 ; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v1, v1, 16, v11
; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v7, v4, v7 ; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v7, v4, v7
; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v11, v16, 16, v13 ; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v11, v16, 16, v13
; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v19, v0, 24, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v12, v0, 24, 4
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v14, 28, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v14, 28, v0
; GFX10-DL-NOXNACK-NEXT: v_pk_lshlrev_b16 v1, 12, v1 op_sel_hi:[0,1] ; GFX10-DL-NOXNACK-NEXT: v_pk_lshlrev_b16 v1, 12, v1 op_sel_hi:[0,1]
; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v15, v0, 16, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v15, v0, 16, 4
; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v17, v0, 20, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v17, v0, 20, 4
; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v0, v0, 12, 4 ; GFX10-DL-NOXNACK-NEXT: v_bfe_u32 v0, v0, 12, 4
; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v13, v4, v18 ; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v13, v4, v18
; GFX10-DL-NOXNACK-NEXT: v_pk_lshlrev_b16 v11, 12, v11 op_sel_hi:[0,1] ; GFX10-DL-NOXNACK-NEXT: v_pk_lshlrev_b16 v11, 12, v11 op_sel_hi:[0,1]
; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v9, v10, 16, v9 ; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v9, v10, 16, v9
Show All 10 Lines
; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v8, v17, 16, v10 ; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v8, v17, 16, v10
; GFX10-DL-NOXNACK-NEXT: v_pk_ashrrev_i16 v7, 12, v7 op_sel_hi:[0,1] ; GFX10-DL-NOXNACK-NEXT: v_pk_ashrrev_i16 v7, 12, v7 op_sel_hi:[0,1]
; GFX10-DL-NOXNACK-NEXT: v_pk_ashrrev_i16 v0, 12, v0 op_sel_hi:[0,1] ; GFX10-DL-NOXNACK-NEXT: v_pk_ashrrev_i16 v0, 12, v0 op_sel_hi:[0,1]
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v10, 16, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v10, 16, v1
; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v1, v1, v3 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v1, v1, v3
; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v3, v4, v5 ; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v3, v4, v5
; GFX10-DL-NOXNACK-NEXT: v_pk_lshlrev_b16 v5, 12, v8 op_sel_hi:[0,1] ; GFX10-DL-NOXNACK-NEXT: v_pk_lshlrev_b16 v5, 12, v8 op_sel_hi:[0,1]
; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v4, v4, v19 ; GFX10-DL-NOXNACK-NEXT: v_and_b32_e32 v4, v4, v12
; GFX10-DL-NOXNACK-NEXT: v_pk_mul_lo_u16 v0, v9, v0 ; GFX10-DL-NOXNACK-NEXT: v_pk_mul_lo_u16 v0, v9, v0
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v1, v1, v10 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v1, v1, v10
; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v3, v6, 16, v3 ; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v3, v6, 16, v3
; GFX10-DL-NOXNACK-NEXT: v_pk_ashrrev_i16 v5, 12, v5 op_sel_hi:[0,1] ; GFX10-DL-NOXNACK-NEXT: v_pk_ashrrev_i16 v5, 12, v5 op_sel_hi:[0,1]
; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v4, v14, 16, v4 ; GFX10-DL-NOXNACK-NEXT: v_lshl_or_b32 v4, v14, 16, v4
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v6, 16, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v6, 16, v0
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v1, v0 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v1, v0
; GFX10-DL-NOXNACK-NEXT: v_pk_lshlrev_b16 v1, 12, v3 op_sel_hi:[0,1] ; GFX10-DL-NOXNACK-NEXT: v_pk_lshlrev_b16 v1, 12, v3 op_sel_hi:[0,1]
; GFX10-DL-NOXNACK-NEXT: v_pk_lshlrev_b16 v3, 12, v4 op_sel_hi:[0,1] ; GFX10-DL-NOXNACK-NEXT: v_pk_lshlrev_b16 v3, 12, v4 op_sel_hi:[0,1]
; GFX10-DL-NOXNACK-NEXT: v_pk_mul_lo_u16 v4, v7, v5 ; GFX10-DL-NOXNACK-NEXT: v_pk_mul_lo_u16 v4, v7, v5
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v7, v0, v6 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v6
; GFX10-DL-NOXNACK-NEXT: v_pk_ashrrev_i16 v1, 12, v1 op_sel_hi:[0,1] ; GFX10-DL-NOXNACK-NEXT: v_pk_ashrrev_i16 v1, 12, v1 op_sel_hi:[0,1]
; GFX10-DL-NOXNACK-NEXT: v_pk_ashrrev_i16 v3, 12, v3 op_sel_hi:[0,1] ; GFX10-DL-NOXNACK-NEXT: v_pk_ashrrev_i16 v3, 12, v3 op_sel_hi:[0,1]
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v5, 16, v4 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v5, 16, v4
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v7, v4 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v4
; GFX10-DL-NOXNACK-NEXT: v_pk_mul_lo_u16 v1, v1, v3 ; GFX10-DL-NOXNACK-NEXT: v_pk_mul_lo_u16 v1, v1, v3
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v5 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v5
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v3, 16, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v1 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v1
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v3 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v3
; GFX10-DL-NOXNACK-NEXT: global_store_short v2, v0, s[0:1] ; GFX10-DL-NOXNACK-NEXT: global_store_short v2, v0, s[0:1]
; GFX10-DL-NOXNACK-NEXT: s_endpgm ; GFX10-DL-NOXNACK-NEXT: s_endpgm
; GFX10-DL-LABEL: idot8_acc16_vecMul: ; GFX10-DL-LABEL: idot8_acc16_vecMul:
▲ Show 20 Lines • Show All 499 Lines • ▼ Show 20 Lines
; GFX9-DL-NEXT: global_store_byte v3, v0, s[2:3] ; GFX9-DL-NEXT: global_store_byte v3, v0, s[2:3]
; GFX9-DL-NEXT: s_endpgm ; GFX9-DL-NEXT: s_endpgm
; ;
; GFX10-DL-XNACK-LABEL: idot8_acc8_vecMul: ; GFX10-DL-XNACK-LABEL: idot8_acc8_vecMul:
; GFX10-DL-XNACK: ; %bb.0: ; %entry ; GFX10-DL-XNACK: ; %bb.0: ; %entry
; GFX10-DL-XNACK-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24 ; GFX10-DL-XNACK-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX10-DL-XNACK-NEXT: v_lshlrev_b32_e32 v0, 2, v0 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX10-DL-XNACK-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34 ; GFX10-DL-XNACK-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
; GFX10-DL-XNACK-NEXT: v_mov_b32_e32 v19, 0 ; GFX10-DL-XNACK-NEXT: v_mov_b32_e32 v4, 0
; GFX10-DL-XNACK-NEXT: s_mov_b32 s8, SCRATCH_RSRC_DWORD0 ; GFX10-DL-XNACK-NEXT: s_mov_b32 s8, SCRATCH_RSRC_DWORD0
; GFX10-DL-XNACK-NEXT: s_mov_b32 s9, SCRATCH_RSRC_DWORD1 ; GFX10-DL-XNACK-NEXT: s_mov_b32 s9, SCRATCH_RSRC_DWORD1
; GFX10-DL-XNACK-NEXT: s_mov_b32 s10, -1 ; GFX10-DL-XNACK-NEXT: s_mov_b32 s10, -1
; GFX10-DL-XNACK-NEXT: s_mov_b32 s11, 0x31c16000 ; GFX10-DL-XNACK-NEXT: s_mov_b32 s11, 0x31c16000
; GFX10-DL-XNACK-NEXT: s_add_u32 s8, s8, s3 ; GFX10-DL-XNACK-NEXT: s_add_u32 s8, s8, s3
; GFX10-DL-XNACK-NEXT: s_addc_u32 s9, s9, 0 ; GFX10-DL-XNACK-NEXT: s_addc_u32 s9, s9, 0
; GFX10-DL-XNACK-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-XNACK-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-XNACK-NEXT: s_clause 0x1 ; GFX10-DL-XNACK-NEXT: s_clause 0x1
; GFX10-DL-XNACK-NEXT: global_load_dword v1, v0, s[4:5] ; GFX10-DL-XNACK-NEXT: global_load_dword v1, v0, s[4:5]
; GFX10-DL-XNACK-NEXT: global_load_dword v2, v0, s[6:7] ; GFX10-DL-XNACK-NEXT: global_load_dword v2, v0, s[6:7]
; GFX10-DL-XNACK-NEXT: global_load_ubyte v3, v19, s[0:1] ; GFX10-DL-XNACK-NEXT: global_load_ubyte v3, v4, s[0:1]
; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1
; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v2
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v16, 8, v2 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v16, 8, v2
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v6, 28, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v6, 28, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 12, v8 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 12, v8
Show All 26 Lines
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v13, 12, v13 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v13, 12, v13
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v1, 12, v1 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v1, 12, v1
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v2, 12, v2 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v2, 12, v2
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v10, 12, v10 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v10, 12, v10
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v15, 12, v17 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v15, 12, v17
; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v9, v9, v16 ; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v9, v9, v16
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 8, v8 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v8, 8, v8
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v5, 12, v5 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v5, 12, v5
; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v6, v6, v13
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v12, 12, v12 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v12, 12, v12
; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v6, v6, v13
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v7, 12, v7 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v7, 12, v7
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v0, 12, v0 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v0, 12, v0
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v14, 12, v14 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v14, 12, v14
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v11, 12, v11 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v11, 12, v11
; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v10, v10, v15 ; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v10, v10, v15
; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v8, v9, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v8, v9, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v1, 12, v1 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v1, 12, v1
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v2 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v2, 12, v2
; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v9, v0, v11 ; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v9, v0, v11
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v5, 12, v5 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v5, 12, v5
; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v23, 12, v12 ; GFX10-DL-XNACK-NEXT: v_ashrrev_i16 v12, 12, v12
; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v11, v7, v14 ; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v11, v7, v14
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v6, 8, v6 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v6, 8, v6
; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v1, v1, v2 ; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v1, v1, v2
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v10, 8, v10 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v10, 8, v10
; GFX10-DL-XNACK-NEXT: v_lshlrev_b32_e32 v0, 16, v8 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b32_e32 v0, 16, v8
; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v2, v5, v23 ; GFX10-DL-XNACK-NEXT: v_mul_lo_u16 v2, v5, v12
; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v9, 8, v9 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b16 v9, 8, v9
; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v6, v11, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v6, v11, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v1, v1, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v1, v1, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v11, v10, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD ; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v11, v10, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v2, v2, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v2, v2, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-DL-XNACK-NEXT: v_lshlrev_b32_e32 v9, 16, v6 ; GFX10-DL-XNACK-NEXT: v_lshlrev_b32_e32 v9, 16, v6
; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-XNACK-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v3, v1, v3 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v3, v1, v3
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v10, 8, v11 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v10, 8, v11
; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v1, v2, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD ; GFX10-DL-XNACK-NEXT: v_or_b32_sdwa v1, v2, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v9, v3, v10 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v9, v3, v10
; GFX10-DL-XNACK-NEXT: v_lshrrev_b64 v[2:3], 24, v[0:1] ; GFX10-DL-XNACK-NEXT: v_lshrrev_b64 v[2:3], 24, v[0:1]
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v1, 8, v1 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v1, 8, v1
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v0, v9, v8 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v0, v9, v8
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v0, v0, v2 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v0, v0, v2
; GFX10-DL-XNACK-NEXT: v_mad_u16 v0, v5, v23, v0 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v0, v5, v12, v0
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v0, v0, v1 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v0, v0, v1
; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v1, 8, v6 ; GFX10-DL-XNACK-NEXT: v_lshrrev_b32_e32 v1, 8, v6
; GFX10-DL-XNACK-NEXT: v_mad_u16 v0, v7, v14, v0 ; GFX10-DL-XNACK-NEXT: v_mad_u16 v0, v7, v14, v0
; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v0, v0, v1 ; GFX10-DL-XNACK-NEXT: v_add_nc_u16 v0, v0, v1
; GFX10-DL-XNACK-NEXT: global_store_byte v19, v0, s[0:1] ; GFX10-DL-XNACK-NEXT: global_store_byte v4, v0, s[0:1]
; GFX10-DL-XNACK-NEXT: s_endpgm ; GFX10-DL-XNACK-NEXT: s_endpgm
; ;
; GFX10-DL-NOXNACK-LABEL: idot8_acc8_vecMul: ; GFX10-DL-NOXNACK-LABEL: idot8_acc8_vecMul:
; GFX10-DL-NOXNACK: ; %bb.0: ; %entry ; GFX10-DL-NOXNACK: ; %bb.0: ; %entry
; GFX10-DL-NOXNACK-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24 ; GFX10-DL-NOXNACK-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b32_e32 v0, 2, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX10-DL-NOXNACK-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34 ; GFX10-DL-NOXNACK-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
; GFX10-DL-NOXNACK-NEXT: v_mov_b32_e32 v19, 0 ; GFX10-DL-NOXNACK-NEXT: v_mov_b32_e32 v4, 0
; GFX10-DL-NOXNACK-NEXT: s_mov_b32 s8, SCRATCH_RSRC_DWORD0 ; GFX10-DL-NOXNACK-NEXT: s_mov_b32 s8, SCRATCH_RSRC_DWORD0
; GFX10-DL-NOXNACK-NEXT: s_mov_b32 s9, SCRATCH_RSRC_DWORD1 ; GFX10-DL-NOXNACK-NEXT: s_mov_b32 s9, SCRATCH_RSRC_DWORD1
; GFX10-DL-NOXNACK-NEXT: s_mov_b32 s10, -1 ; GFX10-DL-NOXNACK-NEXT: s_mov_b32 s10, -1
; GFX10-DL-NOXNACK-NEXT: s_mov_b32 s11, 0x31c16000 ; GFX10-DL-NOXNACK-NEXT: s_mov_b32 s11, 0x31c16000
; GFX10-DL-NOXNACK-NEXT: s_add_u32 s8, s8, s3 ; GFX10-DL-NOXNACK-NEXT: s_add_u32 s8, s8, s3
; GFX10-DL-NOXNACK-NEXT: s_addc_u32 s9, s9, 0 ; GFX10-DL-NOXNACK-NEXT: s_addc_u32 s9, s9, 0
; GFX10-DL-NOXNACK-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NOXNACK-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NOXNACK-NEXT: s_clause 0x1 ; GFX10-DL-NOXNACK-NEXT: s_clause 0x1
; GFX10-DL-NOXNACK-NEXT: global_load_dword v1, v0, s[4:5] ; GFX10-DL-NOXNACK-NEXT: global_load_dword v1, v0, s[4:5]
; GFX10-DL-NOXNACK-NEXT: global_load_dword v0, v0, s[6:7] ; GFX10-DL-NOXNACK-NEXT: global_load_dword v0, v0, s[6:7]
; GFX10-DL-NOXNACK-NEXT: global_load_ubyte v2, v19, s[0:1] ; GFX10-DL-NOXNACK-NEXT: global_load_ubyte v2, v4, s[0:1]
; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v8, 12, v1
; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NOXNACK-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v15, 12, v0
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v9, 8, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v11, 20, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v11, 20, v0
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v12, 16, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v12, 16, v0
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v13, 28, v0 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v13, 28, v0
Show All 22 Lines
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v5, 16, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v5, 16, v1
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v7, 12, v7 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v7, 12, v7
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v3, 12, v3 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v3, 12, v3
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v14, 12, v14 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v14, 12, v14
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v11, 12, v11 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v11, 12, v11
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v6, 12, v6 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v6, 12, v6
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v13, 12, v13 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v13, 12, v13
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v1, 12, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v1, 12, v1
; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v23, v9, v0 ; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v0, v9, v0
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v8, 8, v8 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v8, 8, v8
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v10, 12, v10 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v10, 12, v10
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v15, 12, v17 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v15, 12, v17
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v5, 12, v5 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v5, 12, v5
; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v6, v6, v13 ; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v6, v6, v13
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v12, 12, v12 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v12, 12, v12
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v9, 12, v11 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v9, 12, v11
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v3 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v3, 12, v3
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v7, 12, v7 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v7, 12, v7
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v14, 12, v14 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v14, 12, v14
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v1 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v1, 12, v1
; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v10, v10, v15 ; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v10, v10, v15
; GFX10-DL-NOXNACK-NEXT: v_or_b32_sdwa v8, v23, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; GFX10-DL-NOXNACK-NEXT: v_or_b32_sdwa v8, v0, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v3, v3, v9 ; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v3, v3, v9
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v5, 12, v5 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v5, 12, v5
; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v11, 12, v12 ; GFX10-DL-NOXNACK-NEXT: v_ashrrev_i16 v11, 12, v12
; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v9, v7, v14 ; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v9, v7, v14
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v6, 8, v6 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v6, 8, v6
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v10, 8, v10 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b16 v10, 8, v10
; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v1, v1, v18 ; GFX10-DL-NOXNACK-NEXT: v_mul_lo_u16 v1, v1, v18
; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b32_e32 v0, 16, v8 ; GFX10-DL-NOXNACK-NEXT: v_lshlrev_b32_e32 v0, 16, v8
Show All 13 Lines
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v1, 8, v1 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v1, 8, v1
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v9, v8 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v9, v8
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v2 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v2
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v5, v11, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v5, v11, v0
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v1 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v1
; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v1, 8, v6 ; GFX10-DL-NOXNACK-NEXT: v_lshrrev_b32_e32 v1, 8, v6
; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v7, v14, v0 ; GFX10-DL-NOXNACK-NEXT: v_mad_u16 v0, v7, v14, v0
; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v1 ; GFX10-DL-NOXNACK-NEXT: v_add_nc_u16 v0, v0, v1
; GFX10-DL-NOXNACK-NEXT: global_store_byte v19, v0, s[0:1] ; GFX10-DL-NOXNACK-NEXT: global_store_byte v4, v0, s[0:1]
; GFX10-DL-NOXNACK-NEXT: s_endpgm ; GFX10-DL-NOXNACK-NEXT: s_endpgm
; GFX10-DL-LABEL: idot8_acc8_vecMul: ; GFX10-DL-LABEL: idot8_acc8_vecMul:
; GFX10-DL: ; %bb.0: ; %entry ; GFX10-DL: ; %bb.0: ; %entry
; GFX10-DL-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x34 ; GFX10-DL-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x34
; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0
; GFX10-DL-NEXT: s_mov_b32 s20, SCRATCH_RSRC_DWORD0 ; GFX10-DL-NEXT: s_mov_b32 s20, SCRATCH_RSRC_DWORD0
; GFX10-DL-NEXT: s_mov_b32 s21, SCRATCH_RSRC_DWORD1 ; GFX10-DL-NEXT: s_mov_b32 s21, SCRATCH_RSRC_DWORD1
; GFX10-DL-NEXT: s_mov_b32 s22, -1 ; GFX10-DL-NEXT: s_mov_b32 s22, -1
▲ Show 20 Lines • Show All 126 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/idot8u.ll

Show First 20 Lines • Show All 488 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0
; GFX10-DL-NEXT: s_addc_u32 s9, s9, 0 ; GFX10-DL-NEXT: s_addc_u32 s9, s9, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7]
; GFX10-DL-NEXT: global_load_ushort v4, v1, s[2:3] ; GFX10-DL-NEXT: global_load_ushort v4, v1, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_and_b32_e32 v11, 15, v2 ; GFX10-DL-NEXT: v_and_b32_e32 v0, 15, v2
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_and_b32_e32 v5, 15, v3 ; GFX10-DL-NEXT: v_and_b32_e32 v5, 15, v3
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 4, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 4, 4
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v4 ; GFX10-DL-NEXT: v_mad_u16 v0, v0, v5, v4
; GFX10-DL-NEXT: v_bfe_u32 v11, v2, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v4, v2, 8, 4
; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 8, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 12, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 12, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_bfe_u32 v11, v2, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v4, v2, 16, 4
; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 16, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 20, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 20, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_bfe_u32 v11, v2, 24, 4 ; GFX10-DL-NEXT: v_bfe_u32 v4, v2, 24, 4
; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 24, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 24, 4
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 28, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 28, v3
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0
; GFX10-DL-NEXT: global_store_short v1, v0, s[2:3] ; GFX10-DL-NEXT: global_store_short v1, v0, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<8 x i4> addrspace(1)* %src2, <8 x i4> addrspace(1)* %src2,
i16 addrspace(1)* nocapture %dst) { i16 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
%gep1 = getelementptr <8 x i4>, <8 x i4> addrspace(1)* %src1, i32 %idx %gep1 = getelementptr <8 x i4>, <8 x i4> addrspace(1)* %src1, i32 %idx
▲ Show 20 Lines • Show All 277 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0
; GFX10-DL-NEXT: s_addc_u32 s9, s9, 0 ; GFX10-DL-NEXT: s_addc_u32 s9, s9, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7]
; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3] ; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_and_b32_e32 v11, 15, v2 ; GFX10-DL-NEXT: v_and_b32_e32 v0, 15, v2
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_and_b32_e32 v5, 15, v3 ; GFX10-DL-NEXT: v_and_b32_e32 v5, 15, v3
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 4, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 4, 4
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v4 ; GFX10-DL-NEXT: v_mad_u16 v0, v0, v5, v4
; GFX10-DL-NEXT: v_bfe_u32 v11, v2, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v4, v2, 8, 4
; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 8, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 12, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 12, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_bfe_u32 v11, v2, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v4, v2, 16, 4
; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 16, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 20, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 20, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_bfe_u32 v11, v2, 24, 4 ; GFX10-DL-NEXT: v_bfe_u32 v4, v2, 24, 4
; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 24, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 24, 4
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 28, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 28, v3
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0
; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3] ; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<8 x i4> addrspace(1)* %src2, <8 x i4> addrspace(1)* %src2,
i8 addrspace(1)* nocapture %dst) { i8 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
%gep1 = getelementptr <8 x i4>, <8 x i4> addrspace(1)* %src1, i32 %idx %gep1 = getelementptr <8 x i4>, <8 x i4> addrspace(1)* %src1, i32 %idx
▲ Show 20 Lines • Show All 281 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0
; GFX10-DL-NEXT: s_addc_u32 s9, s9, 0 ; GFX10-DL-NEXT: s_addc_u32 s9, s9, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7]
; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3] ; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_and_b32_e32 v11, 15, v2 ; GFX10-DL-NEXT: v_and_b32_e32 v0, 15, v2
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_and_b32_e32 v5, 15, v3 ; GFX10-DL-NEXT: v_and_b32_e32 v5, 15, v3
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 4, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 4, 4
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v4 ; GFX10-DL-NEXT: v_mad_u16 v0, v0, v5, v4
; GFX10-DL-NEXT: v_bfe_u32 v11, v2, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v4, v2, 8, 4
; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 8, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 12, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 12, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_bfe_u32 v11, v2, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v4, v2, 16, 4
; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 16, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 20, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 20, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v11, 24, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v4, 24, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 24, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 24, v3
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 28, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 28, v3
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0
; GFX10-DL-NEXT: v_and_b32_e32 v0, 15, v0 ; GFX10-DL-NEXT: v_and_b32_e32 v0, 15, v0
; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3] ; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<8 x i4> addrspace(1)* %src2, <8 x i4> addrspace(1)* %src2,
i4 addrspace(1)* nocapture %dst) { i4 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
▲ Show 20 Lines • Show All 266 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v1, 0
; GFX10-DL-NEXT: s_addc_u32 s9, s9, 0 ; GFX10-DL-NEXT: s_addc_u32 s9, s9, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v3, v0, s[6:7]
; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3] ; GFX10-DL-NEXT: global_load_ubyte v4, v1, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_and_b32_e32 v11, 15, v2 ; GFX10-DL-NEXT: v_and_b32_e32 v0, 15, v2
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_and_b32_e32 v5, 15, v3 ; GFX10-DL-NEXT: v_and_b32_e32 v5, 15, v3
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 4, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 4, 4
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v4 ; GFX10-DL-NEXT: v_mad_u16 v0, v0, v5, v4
; GFX10-DL-NEXT: v_bfe_u32 v11, v2, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v4, v2, 8, 4
; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 8, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 12, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 12, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_bfe_u32 v11, v2, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v4, v2, 16, 4
; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v3, 16, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 20, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v3, 20, 4
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v11, 24, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v4, 24, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 24, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 24, v3
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 28, v3 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 28, v3
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v7, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v11, v5, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v4, v5, v0
; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v2, v3, v0
; GFX10-DL-NEXT: v_and_b32_e32 v0, 15, v0 ; GFX10-DL-NEXT: v_and_b32_e32 v0, 15, v0
; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3] ; GFX10-DL-NEXT: global_store_byte v1, v0, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<8 x i4> addrspace(1)* %src2, <8 x i4> addrspace(1)* %src2,
i4 addrspace(1)* nocapture %dst) { i4 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
▲ Show 20 Lines • Show All 891 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v0, 0
; GFX10-DL-NEXT: global_load_ushort v3, v0, s[0:1] ; GFX10-DL-NEXT: global_load_ushort v3, v0, s[0:1]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_and_b32_e32 v7, 15, v1 ; GFX10-DL-NEXT: v_and_b32_e32 v7, 15, v1
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_and_b32_e32 v6, 15, v2 ; GFX10-DL-NEXT: v_and_b32_e32 v6, 15, v2
; GFX10-DL-NEXT: v_bfe_u32 v9, v1, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v9, v1, 4, 4
; GFX10-DL-NEXT: v_bfe_u32 v15, v2, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v10, v2, 4, 4
; GFX10-DL-NEXT: v_bfe_u32 v13, v2, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v12, v1, 8, 4
; GFX10-DL-NEXT: v_and_b32_e32 v7, v4, v7 ; GFX10-DL-NEXT: v_and_b32_e32 v7, v4, v7
; GFX10-DL-NEXT: v_and_b32_e32 v6, v4, v6 ; GFX10-DL-NEXT: v_and_b32_e32 v6, v4, v6
; GFX10-DL-NEXT: v_bfe_u32 v19, v1, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v13, v2, 8, 4
; GFX10-DL-NEXT: v_bfe_u32 v10, v2, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v11, v1, 16, 4
; GFX10-DL-NEXT: v_and_b32_e32 v13, v4, v13 ; GFX10-DL-NEXT: v_and_b32_e32 v12, v4, v12
; GFX10-DL-NEXT: v_lshl_or_b32 v7, v9, 16, v7 ; GFX10-DL-NEXT: v_lshl_or_b32 v7, v9, 16, v7
; GFX10-DL-NEXT: v_lshl_or_b32 v6, v15, 16, v6 ; GFX10-DL-NEXT: v_lshl_or_b32 v6, v10, 16, v6
; GFX10-DL-NEXT: v_bfe_u32 v9, v1, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v9, v1, 12, 4
; GFX10-DL-NEXT: v_and_b32_e32 v12, v4, v19 ; GFX10-DL-NEXT: v_bfe_u32 v10, v2, 12, 4
; GFX10-DL-NEXT: v_bfe_u32 v11, v1, 16, 4 ; GFX10-DL-NEXT: v_and_b32_e32 v13, v4, v13
; GFX10-DL-NEXT: v_lshl_or_b32 v10, v10, 16, v13 ; GFX10-DL-NEXT: v_bfe_u32 v5, v1, 24, 4
; GFX10-DL-NEXT: v_pk_mul_lo_u16 v6, v7, v6 ; GFX10-DL-NEXT: v_pk_mul_lo_u16 v6, v7, v6
; GFX10-DL-NEXT: v_bfe_u32 v7, v2, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v2, 16, 4
; GFX10-DL-NEXT: v_lshl_or_b32 v9, v9, 16, v12 ; GFX10-DL-NEXT: v_lshl_or_b32 v9, v9, 16, v12
; GFX10-DL-NEXT: v_bfe_u32 v5, v1, 24, 4 ; GFX10-DL-NEXT: v_lshl_or_b32 v10, v10, 16, v13
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v23, 28, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v8, 28, v1
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v12, 16, v6 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v12, 16, v6
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_add_nc_u16 v3, v6, v3 ; GFX10-DL-NEXT: v_add_nc_u16 v3, v6, v3
; GFX10-DL-NEXT: v_pk_mul_lo_u16 v9, v9, v10
; GFX10-DL-NEXT: v_bfe_u32 v1, v1, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v1, v1, 20, 4
; GFX10-DL-NEXT: v_and_b32_e32 v11, v4, v11 ; GFX10-DL-NEXT: v_and_b32_e32 v11, v4, v11
; GFX10-DL-NEXT: v_pk_mul_lo_u16 v9, v9, v10
; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v2, 20, 4
; GFX10-DL-NEXT: v_and_b32_e32 v7, v4, v7
; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v12 ; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v12
; GFX10-DL-NEXT: v_and_b32_e32 v7, v4, v7
; GFX10-DL-NEXT: v_bfe_u32 v10, v2, 24, 4 ; GFX10-DL-NEXT: v_bfe_u32 v10, v2, 24, 4
; GFX10-DL-NEXT: v_lshl_or_b32 v1, v1, 16, v11 ; GFX10-DL-NEXT: v_lshl_or_b32 v1, v1, 16, v11
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2
; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v9
; GFX10-DL-NEXT: v_lshl_or_b32 v6, v6, 16, v7 ; GFX10-DL-NEXT: v_lshl_or_b32 v6, v6, 16, v7
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v9 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 16, v9
; GFX10-DL-NEXT: v_add_nc_u16 v14, v3, v9
; GFX10-DL-NEXT: v_and_b32_e32 v9, v4, v10 ; GFX10-DL-NEXT: v_and_b32_e32 v9, v4, v10
; GFX10-DL-NEXT: v_and_b32_e32 v4, v4, v5 ; GFX10-DL-NEXT: v_and_b32_e32 v4, v4, v5
; GFX10-DL-NEXT: v_pk_mul_lo_u16 v1, v1, v6 ; GFX10-DL-NEXT: v_pk_mul_lo_u16 v1, v1, v6
; GFX10-DL-NEXT: v_add_nc_u16 v3, v14, v7 ; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v7
; GFX10-DL-NEXT: v_lshl_or_b32 v2, v2, 16, v9 ; GFX10-DL-NEXT: v_lshl_or_b32 v2, v2, 16, v9
; GFX10-DL-NEXT: v_lshl_or_b32 v4, v23, 16, v4 ; GFX10-DL-NEXT: v_lshl_or_b32 v4, v8, 16, v4
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 16, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 16, v1
; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v1 ; GFX10-DL-NEXT: v_add_nc_u16 v1, v3, v1
; GFX10-DL-NEXT: v_pk_mul_lo_u16 v2, v4, v2 ; GFX10-DL-NEXT: v_pk_mul_lo_u16 v2, v4, v2
; GFX10-DL-NEXT: v_add_nc_u16 v1, v3, v5 ; GFX10-DL-NEXT: v_add_nc_u16 v1, v1, v5
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 16, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; GFX10-DL-NEXT: v_add_nc_u16 v1, v1, v2 ; GFX10-DL-NEXT: v_add_nc_u16 v1, v1, v2
; GFX10-DL-NEXT: v_add_nc_u16 v1, v1, v3 ; GFX10-DL-NEXT: v_add_nc_u16 v1, v1, v3
; GFX10-DL-NEXT: global_store_short v0, v1, s[0:1] ; GFX10-DL-NEXT: global_store_short v0, v1, s[0:1]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<8 x i4> addrspace(1)* %src2, <8 x i4> addrspace(1)* %src2,
i16 addrspace(1)* nocapture %dst) { i16 addrspace(1)* nocapture %dst) {
entry: entry:
▲ Show 20 Lines • Show All 330 Lines • ▼ Show 20 Lines
; GFX9-DL-NEXT: global_store_byte v3, v0, s[2:3] ; GFX9-DL-NEXT: global_store_byte v3, v0, s[2:3]
; GFX9-DL-NEXT: s_endpgm ; GFX9-DL-NEXT: s_endpgm
; ;
; GFX10-DL-LABEL: udot8_acc8_vecMul: ; GFX10-DL-LABEL: udot8_acc8_vecMul:
; GFX10-DL: ; %bb.0: ; %entry ; GFX10-DL: ; %bb.0: ; %entry
; GFX10-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24 ; GFX10-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0 ; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX10-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34 ; GFX10-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
; GFX10-DL-NEXT: v_mov_b32_e32 v19, 0 ; GFX10-DL-NEXT: v_mov_b32_e32 v4, 0
; GFX10-DL-NEXT: s_mov_b32 s8, SCRATCH_RSRC_DWORD0 ; GFX10-DL-NEXT: s_mov_b32 s8, SCRATCH_RSRC_DWORD0
; GFX10-DL-NEXT: s_mov_b32 s9, SCRATCH_RSRC_DWORD1 ; GFX10-DL-NEXT: s_mov_b32 s9, SCRATCH_RSRC_DWORD1
; GFX10-DL-NEXT: s_mov_b32 s10, -1 ; GFX10-DL-NEXT: s_mov_b32 s10, -1
; GFX10-DL-NEXT: s_mov_b32 s11, 0x31c16000 ; GFX10-DL-NEXT: s_mov_b32 s11, 0x31c16000
; GFX10-DL-NEXT: s_add_u32 s8, s8, s3 ; GFX10-DL-NEXT: s_add_u32 s8, s8, s3
; GFX10-DL-NEXT: s_addc_u32 s9, s9, 0 ; GFX10-DL-NEXT: s_addc_u32 s9, s9, 0
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: s_clause 0x1 ; GFX10-DL-NEXT: s_clause 0x1
; GFX10-DL-NEXT: global_load_dword v1, v0, s[4:5] ; GFX10-DL-NEXT: global_load_dword v1, v0, s[4:5]
; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7] ; GFX10-DL-NEXT: global_load_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: global_load_ubyte v3, v19, s[0:1] ; GFX10-DL-NEXT: global_load_ubyte v3, v4, s[0:1]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_bfe_u32 v9, v1, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v9, v1, 12, 4
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_bfe_u32 v10, v2, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v10, v2, 12, 4
; GFX10-DL-NEXT: v_bfe_u32 v8, v1, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v8, v1, 8, 4
; GFX10-DL-NEXT: v_bfe_u32 v13, v2, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v13, v2, 8, 4
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 28, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v7, 28, v1
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v14, 28, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v14, 28, v2
; GFX10-DL-NEXT: v_mul_lo_u16 v9, v9, v10 ; GFX10-DL-NEXT: v_mul_lo_u16 v9, v9, v10
; GFX10-DL-NEXT: v_bfe_u32 v5, v1, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v1, 16, 4
; GFX10-DL-NEXT: v_mul_lo_u16 v8, v8, v13 ; GFX10-DL-NEXT: v_mul_lo_u16 v8, v8, v13
; GFX10-DL-NEXT: v_bfe_u32 v0, v1, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v0, v1, 20, 4
; GFX10-DL-NEXT: v_bfe_u32 v6, v1, 24, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v1, 24, 4
; GFX10-DL-NEXT: v_and_b32_e32 v11, 15, v1 ; GFX10-DL-NEXT: v_and_b32_e32 v11, 15, v1
; GFX10-DL-NEXT: v_lshlrev_b16 v9, 8, v9 ; GFX10-DL-NEXT: v_lshlrev_b16 v9, 8, v9
; GFX10-DL-NEXT: v_bfe_u32 v1, v1, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v1, v1, 4, 4
; GFX10-DL-NEXT: v_bfe_u32 v15, v2, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v15, v2, 4, 4
; GFX10-DL-NEXT: v_mul_lo_u16 v7, v7, v14 ; GFX10-DL-NEXT: v_mul_lo_u16 v7, v7, v14
; GFX10-DL-NEXT: v_bfe_u32 v10, v2, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v10, v2, 20, 4
; GFX10-DL-NEXT: v_bfe_u32 v13, v2, 24, 4 ; GFX10-DL-NEXT: v_bfe_u32 v13, v2, 24, 4
; GFX10-DL-NEXT: v_bfe_u32 v23, v2, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v12, v2, 16, 4
; GFX10-DL-NEXT: v_and_b32_e32 v2, 15, v2 ; GFX10-DL-NEXT: v_and_b32_e32 v2, 15, v2
; GFX10-DL-NEXT: v_mul_lo_u16 v1, v1, v15 ; GFX10-DL-NEXT: v_mul_lo_u16 v1, v1, v15
; GFX10-DL-NEXT: v_or_b32_e32 v8, v8, v9 ; GFX10-DL-NEXT: v_or_b32_e32 v8, v8, v9
; GFX10-DL-NEXT: v_mul_lo_u16 v9, v0, v10 ; GFX10-DL-NEXT: v_mul_lo_u16 v9, v0, v10
; GFX10-DL-NEXT: v_mul_lo_u16 v10, v6, v13 ; GFX10-DL-NEXT: v_mul_lo_u16 v10, v6, v13
; GFX10-DL-NEXT: v_lshlrev_b16 v7, 8, v7 ; GFX10-DL-NEXT: v_lshlrev_b16 v7, 8, v7
; GFX10-DL-NEXT: v_mul_lo_u16 v2, v11, v2 ; GFX10-DL-NEXT: v_mul_lo_u16 v2, v11, v2
; GFX10-DL-NEXT: v_lshlrev_b16 v1, 8, v1 ; GFX10-DL-NEXT: v_lshlrev_b16 v1, 8, v1
; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 16, v8 ; GFX10-DL-NEXT: v_lshlrev_b32_e32 v0, 16, v8
; GFX10-DL-NEXT: v_mul_lo_u16 v11, v5, v23 ; GFX10-DL-NEXT: v_mul_lo_u16 v11, v5, v12
; GFX10-DL-NEXT: v_or_b32_e32 v7, v10, v7 ; GFX10-DL-NEXT: v_or_b32_e32 v7, v10, v7
; GFX10-DL-NEXT: v_lshlrev_b16 v9, 8, v9 ; GFX10-DL-NEXT: v_lshlrev_b16 v9, 8, v9
; GFX10-DL-NEXT: v_or_b32_sdwa v10, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD ; GFX10-DL-NEXT: v_or_b32_sdwa v10, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX10-DL-NEXT: v_or_b32_e32 v1, v2, v1 ; GFX10-DL-NEXT: v_or_b32_e32 v1, v2, v1
; GFX10-DL-NEXT: v_or_b32_e32 v2, v11, v9 ; GFX10-DL-NEXT: v_or_b32_e32 v2, v11, v9
; GFX10-DL-NEXT: v_lshlrev_b32_e32 v9, 16, v7 ; GFX10-DL-NEXT: v_lshlrev_b32_e32 v9, 16, v7
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v10, 8, v10 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v10, 8, v10
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_add_nc_u16 v3, v1, v3 ; GFX10-DL-NEXT: v_add_nc_u16 v3, v1, v3
; GFX10-DL-NEXT: v_or_b32_sdwa v1, v2, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD ; GFX10-DL-NEXT: v_or_b32_sdwa v1, v2, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX10-DL-NEXT: v_add_nc_u16 v9, v3, v10 ; GFX10-DL-NEXT: v_add_nc_u16 v9, v3, v10
; GFX10-DL-NEXT: v_lshrrev_b64 v[2:3], 24, v[0:1] ; GFX10-DL-NEXT: v_lshrrev_b64 v[2:3], 24, v[0:1]
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 8, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 8, v1
; GFX10-DL-NEXT: v_add_nc_u16 v0, v9, v8 ; GFX10-DL-NEXT: v_add_nc_u16 v0, v9, v8
; GFX10-DL-NEXT: v_add_nc_u16 v0, v0, v2 ; GFX10-DL-NEXT: v_add_nc_u16 v0, v0, v2
; GFX10-DL-NEXT: v_mad_u16 v0, v5, v23, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v5, v12, v0
; GFX10-DL-NEXT: v_add_nc_u16 v0, v0, v1 ; GFX10-DL-NEXT: v_add_nc_u16 v0, v0, v1
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 8, v7 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 8, v7
; GFX10-DL-NEXT: v_mad_u16 v0, v6, v13, v0 ; GFX10-DL-NEXT: v_mad_u16 v0, v6, v13, v0
; GFX10-DL-NEXT: v_add_nc_u16 v0, v0, v1 ; GFX10-DL-NEXT: v_add_nc_u16 v0, v0, v1
; GFX10-DL-NEXT: global_store_byte v19, v0, s[0:1] ; GFX10-DL-NEXT: global_store_byte v4, v0, s[0:1]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<8 x i4> addrspace(1)* %src2, <8 x i4> addrspace(1)* %src2,
i8 addrspace(1)* nocapture %dst) { i8 addrspace(1)* nocapture %dst) {
entry: entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x() %idx = call i32 @llvm.amdgcn.workitem.id.x()
%gep1 = getelementptr <8 x i4>, <8 x i4> addrspace(1)* %src1, i32 %idx %gep1 = getelementptr <8 x i4>, <8 x i4> addrspace(1)* %src1, i32 %idx
%vec1 = load <8 x i4>, <8 x i4> addrspace(1)* %gep1 %vec1 = load <8 x i4>, <8 x i4> addrspace(1)* %gep1
%gep2 = getelementptr <8 x i4>, <8 x i4> addrspace(1)* %src2, i32 %idx %gep2 = getelementptr <8 x i4>, <8 x i4> addrspace(1)* %src2, i32 %idx
▲ Show 20 Lines • Show All 273 Lines • ▼ Show 20 Lines
; GFX10-DL-NEXT: global_load_ubyte v3, v0, s[2:3] ; GFX10-DL-NEXT: global_load_ubyte v3, v0, s[2:3]
; GFX10-DL-NEXT: s_waitcnt vmcnt(2) ; GFX10-DL-NEXT: s_waitcnt vmcnt(2)
; GFX10-DL-NEXT: v_and_b32_e32 v4, 15, v1 ; GFX10-DL-NEXT: v_and_b32_e32 v4, 15, v1
; GFX10-DL-NEXT: s_waitcnt vmcnt(1) ; GFX10-DL-NEXT: s_waitcnt vmcnt(1)
; GFX10-DL-NEXT: v_and_b32_e32 v5, 15, v2 ; GFX10-DL-NEXT: v_and_b32_e32 v5, 15, v2
; GFX10-DL-NEXT: v_bfe_u32 v6, v1, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v1, 4, 4
; GFX10-DL-NEXT: v_bfe_u32 v7, v2, 4, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v2, 4, 4
; GFX10-DL-NEXT: v_bfe_u32 v8, v2, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v8, v2, 8, 4
; GFX10-DL-NEXT: v_bfe_u32 v11, v1, 24, 4
; GFX10-DL-NEXT: v_mul_u32_u24_e32 v4, v4, v5 ; GFX10-DL-NEXT: v_mul_u32_u24_e32 v4, v4, v5
; GFX10-DL-NEXT: v_bfe_u32 v5, v1, 8, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v1, 8, 4
; GFX10-DL-NEXT: v_mul_u32_u24_e32 v6, v6, v7 ; GFX10-DL-NEXT: v_mul_u32_u24_e32 v6, v6, v7
; GFX10-DL-NEXT: v_bfe_u32 v7, v2, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v2, 12, 4
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) ; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_add_nc_u16 v3, v4, v3 ; GFX10-DL-NEXT: v_add_nc_u16 v3, v4, v3
; GFX10-DL-NEXT: v_bfe_u32 v4, v1, 12, 4 ; GFX10-DL-NEXT: v_bfe_u32 v4, v1, 12, 4
; GFX10-DL-NEXT: v_mul_u32_u24_e32 v5, v5, v8 ; GFX10-DL-NEXT: v_mul_u32_u24_e32 v5, v5, v8
; GFX10-DL-NEXT: v_bfe_u32 v8, v2, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v8, v2, 16, 4
; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v6 ; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v6
; GFX10-DL-NEXT: v_bfe_u32 v6, v1, 16, 4 ; GFX10-DL-NEXT: v_bfe_u32 v6, v1, 16, 4
; GFX10-DL-NEXT: v_mul_u32_u24_e32 v4, v4, v7 ; GFX10-DL-NEXT: v_mul_u32_u24_e32 v4, v4, v7
; GFX10-DL-NEXT: v_bfe_u32 v7, v2, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v7, v2, 20, 4
; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v5 ; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v5
; GFX10-DL-NEXT: v_bfe_u32 v5, v1, 20, 4 ; GFX10-DL-NEXT: v_bfe_u32 v5, v1, 20, 4
; GFX10-DL-NEXT: v_mul_u32_u24_e32 v6, v6, v8 ; GFX10-DL-NEXT: v_mul_u32_u24_e32 v6, v6, v8
; GFX10-DL-NEXT: v_bfe_u32 v8, v2, 24, 4 ; GFX10-DL-NEXT: v_bfe_u32 v8, v2, 24, 4
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 28, v1 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2
; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v4 ; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v4
; GFX10-DL-NEXT: v_bfe_u32 v4, v1, 24, 4
; GFX10-DL-NEXT: v_mul_u32_u24_e32 v5, v5, v7 ; GFX10-DL-NEXT: v_mul_u32_u24_e32 v5, v5, v7
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v2, 28, v2 ; GFX10-DL-NEXT: v_lshrrev_b32_e32 v1, 28, v1
; GFX10-DL-NEXT: v_mul_u32_u24_e32 v4, v11, v8
; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v6 ; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v6
; GFX10-DL-NEXT: v_mul_u32_u24_e32 v4, v4, v8
; GFX10-DL-NEXT: v_mul_u32_u24_e32 v1, v1, v2 ; GFX10-DL-NEXT: v_mul_u32_u24_e32 v1, v1, v2
; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v5 ; GFX10-DL-NEXT: v_add_nc_u16 v3, v3, v5
; GFX10-DL-NEXT: v_add_nc_u16 v2, v3, v4 ; GFX10-DL-NEXT: v_add_nc_u16 v2, v3, v4
; GFX10-DL-NEXT: v_add_nc_u16 v1, v2, v1 ; GFX10-DL-NEXT: v_add_nc_u16 v1, v2, v1
; GFX10-DL-NEXT: v_and_b32_e32 v1, 15, v1 ; GFX10-DL-NEXT: v_and_b32_e32 v1, 15, v1
; GFX10-DL-NEXT: global_store_byte v0, v1, s[2:3] ; GFX10-DL-NEXT: global_store_byte v0, v1, s[2:3]
; GFX10-DL-NEXT: s_endpgm ; GFX10-DL-NEXT: s_endpgm
<8 x i4> addrspace(1)* %src2, <8 x i4> addrspace(1)* %src2,
▲ Show 20 Lines • Show All 266 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.dim.ll

Show First 20 Lines • Show All 442 Lines • ▼ Show 20 Lines
; NOPRT-NEXT: v_mov_b32_e32 v5, 0 ; NOPRT-NEXT: v_mov_b32_e32 v5, 0
; NOPRT-NEXT: s_waitcnt vmcnt(0) ; NOPRT-NEXT: s_waitcnt vmcnt(0)
; NOPRT-NEXT: global_store_dword v5, v4, s[8:9] ; NOPRT-NEXT: global_store_dword v5, v4, s[8:9]
; NOPRT-NEXT: s_waitcnt vmcnt(0) ; NOPRT-NEXT: s_waitcnt vmcnt(0)
; NOPRT-NEXT: ; return to shader part epilog ; NOPRT-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: load_3d_tfe_lwe: ; GFX10-LABEL: load_3d_tfe_lwe:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v11, 0 ; encoding: [0x80,0x02,0x16,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v8, 0 ; encoding: [0x80,0x02,0x10,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; encoding: [0x02,0x03,0x0e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; encoding: [0x02,0x03,0x0e,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; encoding: [0x01,0x03,0x0c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; encoding: [0x01,0x03,0x0c,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v12, v11 ; encoding: [0x0b,0x03,0x18,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v9, v8 ; encoding: [0x08,0x03,0x12,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v13, v11 ; encoding: [0x0b,0x03,0x1a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v10, v8 ; encoding: [0x08,0x03,0x14,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v14, v11 ; encoding: [0x0b,0x03,0x1c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v11, v8 ; encoding: [0x08,0x03,0x16,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v15, v11 ; encoding: [0x0b,0x03,0x1e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v12, v8 ; encoding: [0x08,0x03,0x18,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v0, v11 ; encoding: [0x0b,0x03,0x00,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v0, v8 ; encoding: [0x08,0x03,0x00,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v1, v12 ; encoding: [0x0c,0x03,0x02,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v1, v9 ; encoding: [0x09,0x03,0x02,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v2, v13 ; encoding: [0x0d,0x03,0x04,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v2, v10 ; encoding: [0x0a,0x03,0x04,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v3, v14 ; encoding: [0x0e,0x03,0x06,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v3, v11 ; encoding: [0x0b,0x03,0x06,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v4, v15 ; encoding: [0x0f,0x03,0x08,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v4, v12 ; encoding: [0x0c,0x03,0x08,0x7e]
; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm tfe lwe ; encoding: [0x10,0x1f,0x03,0xf0,0x05,0x00,0x00,0x00] ; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm tfe lwe ; encoding: [0x10,0x1f,0x03,0xf0,0x05,0x00,0x00,0x00]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: global_store_dword v11, v4, s[8:9] ; encoding: [0x00,0x80,0x70,0xdc,0x0b,0x04,0x08,0x00] ; GFX10-NEXT: global_store_dword v8, v4, s[8:9] ; encoding: [0x00,0x80,0x70,0xdc,0x08,0x04,0x08,0x00]
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb] ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call {<4 x float>,i32} @llvm.amdgcn.image.load.3d.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 3, i32 0) %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.3d.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 3, i32 0)
%v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.vec = extractvalue {<4 x float>, i32} %v, 0
%v.err = extractvalue {<4 x float>, i32} %v, 1 %v.err = extractvalue {<4 x float>, i32} %v, 1
store i32 %v.err, i32 addrspace(1)* %out, align 4 store i32 %v.err, i32 addrspace(1)* %out, align 4
ret <4 x float> %v.vec ret <4 x float> %v.vec
▲ Show 20 Lines • Show All 99 Lines • ▼ Show 20 Lines
; NOPRT-NEXT: v_mov_b32_e32 v5, 0 ; NOPRT-NEXT: v_mov_b32_e32 v5, 0
; NOPRT-NEXT: s_waitcnt vmcnt(0) ; NOPRT-NEXT: s_waitcnt vmcnt(0)
; NOPRT-NEXT: global_store_dword v5, v4, s[8:9] ; NOPRT-NEXT: global_store_dword v5, v4, s[8:9]
; NOPRT-NEXT: s_waitcnt vmcnt(0) ; NOPRT-NEXT: s_waitcnt vmcnt(0)
; NOPRT-NEXT: ; return to shader part epilog ; NOPRT-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: load_cube_lwe: ; GFX10-LABEL: load_cube_lwe:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v11, 0 ; encoding: [0x80,0x02,0x16,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v8, 0 ; encoding: [0x80,0x02,0x10,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; encoding: [0x02,0x03,0x0e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; encoding: [0x02,0x03,0x0e,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; encoding: [0x01,0x03,0x0c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; encoding: [0x01,0x03,0x0c,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v12, v11 ; encoding: [0x0b,0x03,0x18,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v9, v8 ; encoding: [0x08,0x03,0x12,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v13, v11 ; encoding: [0x0b,0x03,0x1a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v10, v8 ; encoding: [0x08,0x03,0x14,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v14, v11 ; encoding: [0x0b,0x03,0x1c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v11, v8 ; encoding: [0x08,0x03,0x16,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v15, v11 ; encoding: [0x0b,0x03,0x1e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v12, v8 ; encoding: [0x08,0x03,0x18,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v0, v11 ; encoding: [0x0b,0x03,0x00,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v0, v8 ; encoding: [0x08,0x03,0x00,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v1, v12 ; encoding: [0x0c,0x03,0x02,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v1, v9 ; encoding: [0x09,0x03,0x02,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v2, v13 ; encoding: [0x0d,0x03,0x04,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v2, v10 ; encoding: [0x0a,0x03,0x04,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v3, v14 ; encoding: [0x0e,0x03,0x06,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v3, v11 ; encoding: [0x0b,0x03,0x06,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v4, v15 ; encoding: [0x0f,0x03,0x08,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v4, v12 ; encoding: [0x0c,0x03,0x08,0x7e]
; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm lwe ; encoding: [0x18,0x1f,0x02,0xf0,0x05,0x00,0x00,0x00] ; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm lwe ; encoding: [0x18,0x1f,0x02,0xf0,0x05,0x00,0x00,0x00]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: global_store_dword v11, v4, s[8:9] ; encoding: [0x00,0x80,0x70,0xdc,0x0b,0x04,0x08,0x00] ; GFX10-NEXT: global_store_dword v8, v4, s[8:9] ; encoding: [0x00,0x80,0x70,0xdc,0x08,0x04,0x08,0x00]
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb] ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call {<4 x float>,i32} @llvm.amdgcn.image.load.cube.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0) %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.cube.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0)
%v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.vec = extractvalue {<4 x float>, i32} %v, 0
%v.err = extractvalue {<4 x float>, i32} %v, 1 %v.err = extractvalue {<4 x float>, i32} %v, 1
store i32 %v.err, i32 addrspace(1)* %out, align 4 store i32 %v.err, i32 addrspace(1)* %out, align 4
ret <4 x float> %v.vec ret <4 x float> %v.vec
▲ Show 20 Lines • Show All 226 Lines • ▼ Show 20 Lines
; NOPRT-NEXT: v_mov_b32_e32 v5, 0 ; NOPRT-NEXT: v_mov_b32_e32 v5, 0
; NOPRT-NEXT: s_waitcnt vmcnt(0) ; NOPRT-NEXT: s_waitcnt vmcnt(0)
; NOPRT-NEXT: global_store_dword v5, v4, s[8:9] ; NOPRT-NEXT: global_store_dword v5, v4, s[8:9]
; NOPRT-NEXT: s_waitcnt vmcnt(0) ; NOPRT-NEXT: s_waitcnt vmcnt(0)
; NOPRT-NEXT: ; return to shader part epilog ; NOPRT-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: load_2darray_lwe: ; GFX10-LABEL: load_2darray_lwe:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v11, 0 ; encoding: [0x80,0x02,0x16,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v8, 0 ; encoding: [0x80,0x02,0x10,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; encoding: [0x02,0x03,0x0e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; encoding: [0x02,0x03,0x0e,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; encoding: [0x01,0x03,0x0c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; encoding: [0x01,0x03,0x0c,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v12, v11 ; encoding: [0x0b,0x03,0x18,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v9, v8 ; encoding: [0x08,0x03,0x12,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v13, v11 ; encoding: [0x0b,0x03,0x1a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v10, v8 ; encoding: [0x08,0x03,0x14,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v14, v11 ; encoding: [0x0b,0x03,0x1c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v11, v8 ; encoding: [0x08,0x03,0x16,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v15, v11 ; encoding: [0x0b,0x03,0x1e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v12, v8 ; encoding: [0x08,0x03,0x18,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v0, v11 ; encoding: [0x0b,0x03,0x00,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v0, v8 ; encoding: [0x08,0x03,0x00,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v1, v12 ; encoding: [0x0c,0x03,0x02,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v1, v9 ; encoding: [0x09,0x03,0x02,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v2, v13 ; encoding: [0x0d,0x03,0x04,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v2, v10 ; encoding: [0x0a,0x03,0x04,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v3, v14 ; encoding: [0x0e,0x03,0x06,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v3, v11 ; encoding: [0x0b,0x03,0x06,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v4, v15 ; encoding: [0x0f,0x03,0x08,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v4, v12 ; encoding: [0x0c,0x03,0x08,0x7e]
; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm lwe ; encoding: [0x28,0x1f,0x02,0xf0,0x05,0x00,0x00,0x00] ; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm lwe ; encoding: [0x28,0x1f,0x02,0xf0,0x05,0x00,0x00,0x00]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: global_store_dword v11, v4, s[8:9] ; encoding: [0x00,0x80,0x70,0xdc,0x0b,0x04,0x08,0x00] ; GFX10-NEXT: global_store_dword v8, v4, s[8:9] ; encoding: [0x00,0x80,0x70,0xdc,0x08,0x04,0x08,0x00]
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb] ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2darray.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0) %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2darray.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0)
%v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.vec = extractvalue {<4 x float>, i32} %v, 0
%v.err = extractvalue {<4 x float>, i32} %v, 1 %v.err = extractvalue {<4 x float>, i32} %v, 1
store i32 %v.err, i32 addrspace(1)* %out, align 4 store i32 %v.err, i32 addrspace(1)* %out, align 4
ret <4 x float> %v.vec ret <4 x float> %v.vec
▲ Show 20 Lines • Show All 99 Lines • ▼ Show 20 Lines
; NOPRT-NEXT: v_mov_b32_e32 v5, 0 ; NOPRT-NEXT: v_mov_b32_e32 v5, 0
; NOPRT-NEXT: s_waitcnt vmcnt(0) ; NOPRT-NEXT: s_waitcnt vmcnt(0)
; NOPRT-NEXT: global_store_dword v5, v4, s[8:9] ; NOPRT-NEXT: global_store_dword v5, v4, s[8:9]
; NOPRT-NEXT: s_waitcnt vmcnt(0) ; NOPRT-NEXT: s_waitcnt vmcnt(0)
; NOPRT-NEXT: ; return to shader part epilog ; NOPRT-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: load_2dmsaa_both: ; GFX10-LABEL: load_2dmsaa_both:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v11, 0 ; encoding: [0x80,0x02,0x16,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v8, 0 ; encoding: [0x80,0x02,0x10,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; encoding: [0x02,0x03,0x0e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; encoding: [0x02,0x03,0x0e,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; encoding: [0x01,0x03,0x0c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; encoding: [0x01,0x03,0x0c,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v12, v11 ; encoding: [0x0b,0x03,0x18,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v9, v8 ; encoding: [0x08,0x03,0x12,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v13, v11 ; encoding: [0x0b,0x03,0x1a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v10, v8 ; encoding: [0x08,0x03,0x14,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v14, v11 ; encoding: [0x0b,0x03,0x1c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v11, v8 ; encoding: [0x08,0x03,0x16,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v15, v11 ; encoding: [0x0b,0x03,0x1e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v12, v8 ; encoding: [0x08,0x03,0x18,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v0, v11 ; encoding: [0x0b,0x03,0x00,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v0, v8 ; encoding: [0x08,0x03,0x00,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v1, v12 ; encoding: [0x0c,0x03,0x02,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v1, v9 ; encoding: [0x09,0x03,0x02,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v2, v13 ; encoding: [0x0d,0x03,0x04,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v2, v10 ; encoding: [0x0a,0x03,0x04,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v3, v14 ; encoding: [0x0e,0x03,0x06,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v3, v11 ; encoding: [0x0b,0x03,0x06,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v4, v15 ; encoding: [0x0f,0x03,0x08,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v4, v12 ; encoding: [0x0c,0x03,0x08,0x7e]
; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm tfe lwe ; encoding: [0x30,0x1f,0x03,0xf0,0x05,0x00,0x00,0x00] ; GFX10-NEXT: image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm tfe lwe ; encoding: [0x30,0x1f,0x03,0xf0,0x05,0x00,0x00,0x00]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: global_store_dword v11, v4, s[8:9] ; encoding: [0x00,0x80,0x70,0xdc,0x0b,0x04,0x08,0x00] ; GFX10-NEXT: global_store_dword v8, v4, s[8:9] ; encoding: [0x00,0x80,0x70,0xdc,0x08,0x04,0x08,0x00]
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb] ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2dmsaa.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 3, i32 0) %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2dmsaa.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 3, i32 0)
%v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.vec = extractvalue {<4 x float>, i32} %v, 0
%v.err = extractvalue {<4 x float>, i32} %v, 1 %v.err = extractvalue {<4 x float>, i32} %v, 1
store i32 %v.err, i32 addrspace(1)* %out, align 4 store i32 %v.err, i32 addrspace(1)* %out, align 4
ret <4 x float> %v.vec ret <4 x float> %v.vec
▲ Show 20 Lines • Show All 361 Lines • ▼ Show 20 Lines
; NOPRT-NEXT: v_mov_b32_e32 v5, 0 ; NOPRT-NEXT: v_mov_b32_e32 v5, 0
; NOPRT-NEXT: s_waitcnt vmcnt(0) ; NOPRT-NEXT: s_waitcnt vmcnt(0)
; NOPRT-NEXT: global_store_dword v5, v4, s[8:9] ; NOPRT-NEXT: global_store_dword v5, v4, s[8:9]
; NOPRT-NEXT: s_waitcnt vmcnt(0) ; NOPRT-NEXT: s_waitcnt vmcnt(0)
; NOPRT-NEXT: ; return to shader part epilog ; NOPRT-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: load_mip_2d_tfe: ; GFX10-LABEL: load_mip_2d_tfe:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v11, 0 ; encoding: [0x80,0x02,0x16,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v8, 0 ; encoding: [0x80,0x02,0x10,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; encoding: [0x02,0x03,0x0e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v7, v2 ; encoding: [0x02,0x03,0x0e,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; encoding: [0x01,0x03,0x0c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v6, v1 ; encoding: [0x01,0x03,0x0c,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v12, v11 ; encoding: [0x0b,0x03,0x18,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v9, v8 ; encoding: [0x08,0x03,0x12,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v13, v11 ; encoding: [0x0b,0x03,0x1a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v10, v8 ; encoding: [0x08,0x03,0x14,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v14, v11 ; encoding: [0x0b,0x03,0x1c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v11, v8 ; encoding: [0x08,0x03,0x16,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v15, v11 ; encoding: [0x0b,0x03,0x1e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v12, v8 ; encoding: [0x08,0x03,0x18,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v0, v11 ; encoding: [0x0b,0x03,0x00,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v0, v8 ; encoding: [0x08,0x03,0x00,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v1, v12 ; encoding: [0x0c,0x03,0x02,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v1, v9 ; encoding: [0x09,0x03,0x02,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v2, v13 ; encoding: [0x0d,0x03,0x04,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v2, v10 ; encoding: [0x0a,0x03,0x04,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v3, v14 ; encoding: [0x0e,0x03,0x06,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v3, v11 ; encoding: [0x0b,0x03,0x06,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v4, v15 ; encoding: [0x0f,0x03,0x08,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v4, v12 ; encoding: [0x0c,0x03,0x08,0x7e]
; GFX10-NEXT: image_load_mip v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm tfe ; encoding: [0x08,0x1f,0x05,0xf0,0x05,0x00,0x00,0x00] ; GFX10-NEXT: image_load_mip v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm tfe ; encoding: [0x08,0x1f,0x05,0xf0,0x05,0x00,0x00,0x00]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: global_store_dword v11, v4, s[8:9] ; encoding: [0x00,0x80,0x70,0xdc,0x0b,0x04,0x08,0x00] ; GFX10-NEXT: global_store_dword v8, v4, s[8:9] ; encoding: [0x00,0x80,0x70,0xdc,0x08,0x04,0x08,0x00]
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb] ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call {<4 x float>,i32} @llvm.amdgcn.image.load.mip.2d.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 1, i32 0) %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.mip.2d.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 1, i32 0)
%v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.vec = extractvalue {<4 x float>, i32} %v, 0
%v.err = extractvalue {<4 x float>, i32} %v, 1 %v.err = extractvalue {<4 x float>, i32} %v, 1
store i32 %v.err, i32 addrspace(1)* %out, align 4 store i32 %v.err, i32 addrspace(1)* %out, align 4
ret <4 x float> %v.vec ret <4 x float> %v.vec
▲ Show 20 Lines • Show All 2,165 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.a16.dim.ll

Show First 20 Lines • Show All 560 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_lshl_or_b32 v4, v5, 16, v4 ; GFX9-NEXT: v_lshl_or_b32 v4, v5, 16, v4
; GFX9-NEXT: v_lshl_or_b32 v2, v1, 16, v0 ; GFX9-NEXT: v_lshl_or_b32 v2, v1, 16, v0
; GFX9-NEXT: image_sample_d v[0:3], v[2:4], s[0:7], s[8:11] dmask:0xf a16 ; GFX9-NEXT: image_sample_d v[0:3], v[2:4], s[0:7], s[8:11] dmask:0xf a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: sample_d_2d: ; GFX10-LABEL: sample_d_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v6, 0xffff
; GFX10-NEXT: v_and_b32_e32 v4, v7, v4 ; GFX10-NEXT: v_and_b32_e32 v4, v6, v4
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; GFX10-NEXT: v_and_b32_e32 v2, v6, v2
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; GFX10-NEXT: v_and_b32_e32 v0, v6, v0
; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4 ; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4
; GFX10-NEXT: v_lshl_or_b32 v3, v3, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v3, v3, 16, v2
; GFX10-NEXT: v_lshl_or_b32 v2, v1, 16, v0 ; GFX10-NEXT: v_lshl_or_b32 v2, v1, 16, v0
; GFX10-NEXT: image_sample_d v[0:3], v[2:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; GFX10-NEXT: image_sample_d v[0:3], v[2:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 64 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_lshl_or_b32 v2, v4, 16, v2 ; GFX9-NEXT: v_lshl_or_b32 v2, v4, 16, v2
; GFX9-NEXT: v_lshl_or_b32 v1, v8, 16, v1 ; GFX9-NEXT: v_lshl_or_b32 v1, v8, 16, v1
; GFX9-NEXT: image_sample_c_d v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf a16 ; GFX9-NEXT: image_sample_c_d v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: sample_c_d_2d: ; GFX10-LABEL: sample_c_d_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v10, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_and_b32_e32 v5, v10, v5 ; GFX10-NEXT: v_and_b32_e32 v5, v7, v5
; GFX10-NEXT: v_and_b32_e32 v3, v10, v3 ; GFX10-NEXT: v_and_b32_e32 v3, v7, v3
; GFX10-NEXT: v_and_b32_e32 v1, v10, v1 ; GFX10-NEXT: v_and_b32_e32 v1, v7, v1
; GFX10-NEXT: v_lshl_or_b32 v5, v6, 16, v5 ; GFX10-NEXT: v_lshl_or_b32 v5, v6, 16, v5
; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3 ; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3
; GFX10-NEXT: v_lshl_or_b32 v2, v2, 16, v1 ; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1
; GFX10-NEXT: image_sample_c_d v[0:3], [v0, v2, v3, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; GFX10-NEXT: image_sample_c_d v[0:3], [v0, v1, v3, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f32.f16(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f32.f16(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, half %s, half %clamp) { define amdgpu_ps <4 x float> @sample_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, half %s, half %clamp) {
Show All 33 Lines
; ;
; GFX10-LABEL: sample_d_cl_2d: ; GFX10-LABEL: sample_d_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_and_b32_e32 v4, v7, v4 ; GFX10-NEXT: v_and_b32_e32 v4, v7, v4
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; GFX10-NEXT: v_and_b32_e32 v2, v7, v2
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; GFX10-NEXT: v_and_b32_e32 v0, v7, v0
; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4 ; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4
; GFX10-NEXT: v_lshl_or_b32 v5, v3, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2
; GFX10-NEXT: v_lshl_or_b32 v3, v1, 16, v0 ; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v0
; GFX10-NEXT: image_sample_d_cl v[0:3], [v3, v5, v4, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; GFX10-NEXT: image_sample_d_cl v[0:3], [v0, v2, v4, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, half %s, half %clamp) { define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, half %s, half %clamp) {
Show All 36 Lines
; GFX10-LABEL: sample_c_d_cl_2d: ; GFX10-LABEL: sample_c_d_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v8, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v8, 0xffff
; GFX10-NEXT: v_and_b32_e32 v5, v8, v5 ; GFX10-NEXT: v_and_b32_e32 v5, v8, v5
; GFX10-NEXT: v_and_b32_e32 v3, v8, v3 ; GFX10-NEXT: v_and_b32_e32 v3, v8, v3
; GFX10-NEXT: v_and_b32_e32 v1, v8, v1 ; GFX10-NEXT: v_and_b32_e32 v1, v8, v1
; GFX10-NEXT: v_lshl_or_b32 v5, v6, 16, v5 ; GFX10-NEXT: v_lshl_or_b32 v5, v6, 16, v5
; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3 ; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3
; GFX10-NEXT: v_lshl_or_b32 v2, v2, 16, v1 ; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1
; GFX10-NEXT: image_sample_c_d_cl v[0:3], [v0, v2, v3, v5, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; GFX10-NEXT: image_sample_c_d_cl v[0:3], [v0, v1, v3, v5, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.2d.v4f32.f32.f16(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.2d.v4f32.f32.f16(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_cd_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, half %s) { define amdgpu_ps <4 x float> @sample_cd_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, half %s) {
Show All 24 Lines
; GFX9-NEXT: v_lshl_or_b32 v4, v5, 16, v4 ; GFX9-NEXT: v_lshl_or_b32 v4, v5, 16, v4
; GFX9-NEXT: v_lshl_or_b32 v2, v1, 16, v0 ; GFX9-NEXT: v_lshl_or_b32 v2, v1, 16, v0
; GFX9-NEXT: image_sample_cd v[0:3], v[2:4], s[0:7], s[8:11] dmask:0xf a16 ; GFX9-NEXT: image_sample_cd v[0:3], v[2:4], s[0:7], s[8:11] dmask:0xf a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: sample_cd_2d: ; GFX10-LABEL: sample_cd_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v6, 0xffff
; GFX10-NEXT: v_and_b32_e32 v4, v7, v4 ; GFX10-NEXT: v_and_b32_e32 v4, v6, v4
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; GFX10-NEXT: v_and_b32_e32 v2, v6, v2
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; GFX10-NEXT: v_and_b32_e32 v0, v6, v0
; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4 ; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4
; GFX10-NEXT: v_lshl_or_b32 v3, v3, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v3, v3, 16, v2
; GFX10-NEXT: v_lshl_or_b32 v2, v1, 16, v0 ; GFX10-NEXT: v_lshl_or_b32 v2, v1, 16, v0
; GFX10-NEXT: image_sample_cd v[0:3], v[2:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; GFX10-NEXT: image_sample_cd v[0:3], v[2:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
Show All 30 Lines
; GFX9-NEXT: v_lshl_or_b32 v2, v4, 16, v2 ; GFX9-NEXT: v_lshl_or_b32 v2, v4, 16, v2
; GFX9-NEXT: v_lshl_or_b32 v1, v8, 16, v1 ; GFX9-NEXT: v_lshl_or_b32 v1, v8, 16, v1
; GFX9-NEXT: image_sample_c_cd v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf a16 ; GFX9-NEXT: image_sample_c_cd v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf a16
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: sample_c_cd_2d: ; GFX10-LABEL: sample_c_cd_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v10, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_and_b32_e32 v5, v10, v5 ; GFX10-NEXT: v_and_b32_e32 v5, v7, v5
; GFX10-NEXT: v_and_b32_e32 v3, v10, v3 ; GFX10-NEXT: v_and_b32_e32 v3, v7, v3
; GFX10-NEXT: v_and_b32_e32 v1, v10, v1 ; GFX10-NEXT: v_and_b32_e32 v1, v7, v1
; GFX10-NEXT: v_lshl_or_b32 v5, v6, 16, v5 ; GFX10-NEXT: v_lshl_or_b32 v5, v6, 16, v5
; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3 ; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3
; GFX10-NEXT: v_lshl_or_b32 v2, v2, 16, v1 ; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1
; GFX10-NEXT: image_sample_c_cd v[0:3], [v0, v2, v3, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; GFX10-NEXT: image_sample_c_cd v[0:3], [v0, v1, v3, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f32.f16(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f32.f16(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, half %s, half %clamp) { define amdgpu_ps <4 x float> @sample_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, half %s, half %clamp) {
Show All 33 Lines
; ;
; GFX10-LABEL: sample_cd_cl_2d: ; GFX10-LABEL: sample_cd_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_and_b32_e32 v4, v7, v4 ; GFX10-NEXT: v_and_b32_e32 v4, v7, v4
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; GFX10-NEXT: v_and_b32_e32 v2, v7, v2
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; GFX10-NEXT: v_and_b32_e32 v0, v7, v0
; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4 ; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4
; GFX10-NEXT: v_lshl_or_b32 v5, v3, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2
; GFX10-NEXT: v_lshl_or_b32 v3, v1, 16, v0 ; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v0
; GFX10-NEXT: image_sample_cd_cl v[0:3], [v3, v5, v4, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; GFX10-NEXT: image_sample_cd_cl v[0:3], [v0, v2, v4, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, half %s, half %clamp) { define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, half %s, half %clamp) {
Show All 36 Lines
; GFX10-LABEL: sample_c_cd_cl_2d: ; GFX10-LABEL: sample_c_cd_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v8, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v8, 0xffff
; GFX10-NEXT: v_and_b32_e32 v5, v8, v5 ; GFX10-NEXT: v_and_b32_e32 v5, v8, v5
; GFX10-NEXT: v_and_b32_e32 v3, v8, v3 ; GFX10-NEXT: v_and_b32_e32 v3, v8, v3
; GFX10-NEXT: v_and_b32_e32 v1, v8, v1 ; GFX10-NEXT: v_and_b32_e32 v1, v8, v1
; GFX10-NEXT: v_lshl_or_b32 v5, v6, 16, v5 ; GFX10-NEXT: v_lshl_or_b32 v5, v6, 16, v5
; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3 ; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3
; GFX10-NEXT: v_lshl_or_b32 v2, v2, 16, v1 ; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1
; GFX10-NEXT: image_sample_c_cd_cl v[0:3], [v0, v2, v3, v5, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; GFX10-NEXT: image_sample_c_cd_cl v[0:3], [v0, v1, v3, v5, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.2d.v4f32.f32.f16(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.2d.v4f32.f32.f16(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_l_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %lod) { define amdgpu_ps <4 x float> @sample_l_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %lod) {
▲ Show 20 Lines • Show All 178 Lines • ▼ Show 20 Lines
; GFX10-LABEL: sample_c_d_o_2darray_V1: ; GFX10-LABEL: sample_c_d_o_2darray_V1:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v9, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v9, 0xffff
; GFX10-NEXT: v_and_b32_e32 v6, v9, v6 ; GFX10-NEXT: v_and_b32_e32 v6, v9, v6
; GFX10-NEXT: v_and_b32_e32 v4, v9, v4 ; GFX10-NEXT: v_and_b32_e32 v4, v9, v4
; GFX10-NEXT: v_and_b32_e32 v2, v9, v2 ; GFX10-NEXT: v_and_b32_e32 v2, v9, v2
; GFX10-NEXT: v_lshl_or_b32 v6, v7, 16, v6 ; GFX10-NEXT: v_lshl_or_b32 v6, v7, 16, v6
; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4 ; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4
; GFX10-NEXT: v_lshl_or_b32 v3, v3, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2
; GFX10-NEXT: image_sample_c_d_o v0, [v0, v1, v3, v4, v6, v8], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D_ARRAY a16 ; GFX10-NEXT: image_sample_c_d_o v0, [v0, v1, v2, v4, v6, v8], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D_ARRAY a16
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call float @llvm.amdgcn.image.sample.c.d.o.2darray.f32.f16.f16(i32 4, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call float @llvm.amdgcn.image.sample.c.d.o.2darray.f32.f16.f16(i32 4, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret float %v ret float %v
} }
define amdgpu_ps <2 x float> @sample_c_d_o_2darray_V2(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %slice) { define amdgpu_ps <2 x float> @sample_c_d_o_2darray_V2(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %slice) {
Show All 16 Lines
; GFX10-LABEL: sample_c_d_o_2darray_V2: ; GFX10-LABEL: sample_c_d_o_2darray_V2:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v9, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v9, 0xffff
; GFX10-NEXT: v_and_b32_e32 v6, v9, v6 ; GFX10-NEXT: v_and_b32_e32 v6, v9, v6
; GFX10-NEXT: v_and_b32_e32 v4, v9, v4 ; GFX10-NEXT: v_and_b32_e32 v4, v9, v4
; GFX10-NEXT: v_and_b32_e32 v2, v9, v2 ; GFX10-NEXT: v_and_b32_e32 v2, v9, v2
; GFX10-NEXT: v_lshl_or_b32 v6, v7, 16, v6 ; GFX10-NEXT: v_lshl_or_b32 v6, v7, 16, v6
; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4 ; GFX10-NEXT: v_lshl_or_b32 v4, v5, 16, v4
; GFX10-NEXT: v_lshl_or_b32 v3, v3, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2
; GFX10-NEXT: image_sample_c_d_o v[0:1], [v0, v1, v3, v4, v6, v8], s[0:7], s[8:11] dmask:0x6 dim:SQ_RSRC_IMG_2D_ARRAY a16 ; GFX10-NEXT: image_sample_c_d_o v[0:1], [v0, v1, v2, v4, v6, v8], s[0:7], s[8:11] dmask:0x6 dim:SQ_RSRC_IMG_2D_ARRAY a16
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <2 x float> @llvm.amdgcn.image.sample.c.d.o.2darray.v2f32.f32.f16(i32 6, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <2 x float> @llvm.amdgcn.image.sample.c.d.o.2darray.v2f32.f32.f16(i32 6, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <2 x float> %v ret <2 x float> %v
} }
declare <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f16(i32, half, <8 x i32>, <4 x i32>, i1, i32, i32) #1 declare <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f16(i32, half, <8 x i32>, <4 x i32>, i1, i32, i32) #1
▲ Show 20 Lines • Show All 58 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.d16.dim.ll

Show First 20 Lines • Show All 90 Lines • ▼ Show 20 Lines
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v0, v2 ; GFX9-NEXT: v_mov_b32_e32 v0, v2
; GFX9-NEXT: global_store_dword v4, v3, s[12:13] ; GFX9-NEXT: global_store_dword v4, v3, s[12:13]
; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ; return to shader part epilog ; GFX9-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: image_sample_2d_f16_tfe: ; GFX10-LABEL: image_sample_2d_f16_tfe:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: s_mov_b32 s28, exec_lo ; GFX10-NEXT: s_mov_b32 s14, exec_lo
; GFX10-NEXT: s_wqm_b32 exec_lo, exec_lo ; GFX10-NEXT: s_wqm_b32 exec_lo, exec_lo
; GFX10-NEXT: v_mov_b32_e32 v4, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: v_mov_b32_e32 v5, v4 ; GFX10-NEXT: v_mov_b32_e32 v5, v4
; GFX10-NEXT: v_mov_b32_e32 v2, v4 ; GFX10-NEXT: v_mov_b32_e32 v2, v4
; GFX10-NEXT: v_mov_b32_e32 v3, v5 ; GFX10-NEXT: v_mov_b32_e32 v3, v5
; GFX10-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10-NEXT: image_sample v[2:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D tfe d16 ; GFX10-NEXT: image_sample v[2:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D tfe d16
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, v2 ; GFX10-NEXT: v_mov_b32_e32 v0, v2
; GFX10-NEXT: global_store_dword v4, v3, s[12:13] ; GFX10-NEXT: global_store_dword v4, v3, s[12:13]
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%tex = call {half,i32} @llvm.amdgcn.image.sample.2d.f16i32.f32(i32 1, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 1, i32 0) %tex = call {half,i32} @llvm.amdgcn.image.sample.2d.f16i32.f32(i32 1, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 1, i32 0)
▲ Show 20 Lines • Show All 353 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.dim.ll

Show First 20 Lines • Show All 73 Lines • ▼ Show 20 Lines
; GFX6789-NEXT: image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf tfe ; GFX6789-NEXT: image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf tfe
; GFX6789-NEXT: s_waitcnt vmcnt(0) ; GFX6789-NEXT: s_waitcnt vmcnt(0)
; GFX6789-NEXT: global_store_dword v6, v4, s[12:13] ; GFX6789-NEXT: global_store_dword v6, v4, s[12:13]
; GFX6789-NEXT: s_waitcnt vmcnt(0) ; GFX6789-NEXT: s_waitcnt vmcnt(0)
; GFX6789-NEXT: ; return to shader part epilog ; GFX6789-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: sample_1d_tfe: ; GFX10-LABEL: sample_1d_tfe:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: s_mov_b32 s28, exec_lo ; encoding: [0x7e,0x03,0x9c,0xbe] ; GFX10-NEXT: s_mov_b32 s14, exec_lo ; encoding: [0x7e,0x03,0x8e,0xbe]
; GFX10-NEXT: s_wqm_b32 exec_lo, exec_lo ; encoding: [0x7e,0x09,0xfe,0xbe] ; GFX10-NEXT: s_wqm_b32 exec_lo, exec_lo ; encoding: [0x7e,0x09,0xfe,0xbe]
; GFX10-NEXT: v_mov_b32_e32 v6, 0 ; encoding: [0x80,0x02,0x0c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v6, 0 ; encoding: [0x80,0x02,0x0c,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v7, v6 ; encoding: [0x06,0x03,0x0e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v7, v6 ; encoding: [0x06,0x03,0x0e,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v8, v6 ; encoding: [0x06,0x03,0x10,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v8, v6 ; encoding: [0x06,0x03,0x10,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v9, v6 ; encoding: [0x06,0x03,0x12,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v9, v6 ; encoding: [0x06,0x03,0x12,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v10, v6 ; encoding: [0x06,0x03,0x14,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v10, v6 ; encoding: [0x06,0x03,0x14,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v0, v6 ; encoding: [0x06,0x03,0x00,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v0, v6 ; encoding: [0x06,0x03,0x00,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v1, v7 ; encoding: [0x07,0x03,0x02,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v1, v7 ; encoding: [0x07,0x03,0x02,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v2, v8 ; encoding: [0x08,0x03,0x04,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v2, v8 ; encoding: [0x08,0x03,0x04,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v3, v9 ; encoding: [0x09,0x03,0x06,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v3, v9 ; encoding: [0x09,0x03,0x06,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v4, v10 ; encoding: [0x0a,0x03,0x08,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v4, v10 ; encoding: [0x0a,0x03,0x08,0x7e]
; GFX10-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; encoding: [0x7e,0x1c,0x7e,0x87] ; GFX10-NEXT: s_and_b32 exec_lo, exec_lo, s14 ; encoding: [0x7e,0x0e,0x7e,0x87]
; GFX10-NEXT: image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D tfe ; encoding: [0x00,0x0f,0x81,0xf0,0x05,0x00,0x40,0x00] ; GFX10-NEXT: image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D tfe ; encoding: [0x00,0x0f,0x81,0xf0,0x05,0x00,0x40,0x00]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: global_store_dword v6, v4, s[12:13] ; encoding: [0x00,0x80,0x70,0xdc,0x06,0x04,0x0c,0x00] ; GFX10-NEXT: global_store_dword v6, v4, s[12:13] ; encoding: [0x00,0x80,0x70,0xdc,0x06,0x04,0x0c,0x00]
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb] ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call {<4 x float>,i32} @llvm.amdgcn.image.sample.1d.v4f32i32.f32(i32 15, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 1, i32 0) %v = call {<4 x float>,i32} @llvm.amdgcn.image.sample.1d.v4f32i32.f32(i32 15, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 1, i32 0)
%v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.vec = extractvalue {<4 x float>, i32} %v, 0
▲ Show 20 Lines • Show All 390 Lines • ▼ Show 20 Lines
; GFX6789-NEXT: image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf lwe ; GFX6789-NEXT: image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf lwe
; GFX6789-NEXT: s_waitcnt vmcnt(0) ; GFX6789-NEXT: s_waitcnt vmcnt(0)
; GFX6789-NEXT: global_store_dword v6, v4, s[12:13] ; GFX6789-NEXT: global_store_dword v6, v4, s[12:13]
; GFX6789-NEXT: s_waitcnt vmcnt(0) ; GFX6789-NEXT: s_waitcnt vmcnt(0)
; GFX6789-NEXT: ; return to shader part epilog ; GFX6789-NEXT: ; return to shader part epilog
; ;
; GFX10-LABEL: sample_1d_lwe: ; GFX10-LABEL: sample_1d_lwe:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: s_mov_b32 s28, exec_lo ; encoding: [0x7e,0x03,0x9c,0xbe] ; GFX10-NEXT: s_mov_b32 s14, exec_lo ; encoding: [0x7e,0x03,0x8e,0xbe]
; GFX10-NEXT: s_wqm_b32 exec_lo, exec_lo ; encoding: [0x7e,0x09,0xfe,0xbe] ; GFX10-NEXT: s_wqm_b32 exec_lo, exec_lo ; encoding: [0x7e,0x09,0xfe,0xbe]
; GFX10-NEXT: v_mov_b32_e32 v6, 0 ; encoding: [0x80,0x02,0x0c,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v6, 0 ; encoding: [0x80,0x02,0x0c,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v7, v6 ; encoding: [0x06,0x03,0x0e,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v7, v6 ; encoding: [0x06,0x03,0x0e,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v8, v6 ; encoding: [0x06,0x03,0x10,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v8, v6 ; encoding: [0x06,0x03,0x10,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v9, v6 ; encoding: [0x06,0x03,0x12,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v9, v6 ; encoding: [0x06,0x03,0x12,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v10, v6 ; encoding: [0x06,0x03,0x14,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v10, v6 ; encoding: [0x06,0x03,0x14,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v0, v6 ; encoding: [0x06,0x03,0x00,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v0, v6 ; encoding: [0x06,0x03,0x00,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v1, v7 ; encoding: [0x07,0x03,0x02,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v1, v7 ; encoding: [0x07,0x03,0x02,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v2, v8 ; encoding: [0x08,0x03,0x04,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v2, v8 ; encoding: [0x08,0x03,0x04,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v3, v9 ; encoding: [0x09,0x03,0x06,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v3, v9 ; encoding: [0x09,0x03,0x06,0x7e]
; GFX10-NEXT: v_mov_b32_e32 v4, v10 ; encoding: [0x0a,0x03,0x08,0x7e] ; GFX10-NEXT: v_mov_b32_e32 v4, v10 ; encoding: [0x0a,0x03,0x08,0x7e]
; GFX10-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; encoding: [0x7e,0x1c,0x7e,0x87] ; GFX10-NEXT: s_and_b32 exec_lo, exec_lo, s14 ; encoding: [0x7e,0x0e,0x7e,0x87]
; GFX10-NEXT: image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D lwe ; encoding: [0x00,0x0f,0x82,0xf0,0x05,0x00,0x40,0x00] ; GFX10-NEXT: image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D lwe ; encoding: [0x00,0x0f,0x82,0xf0,0x05,0x00,0x40,0x00]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: global_store_dword v6, v4, s[12:13] ; encoding: [0x00,0x80,0x70,0xdc,0x06,0x04,0x0c,0x00] ; GFX10-NEXT: global_store_dword v6, v4, s[12:13] ; encoding: [0x00,0x80,0x70,0xdc,0x06,0x04,0x0c,0x00]
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb] ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call {<4 x float>,i32} @llvm.amdgcn.image.sample.1d.v4f32i32.f32(i32 15, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 2, i32 0) %v = call {<4 x float>,i32} @llvm.amdgcn.image.sample.1d.v4f32i32.f32(i32 15, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 2, i32 0)
%v.vec = extractvalue {<4 x float>, i32} %v, 0 %v.vec = extractvalue {<4 x float>, i32} %v, 0
▲ Show 20 Lines • Show All 1,810 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.g16.encode.ll

Show All 9 Lines
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_d_2d: ; GFX10-LABEL: sample_d_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; encoding: [0xff,0x02,0x0e,0x7e,0xff,0xff,0x00,0x00] ; GFX10-NEXT: v_mov_b32_e32 v6, 0xffff ; encoding: [0xff,0x02,0x0c,0x7e,0xff,0xff,0x00,0x00]
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; encoding: [0x07,0x05,0x04,0x36] ; GFX10-NEXT: v_and_b32_e32 v2, v6, v2 ; encoding: [0x06,0x05,0x04,0x36]
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; encoding: [0x07,0x01,0x00,0x36] ; GFX10-NEXT: v_and_b32_e32 v0, v6, v0 ; encoding: [0x06,0x01,0x00,0x36]
; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; encoding: [0x02,0x00,0x6f,0xd7,0x03,0x21,0x09,0x04] ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; encoding: [0x02,0x00,0x6f,0xd7,0x03,0x21,0x09,0x04]
; GFX10-NEXT: v_lshl_or_b32 v3, v1, 16, v0 ; encoding: [0x03,0x00,0x6f,0xd7,0x01,0x21,0x01,0x04] ; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v0 ; encoding: [0x00,0x00,0x6f,0xd7,0x01,0x21,0x01,0x04]
; GFX10-NEXT: image_sample_d_g16 v[0:3], [v3, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0x88,0xf0,0x03,0x00,0x40,0x00,0x02,0x04,0x05,0x00] ; GFX10-NEXT: image_sample_d_g16 v[0:3], [v0, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0x88,0xf0,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x00]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_d_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r) { define amdgpu_ps <4 x float> @sample_d_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r) {
Show All 21 Lines
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.d.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_c_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_c_d_2d: ; GFX10-LABEL: sample_c_d_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v10, 0xffff ; encoding: [0xff,0x02,0x14,0x7e,0xff,0xff,0x00,0x00] ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; encoding: [0xff,0x02,0x0e,0x7e,0xff,0xff,0x00,0x00]
; GFX10-NEXT: v_and_b32_e32 v3, v10, v3 ; encoding: [0x0a,0x07,0x06,0x36] ; GFX10-NEXT: v_and_b32_e32 v3, v7, v3 ; encoding: [0x07,0x07,0x06,0x36]
; GFX10-NEXT: v_and_b32_e32 v1, v10, v1 ; encoding: [0x0a,0x03,0x02,0x36] ; GFX10-NEXT: v_and_b32_e32 v1, v7, v1 ; encoding: [0x07,0x03,0x02,0x36]
; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3 ; encoding: [0x03,0x00,0x6f,0xd7,0x04,0x21,0x0d,0x04] ; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3 ; encoding: [0x03,0x00,0x6f,0xd7,0x04,0x21,0x0d,0x04]
; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1 ; encoding: [0x01,0x00,0x6f,0xd7,0x02,0x21,0x05,0x04] ; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1 ; encoding: [0x01,0x00,0x6f,0xd7,0x02,0x21,0x05,0x04]
; GFX10-NEXT: image_sample_c_d_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa8,0xf0,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06] ; GFX10-NEXT: image_sample_c_d_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa8,0xf0,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
Show All 12 Lines
define amdgpu_ps <4 x float> @sample_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) { define amdgpu_ps <4 x float> @sample_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
; GFX10-LABEL: sample_d_cl_2d: ; GFX10-LABEL: sample_d_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; encoding: [0xff,0x02,0x0e,0x7e,0xff,0xff,0x00,0x00] ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; encoding: [0xff,0x02,0x0e,0x7e,0xff,0xff,0x00,0x00]
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; encoding: [0x07,0x05,0x04,0x36] ; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; encoding: [0x07,0x05,0x04,0x36]
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; encoding: [0x07,0x01,0x00,0x36] ; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; encoding: [0x07,0x01,0x00,0x36]
; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; encoding: [0x02,0x00,0x6f,0xd7,0x03,0x21,0x09,0x04] ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; encoding: [0x02,0x00,0x6f,0xd7,0x03,0x21,0x09,0x04]
; GFX10-NEXT: v_lshl_or_b32 v3, v1, 16, v0 ; encoding: [0x03,0x00,0x6f,0xd7,0x01,0x21,0x01,0x04] ; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v0 ; encoding: [0x00,0x00,0x6f,0xd7,0x01,0x21,0x01,0x04]
; GFX10-NEXT: image_sample_d_cl_g16 v[0:3], [v3, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0x8c,0xf0,0x03,0x00,0x40,0x00,0x02,0x04,0x05,0x06] ; GFX10-NEXT: image_sample_d_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0x8c,0xf0,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x06]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) { define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
Show All 32 Lines
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cd.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cd.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_cd_2d: ; GFX10-LABEL: sample_cd_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; encoding: [0xff,0x02,0x0e,0x7e,0xff,0xff,0x00,0x00] ; GFX10-NEXT: v_mov_b32_e32 v6, 0xffff ; encoding: [0xff,0x02,0x0c,0x7e,0xff,0xff,0x00,0x00]
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; encoding: [0x07,0x05,0x04,0x36] ; GFX10-NEXT: v_and_b32_e32 v2, v6, v2 ; encoding: [0x06,0x05,0x04,0x36]
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; encoding: [0x07,0x01,0x00,0x36] ; GFX10-NEXT: v_and_b32_e32 v0, v6, v0 ; encoding: [0x06,0x01,0x00,0x36]
; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; encoding: [0x02,0x00,0x6f,0xd7,0x03,0x21,0x09,0x04] ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; encoding: [0x02,0x00,0x6f,0xd7,0x03,0x21,0x09,0x04]
; GFX10-NEXT: v_lshl_or_b32 v3, v1, 16, v0 ; encoding: [0x03,0x00,0x6f,0xd7,0x01,0x21,0x01,0x04] ; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v0 ; encoding: [0x00,0x00,0x6f,0xd7,0x01,0x21,0x01,0x04]
; GFX10-NEXT: image_sample_cd_g16 v[0:3], [v3, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa0,0xf1,0x03,0x00,0x40,0x00,0x02,0x04,0x05,0x00] ; GFX10-NEXT: image_sample_cd_g16 v[0:3], [v0, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x00]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_cd_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s) { define amdgpu_ps <4 x float> @sample_c_cd_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s) {
; GFX10-LABEL: sample_c_cd_1d: ; GFX10-LABEL: sample_c_cd_1d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: image_sample_c_cd_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x00] ; GFX10-NEXT: image_sample_c_cd_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x00]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_c_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_c_cd_2d: ; GFX10-LABEL: sample_c_cd_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v10, 0xffff ; encoding: [0xff,0x02,0x14,0x7e,0xff,0xff,0x00,0x00] ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; encoding: [0xff,0x02,0x0e,0x7e,0xff,0xff,0x00,0x00]
; GFX10-NEXT: v_and_b32_e32 v3, v10, v3 ; encoding: [0x0a,0x07,0x06,0x36] ; GFX10-NEXT: v_and_b32_e32 v3, v7, v3 ; encoding: [0x07,0x07,0x06,0x36]
; GFX10-NEXT: v_and_b32_e32 v1, v10, v1 ; encoding: [0x0a,0x03,0x02,0x36] ; GFX10-NEXT: v_and_b32_e32 v1, v7, v1 ; encoding: [0x07,0x03,0x02,0x36]
; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3 ; encoding: [0x03,0x00,0x6f,0xd7,0x04,0x21,0x0d,0x04] ; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3 ; encoding: [0x03,0x00,0x6f,0xd7,0x04,0x21,0x0d,0x04]
; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1 ; encoding: [0x01,0x00,0x6f,0xd7,0x02,0x21,0x05,0x04] ; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1 ; encoding: [0x01,0x00,0x6f,0xd7,0x02,0x21,0x05,0x04]
; GFX10-NEXT: image_sample_c_cd_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06] ; GFX10-NEXT: image_sample_c_cd_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
Show All 12 Lines
define amdgpu_ps <4 x float> @sample_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) { define amdgpu_ps <4 x float> @sample_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
; GFX10-LABEL: sample_cd_cl_2d: ; GFX10-LABEL: sample_cd_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; encoding: [0xff,0x02,0x0e,0x7e,0xff,0xff,0x00,0x00] ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; encoding: [0xff,0x02,0x0e,0x7e,0xff,0xff,0x00,0x00]
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; encoding: [0x07,0x05,0x04,0x36] ; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; encoding: [0x07,0x05,0x04,0x36]
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; encoding: [0x07,0x01,0x00,0x36] ; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; encoding: [0x07,0x01,0x00,0x36]
; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; encoding: [0x02,0x00,0x6f,0xd7,0x03,0x21,0x09,0x04] ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; encoding: [0x02,0x00,0x6f,0xd7,0x03,0x21,0x09,0x04]
; GFX10-NEXT: v_lshl_or_b32 v3, v1, 16, v0 ; encoding: [0x03,0x00,0x6f,0xd7,0x01,0x21,0x01,0x04] ; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v0 ; encoding: [0x00,0x00,0x6f,0xd7,0x01,0x21,0x01,0x04]
; GFX10-NEXT: image_sample_cd_cl_g16 v[0:3], [v3, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa4,0xf1,0x03,0x00,0x40,0x00,0x02,0x04,0x05,0x06] ; GFX10-NEXT: image_sample_cd_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa4,0xf1,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x06]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf] ; GFX10-NEXT: s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) { define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
▲ Show 20 Lines • Show All 83 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.g16.ll

Show All 9 Lines
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_d_2d: ; GFX10-LABEL: sample_d_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v6, 0xffff
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; GFX10-NEXT: v_and_b32_e32 v2, v6, v2
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; GFX10-NEXT: v_and_b32_e32 v0, v6, v0
; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2
; GFX10-NEXT: v_lshl_or_b32 v3, v1, 16, v0 ; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v0
; GFX10-NEXT: image_sample_d_g16 v[0:3], [v3, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_d_g16 v[0:3], [v0, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_d_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r) { define amdgpu_ps <4 x float> @sample_d_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r) {
Show All 21 Lines
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.d.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_c_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_c_d_2d: ; GFX10-LABEL: sample_c_d_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v10, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_and_b32_e32 v3, v10, v3 ; GFX10-NEXT: v_and_b32_e32 v3, v7, v3
; GFX10-NEXT: v_and_b32_e32 v1, v10, v1 ; GFX10-NEXT: v_and_b32_e32 v1, v7, v1
; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3 ; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3
; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1 ; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1
; GFX10-NEXT: image_sample_c_d_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_c_d_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
Show All 12 Lines
define amdgpu_ps <4 x float> @sample_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) { define amdgpu_ps <4 x float> @sample_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
; GFX10-LABEL: sample_d_cl_2d: ; GFX10-LABEL: sample_d_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; GFX10-NEXT: v_and_b32_e32 v2, v7, v2
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; GFX10-NEXT: v_and_b32_e32 v0, v7, v0
; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2
; GFX10-NEXT: v_lshl_or_b32 v3, v1, 16, v0 ; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v0
; GFX10-NEXT: image_sample_d_cl_g16 v[0:3], [v3, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_d_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) { define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
Show All 32 Lines
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cd.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cd.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_cd_2d: ; GFX10-LABEL: sample_cd_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v6, 0xffff
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; GFX10-NEXT: v_and_b32_e32 v2, v6, v2
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; GFX10-NEXT: v_and_b32_e32 v0, v6, v0
; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2
; GFX10-NEXT: v_lshl_or_b32 v3, v1, 16, v0 ; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v0
; GFX10-NEXT: image_sample_cd_g16 v[0:3], [v3, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_cd_g16 v[0:3], [v0, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_cd_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s) { define amdgpu_ps <4 x float> @sample_c_cd_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s) {
; GFX10-LABEL: sample_c_cd_1d: ; GFX10-LABEL: sample_c_cd_1d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: image_sample_c_cd_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-NEXT: image_sample_c_cd_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) { define amdgpu_ps <4 x float> @sample_c_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
; GFX10-LABEL: sample_c_cd_2d: ; GFX10-LABEL: sample_c_cd_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v10, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_and_b32_e32 v3, v10, v3 ; GFX10-NEXT: v_and_b32_e32 v3, v7, v3
; GFX10-NEXT: v_and_b32_e32 v1, v10, v1 ; GFX10-NEXT: v_and_b32_e32 v1, v7, v1
; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3 ; GFX10-NEXT: v_lshl_or_b32 v3, v4, 16, v3
; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1 ; GFX10-NEXT: v_lshl_or_b32 v1, v2, 16, v1
; GFX10-NEXT: image_sample_c_cd_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_c_cd_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
Show All 12 Lines
define amdgpu_ps <4 x float> @sample_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) { define amdgpu_ps <4 x float> @sample_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
; GFX10-LABEL: sample_cd_cl_2d: ; GFX10-LABEL: sample_cd_cl_2d:
; GFX10: ; %bb.0: ; %main_body ; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v7, 0xffff
; GFX10-NEXT: v_and_b32_e32 v2, v7, v2 ; GFX10-NEXT: v_and_b32_e32 v2, v7, v2
; GFX10-NEXT: v_and_b32_e32 v0, v7, v0 ; GFX10-NEXT: v_and_b32_e32 v0, v7, v0
; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v2, v3, 16, v2
; GFX10-NEXT: v_lshl_or_b32 v3, v1, 16, v0 ; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v0
; GFX10-NEXT: image_sample_cd_cl_g16 v[0:3], [v3, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; GFX10-NEXT: image_sample_cd_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog ; GFX10-NEXT: ; return to shader part epilog
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v ret <4 x float> %v
} }
define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) { define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
▲ Show 20 Lines • Show All 83 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wqm.demote.ll

Show First 20 Lines • Show All 350 Lines • ▼ Show 20 Lines
; GFX10-32-NEXT: v_cmp_ngt_f32_e32 vcc_lo, 0, v1 ; GFX10-32-NEXT: v_cmp_ngt_f32_e32 vcc_lo, 0, v1
; GFX10-32-NEXT: s_and_saveexec_b32 s13, vcc_lo ; GFX10-32-NEXT: s_and_saveexec_b32 s13, vcc_lo
; GFX10-32-NEXT: s_xor_b32 s13, exec_lo, s13 ; GFX10-32-NEXT: s_xor_b32 s13, exec_lo, s13
; GFX10-32-NEXT: s_cbranch_execz BB3_3 ; GFX10-32-NEXT: s_cbranch_execz BB3_3
; GFX10-32-NEXT: ; %bb.1: ; %.demote ; GFX10-32-NEXT: ; %bb.1: ; %.demote
; GFX10-32-NEXT: s_andn2_b32 s12, s12, exec_lo ; GFX10-32-NEXT: s_andn2_b32 s12, s12, exec_lo
; GFX10-32-NEXT: s_cbranch_scc0 BB3_4 ; GFX10-32-NEXT: s_cbranch_scc0 BB3_4
; GFX10-32-NEXT: ; %bb.2: ; %.demote ; GFX10-32-NEXT: ; %bb.2: ; %.demote
; GFX10-32-NEXT: s_wqm_b32 s28, s12 ; GFX10-32-NEXT: s_wqm_b32 s14, s12
; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10-32-NEXT: BB3_3: ; %.continue ; GFX10-32-NEXT: BB3_3: ; %.continue
; GFX10-32-NEXT: s_or_b32 exec_lo, exec_lo, s13 ; GFX10-32-NEXT: s_or_b32 exec_lo, exec_lo, s13
; GFX10-32-NEXT: image_sample v0, v0, s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_1D ; GFX10-32-NEXT: image_sample v0, v0, s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_1D
; GFX10-32-NEXT: s_waitcnt vmcnt(0) ; GFX10-32-NEXT: s_waitcnt vmcnt(0)
; GFX10-32-NEXT: v_add_f32_e32 v0, v0, v0 ; GFX10-32-NEXT: v_add_f32_e32 v0, v0, v0
; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s12 ; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s12
; GFX10-32-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-32-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-32-NEXT: s_waitcnt vmcnt(0) ; GFX10-32-NEXT: s_waitcnt vmcnt(0)
; GFX10-32-NEXT: s_branch BB3_5 ; GFX10-32-NEXT: s_branch BB3_5
; GFX10-32-NEXT: BB3_4: ; GFX10-32-NEXT: BB3_4:
; GFX10-32-NEXT: s_mov_b32 exec_lo, 0 ; GFX10-32-NEXT: s_mov_b32 exec_lo, 0
; GFX10-32-NEXT: exp null off, off, off, off done vm ; GFX10-32-NEXT: exp null off, off, off, off done vm
; GFX10-32-NEXT: s_endpgm ; GFX10-32-NEXT: s_endpgm
; GFX10-32-NEXT: BB3_5: ; GFX10-32-NEXT: BB3_5:
; ;
; GFX10-64-LABEL: wqm_demote_1: ; GFX10-64-LABEL: wqm_demote_1:
; GFX10-64: ; %bb.0: ; %.entry ; GFX10-64: ; %bb.0: ; %.entry
; GFX10-64-NEXT: s_mov_b64 s[12:13], exec ; GFX10-64-NEXT: s_mov_b64 s[12:13], exec
; GFX10-64-NEXT: s_wqm_b64 exec, exec ; GFX10-64-NEXT: s_wqm_b64 exec, exec
; GFX10-64-NEXT: v_cmp_ngt_f32_e32 vcc, 0, v1 ; GFX10-64-NEXT: v_cmp_ngt_f32_e32 vcc, 0, v1
; GFX10-64-NEXT: s_and_saveexec_b64 s[14:15], vcc ; GFX10-64-NEXT: s_and_saveexec_b64 s[14:15], vcc
; GFX10-64-NEXT: s_xor_b64 s[28:29], exec, s[14:15] ; GFX10-64-NEXT: s_xor_b64 s[14:15], exec, s[14:15]
; GFX10-64-NEXT: s_cbranch_execz BB3_3 ; GFX10-64-NEXT: s_cbranch_execz BB3_3
; GFX10-64-NEXT: ; %bb.1: ; %.demote ; GFX10-64-NEXT: ; %bb.1: ; %.demote
; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], exec ; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], exec
; GFX10-64-NEXT: s_cbranch_scc0 BB3_4 ; GFX10-64-NEXT: s_cbranch_scc0 BB3_4
; GFX10-64-NEXT: ; %bb.2: ; %.demote ; GFX10-64-NEXT: ; %bb.2: ; %.demote
; GFX10-64-NEXT: s_wqm_b64 s[16:17], s[12:13] ; GFX10-64-NEXT: s_wqm_b64 s[16:17], s[12:13]
; GFX10-64-NEXT: s_and_b64 exec, exec, s[16:17] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[16:17]
; GFX10-64-NEXT: BB3_3: ; %.continue ; GFX10-64-NEXT: BB3_3: ; %.continue
; GFX10-64-NEXT: s_or_b64 exec, exec, s[28:29] ; GFX10-64-NEXT: s_or_b64 exec, exec, s[14:15]
; GFX10-64-NEXT: image_sample v0, v0, s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v0, v0, s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0 ; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0
; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13]
; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: s_branch BB3_5 ; GFX10-64-NEXT: s_branch BB3_5
; GFX10-64-NEXT: BB3_4: ; GFX10-64-NEXT: BB3_4:
▲ Show 20 Lines • Show All 87 Lines • ▼ Show 20 Lines
; GFX10-32-NEXT: v_cmp_ngt_f32_e32 vcc_lo, 0, v0 ; GFX10-32-NEXT: v_cmp_ngt_f32_e32 vcc_lo, 0, v0
; GFX10-32-NEXT: s_and_saveexec_b32 s13, vcc_lo ; GFX10-32-NEXT: s_and_saveexec_b32 s13, vcc_lo
; GFX10-32-NEXT: s_xor_b32 s13, exec_lo, s13 ; GFX10-32-NEXT: s_xor_b32 s13, exec_lo, s13
; GFX10-32-NEXT: s_cbranch_execz BB4_3 ; GFX10-32-NEXT: s_cbranch_execz BB4_3
; GFX10-32-NEXT: ; %bb.1: ; %.demote ; GFX10-32-NEXT: ; %bb.1: ; %.demote
; GFX10-32-NEXT: s_andn2_b32 s12, s12, exec_lo ; GFX10-32-NEXT: s_andn2_b32 s12, s12, exec_lo
; GFX10-32-NEXT: s_cbranch_scc0 BB4_4 ; GFX10-32-NEXT: s_cbranch_scc0 BB4_4
; GFX10-32-NEXT: ; %bb.2: ; %.demote ; GFX10-32-NEXT: ; %bb.2: ; %.demote
; GFX10-32-NEXT: s_wqm_b32 s28, s12 ; GFX10-32-NEXT: s_wqm_b32 s14, s12
; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s28 ; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s14
; GFX10-32-NEXT: BB4_3: ; %.continue ; GFX10-32-NEXT: BB4_3: ; %.continue
; GFX10-32-NEXT: s_or_b32 exec_lo, exec_lo, s13 ; GFX10-32-NEXT: s_or_b32 exec_lo, exec_lo, s13
; GFX10-32-NEXT: v_add_f32_e32 v0, v0, v0 ; GFX10-32-NEXT: v_add_f32_e32 v0, v0, v0
; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s12 ; GFX10-32-NEXT: s_and_b32 exec_lo, exec_lo, s12
; GFX10-32-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-32-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-32-NEXT: s_waitcnt vmcnt(0) ; GFX10-32-NEXT: s_waitcnt vmcnt(0)
; GFX10-32-NEXT: s_branch BB4_5 ; GFX10-32-NEXT: s_branch BB4_5
; GFX10-32-NEXT: BB4_4: ; GFX10-32-NEXT: BB4_4:
; GFX10-32-NEXT: s_mov_b32 exec_lo, 0 ; GFX10-32-NEXT: s_mov_b32 exec_lo, 0
; GFX10-32-NEXT: exp null off, off, off, off done vm ; GFX10-32-NEXT: exp null off, off, off, off done vm
; GFX10-32-NEXT: s_endpgm ; GFX10-32-NEXT: s_endpgm
; GFX10-32-NEXT: BB4_5: ; GFX10-32-NEXT: BB4_5:
; ;
; GFX10-64-LABEL: wqm_demote_2: ; GFX10-64-LABEL: wqm_demote_2:
; GFX10-64: ; %bb.0: ; %.entry ; GFX10-64: ; %bb.0: ; %.entry
; GFX10-64-NEXT: s_mov_b64 s[12:13], exec ; GFX10-64-NEXT: s_mov_b64 s[12:13], exec
; GFX10-64-NEXT: s_wqm_b64 exec, exec ; GFX10-64-NEXT: s_wqm_b64 exec, exec
; GFX10-64-NEXT: image_sample v0, v0, s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v0, v0, s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: v_cmp_ngt_f32_e32 vcc, 0, v0 ; GFX10-64-NEXT: v_cmp_ngt_f32_e32 vcc, 0, v0
; GFX10-64-NEXT: s_and_saveexec_b64 s[14:15], vcc ; GFX10-64-NEXT: s_and_saveexec_b64 s[14:15], vcc
; GFX10-64-NEXT: s_xor_b64 s[28:29], exec, s[14:15] ; GFX10-64-NEXT: s_xor_b64 s[14:15], exec, s[14:15]
; GFX10-64-NEXT: s_cbranch_execz BB4_3 ; GFX10-64-NEXT: s_cbranch_execz BB4_3
; GFX10-64-NEXT: ; %bb.1: ; %.demote ; GFX10-64-NEXT: ; %bb.1: ; %.demote
; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], exec ; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], exec
; GFX10-64-NEXT: s_cbranch_scc0 BB4_4 ; GFX10-64-NEXT: s_cbranch_scc0 BB4_4
; GFX10-64-NEXT: ; %bb.2: ; %.demote ; GFX10-64-NEXT: ; %bb.2: ; %.demote
; GFX10-64-NEXT: s_wqm_b64 s[16:17], s[12:13] ; GFX10-64-NEXT: s_wqm_b64 s[16:17], s[12:13]
; GFX10-64-NEXT: s_and_b64 exec, exec, s[16:17] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[16:17]
; GFX10-64-NEXT: BB4_3: ; %.continue ; GFX10-64-NEXT: BB4_3: ; %.continue
; GFX10-64-NEXT: s_or_b64 exec, exec, s[28:29] ; GFX10-64-NEXT: s_or_b64 exec, exec, s[14:15]
; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0 ; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0
; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13]
; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: s_branch BB4_5 ; GFX10-64-NEXT: s_branch BB4_5
; GFX10-64-NEXT: BB4_4: ; GFX10-64-NEXT: BB4_4:
; GFX10-64-NEXT: s_mov_b64 exec, 0 ; GFX10-64-NEXT: s_mov_b64 exec, 0
; GFX10-64-NEXT: exp null off, off, off, off done vm ; GFX10-64-NEXT: exp null off, off, off, off done vm
▲ Show 20 Lines • Show All 96 Lines • ▼ Show 20 Lines
; GFX10-64-NEXT: s_wqm_b64 exec, exec ; GFX10-64-NEXT: s_wqm_b64 exec, exec
; GFX10-64-NEXT: image_sample v0, v0, s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v0, v0, s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: v_cmp_gt_f32_e32 vcc, 0, v0 ; GFX10-64-NEXT: v_cmp_gt_f32_e32 vcc, 0, v0
; GFX10-64-NEXT: s_xor_b64 s[14:15], vcc, exec ; GFX10-64-NEXT: s_xor_b64 s[14:15], vcc, exec
; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], s[14:15] ; GFX10-64-NEXT: s_andn2_b64 s[12:13], s[12:13], s[14:15]
; GFX10-64-NEXT: s_cbranch_scc0 BB5_2 ; GFX10-64-NEXT: s_cbranch_scc0 BB5_2
; GFX10-64-NEXT: ; %bb.1: ; %.entry ; GFX10-64-NEXT: ; %bb.1: ; %.entry
; GFX10-64-NEXT: s_wqm_b64 s[28:29], s[12:13] ; GFX10-64-NEXT: s_wqm_b64 s[14:15], s[12:13]
; GFX10-64-NEXT: s_and_b64 exec, exec, s[28:29] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[14:15]
; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0 ; GFX10-64-NEXT: v_add_f32_e32 v0, v0, v0
; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13] ; GFX10-64-NEXT: s_and_b64 exec, exec, s[12:13]
; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; GFX10-64-NEXT: image_sample v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX10-64-NEXT: s_waitcnt vmcnt(0) ; GFX10-64-NEXT: s_waitcnt vmcnt(0)
; GFX10-64-NEXT: s_branch BB5_3 ; GFX10-64-NEXT: s_branch BB5_3
; GFX10-64-NEXT: BB5_2: ; GFX10-64-NEXT: BB5_2:
; GFX10-64-NEXT: s_mov_b64 exec, 0 ; GFX10-64-NEXT: s_mov_b64 exec, 0
; GFX10-64-NEXT: exp null off, off, off, off done vm ; GFX10-64-NEXT: exp null off, off, off, off done vm
▲ Show 20 Lines • Show All 552 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/llvm.mulo.ll

Show First 20 Lines • Show All 151 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cmp_ne_u64_e32 vcc, v[3:4], v[5:6] ; GFX9-NEXT: v_cmp_ne_u64_e32 vcc, v[3:4], v[5:6]
; GFX9-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc ; GFX9-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: smulo_i64_s_s: ; GFX10-LABEL: smulo_i64_s_s:
; GFX10: ; %bb.0: ; %bb ; GFX10: ; %bb.0: ; %bb
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mul_lo_u32 v15, v0, v3 ; GFX10-NEXT: v_mul_lo_u32 v4, v0, v3
; GFX10-NEXT: v_mul_hi_u32 v5, v0, v2 ; GFX10-NEXT: v_mul_hi_u32 v5, v0, v2
; GFX10-NEXT: v_mul_hi_u32 v6, v0, v3 ; GFX10-NEXT: v_mul_hi_u32 v6, v0, v3
; GFX10-NEXT: v_mul_lo_u32 v8, v1, v2 ; GFX10-NEXT: v_mul_lo_u32 v8, v1, v2
; GFX10-NEXT: v_mul_hi_u32 v7, v1, v2 ; GFX10-NEXT: v_mul_hi_u32 v7, v1, v2
; GFX10-NEXT: v_mul_hi_i32 v9, v1, v3 ; GFX10-NEXT: v_mul_hi_i32 v9, v1, v3
; GFX10-NEXT: v_mul_lo_u32 v11, v1, v3 ; GFX10-NEXT: v_mul_lo_u32 v11, v1, v3
; GFX10-NEXT: v_add_co_u32 v10, vcc_lo, v5, v15 ; GFX10-NEXT: v_add_co_u32 v10, vcc_lo, v5, v4
; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, 0, v6, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, 0, v6, vcc_lo
; GFX10-NEXT: v_add_co_u32 v10, vcc_lo, v10, v8 ; GFX10-NEXT: v_add_co_u32 v10, vcc_lo, v10, v8
; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, v6, v7, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, v6, v7, vcc_lo
; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, 0, v9, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, 0, v9, vcc_lo
; GFX10-NEXT: v_add_co_u32 v11, vcc_lo, v6, v11 ; GFX10-NEXT: v_add_co_u32 v6, vcc_lo, v6, v11
; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, 0, v7, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v7, vcc_lo, 0, v7, vcc_lo
; GFX10-NEXT: v_sub_co_u32 v9, vcc_lo, v11, v2 ; GFX10-NEXT: v_sub_co_u32 v9, vcc_lo, v6, v2
; GFX10-NEXT: v_subrev_co_ci_u32_e32 v10, vcc_lo, 0, v7, vcc_lo ; GFX10-NEXT: v_subrev_co_ci_u32_e32 v10, vcc_lo, 0, v7, vcc_lo
; GFX10-NEXT: v_cmp_gt_i32_e32 vcc_lo, 0, v1 ; GFX10-NEXT: v_cmp_gt_i32_e32 vcc_lo, 0, v1
; GFX10-NEXT: v_add3_u32 v1, v5, v15, v8 ; GFX10-NEXT: v_add3_u32 v1, v5, v4, v8
; GFX10-NEXT: v_cndmask_b32_e32 v6, v11, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v6, v6, v9, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, v7, v10, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v7, v7, v10, vcc_lo
; GFX10-NEXT: v_ashrrev_i32_e32 v4, 31, v1 ; GFX10-NEXT: v_ashrrev_i32_e32 v4, 31, v1
; GFX10-NEXT: v_sub_co_u32 v8, vcc_lo, v6, v0 ; GFX10-NEXT: v_sub_co_u32 v8, vcc_lo, v6, v0
; GFX10-NEXT: v_mul_lo_u32 v0, v0, v2 ; GFX10-NEXT: v_mul_lo_u32 v0, v0, v2
; GFX10-NEXT: v_subrev_co_ci_u32_e32 v9, vcc_lo, 0, v7, vcc_lo ; GFX10-NEXT: v_subrev_co_ci_u32_e32 v9, vcc_lo, 0, v7, vcc_lo
; GFX10-NEXT: v_cmp_gt_i32_e32 vcc_lo, 0, v3 ; GFX10-NEXT: v_cmp_gt_i32_e32 vcc_lo, 0, v3
; GFX10-NEXT: v_mov_b32_e32 v5, v4 ; GFX10-NEXT: v_mov_b32_e32 v5, v4
; GFX10-NEXT: v_cndmask_b32_e32 v7, v7, v9, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v7, v7, v9, vcc_lo
▲ Show 20 Lines • Show All 269 Lines • ▼ Show 20 Lines
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: smulo_i64_v_4: ; GFX10-LABEL: smulo_i64_v_4:
; GFX10: ; %bb.0: ; %bb ; GFX10: ; %bb.0: ; %bb
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_lshlrev_b64 v[4:5], 2, v[0:1] ; GFX10-NEXT: v_lshlrev_b64 v[4:5], 2, v[0:1]
; GFX10-NEXT: v_alignbit_b32 v3, v1, v0, 30 ; GFX10-NEXT: v_alignbit_b32 v3, v1, v0, 30
; GFX10-NEXT: v_ashrrev_i64 v[6:7], 2, v[4:5] ; GFX10-NEXT: v_ashrrev_i64 v[5:6], 2, v[4:5]
; GFX10-NEXT: v_cmp_ne_u64_e32 vcc_lo, v[6:7], v[0:1] ; GFX10-NEXT: v_cmp_ne_u64_e32 vcc_lo, v[5:6], v[0:1]
; GFX10-NEXT: v_mov_b32_e32 v0, v4 ; GFX10-NEXT: v_mov_b32_e32 v0, v4
; GFX10-NEXT: v_mov_b32_e32 v1, v3 ; GFX10-NEXT: v_mov_b32_e32 v1, v3
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
bb: bb:
%umulo = tail call { i64, i1 } @llvm.smul.with.overflow.i64(i64 %i, i64 4) %umulo = tail call { i64, i1 } @llvm.smul.with.overflow.i64(i64 %i, i64 4)
ret { i64, i1 } %umulo ret { i64, i1 } %umulo
} }
▲ Show 20 Lines • Show All 48 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/lshr.v2i16.ll

Show First 20 Lines • Show All 533 Lines • ▼ Show 20 Lines
; CI-NEXT: v_or_b32_e32 v3, v3, v5 ; CI-NEXT: v_or_b32_e32 v3, v3, v5
; CI-NEXT: v_or_b32_e32 v2, v2, v4 ; CI-NEXT: v_or_b32_e32 v2, v2, v4
; CI-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64 ; CI-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64
; CI-NEXT: s_endpgm ; CI-NEXT: s_endpgm
; ;
; GFX10-LABEL: v_lshr_v4i16: ; GFX10-LABEL: v_lshr_v4i16:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24 ; GFX10-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 3, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 3, v0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx2 v[0:1], v7, s[2:3] ; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[2:3]
; GFX10-NEXT: global_load_dwordx2 v[2:3], v7, s[2:3] offset:8 ; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[2:3] offset:8
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_pk_lshrrev_b16 v1, v3, v1 ; GFX10-NEXT: v_pk_lshrrev_b16 v1, v3, v1
; GFX10-NEXT: v_pk_lshrrev_b16 v0, v2, v0 ; GFX10-NEXT: v_pk_lshrrev_b16 v0, v2, v0
; GFX10-NEXT: global_store_dwordx2 v7, v[0:1], s[0:1] ; GFX10-NEXT: global_store_dwordx2 v4, v[0:1], s[0:1]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x() %tid = call i32 @llvm.amdgcn.workitem.id.x()
%tid.ext = sext i32 %tid to i64 %tid.ext = sext i32 %tid to i64
%in.gep = getelementptr inbounds <4 x i16>, <4 x i16> addrspace(1)* %in, i64 %tid.ext %in.gep = getelementptr inbounds <4 x i16>, <4 x i16> addrspace(1)* %in, i64 %tid.ext
%out.gep = getelementptr inbounds <4 x i16>, <4 x i16> addrspace(1)* %out, i64 %tid.ext %out.gep = getelementptr inbounds <4 x i16>, <4 x i16> addrspace(1)* %out, i64 %tid.ext
%b_ptr = getelementptr <4 x i16>, <4 x i16> addrspace(1)* %in.gep, i32 1 %b_ptr = getelementptr <4 x i16>, <4 x i16> addrspace(1)* %in.gep, i32 1
%a = load <4 x i16>, <4 x i16> addrspace(1)* %in.gep %a = load <4 x i16>, <4 x i16> addrspace(1)* %in.gep
%b = load <4 x i16>, <4 x i16> addrspace(1)* %b_ptr %b = load <4 x i16>, <4 x i16> addrspace(1)* %b_ptr
▲ Show 20 Lines • Show All 85 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/memory_clause.ll

Show First 20 Lines • Show All 324 Lines • ▼ Show 20 Lines
; GCN-NEXT: global_store_dword v[3:4], v2, off offset:128 ; GCN-NEXT: global_store_dword v[3:4], v2, off offset:128
; GCN-NEXT: s_waitcnt vmcnt(0) ; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GCN-SCRATCH-LABEL: load_global_d16_hi: ; GCN-SCRATCH-LABEL: load_global_d16_hi:
; GCN-SCRATCH: ; %bb.0: ; %entry ; GCN-SCRATCH: ; %bb.0: ; %entry
; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SCRATCH-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-SCRATCH-NEXT: s_waitcnt_vscnt null, 0x0
; GCN-SCRATCH-NEXT: v_mov_b32_e32 v6, v2 ; GCN-SCRATCH-NEXT: v_mov_b32_e32 v5, v2
; GCN-SCRATCH-NEXT: s_clause 0x1 ; GCN-SCRATCH-NEXT: s_clause 0x1
; GCN-SCRATCH-NEXT: global_load_short_d16_hi v6, v[0:1], off ; GCN-SCRATCH-NEXT: global_load_short_d16_hi v5, v[0:1], off
; GCN-SCRATCH-NEXT: global_load_short_d16_hi v2, v[0:1], off offset:64 ; GCN-SCRATCH-NEXT: global_load_short_d16_hi v2, v[0:1], off offset:64
; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(1) ; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(1)
; GCN-SCRATCH-NEXT: global_store_dword v[3:4], v6, off ; GCN-SCRATCH-NEXT: global_store_dword v[3:4], v5, off
; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(0) ; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(0)
; GCN-SCRATCH-NEXT: global_store_dword v[3:4], v2, off offset:128 ; GCN-SCRATCH-NEXT: global_store_dword v[3:4], v2, off offset:128
; GCN-SCRATCH-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-SCRATCH-NEXT: s_waitcnt_vscnt null, 0x0
; GCN-SCRATCH-NEXT: s_setpc_b64 s[30:31] ; GCN-SCRATCH-NEXT: s_setpc_b64 s[30:31]
entry: entry:
%gep = getelementptr inbounds i16, i16 addrspace(1)* %in, i64 32 %gep = getelementptr inbounds i16, i16 addrspace(1)* %in, i64 32
%load1 = load i16, i16 addrspace(1)* %in %load1 = load i16, i16 addrspace(1)* %in
%load2 = load i16, i16 addrspace(1)* %gep %load2 = load i16, i16 addrspace(1)* %gep
Show All 21 Lines
; GCN-NEXT: global_store_dword v[3:4], v2, off offset:128 ; GCN-NEXT: global_store_dword v[3:4], v2, off offset:128
; GCN-NEXT: s_waitcnt vmcnt(0) ; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GCN-SCRATCH-LABEL: load_global_d16_lo: ; GCN-SCRATCH-LABEL: load_global_d16_lo:
; GCN-SCRATCH: ; %bb.0: ; %entry ; GCN-SCRATCH: ; %bb.0: ; %entry
; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SCRATCH-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-SCRATCH-NEXT: s_waitcnt_vscnt null, 0x0
; GCN-SCRATCH-NEXT: v_mov_b32_e32 v6, v2 ; GCN-SCRATCH-NEXT: v_mov_b32_e32 v5, v2
; GCN-SCRATCH-NEXT: s_clause 0x1 ; GCN-SCRATCH-NEXT: s_clause 0x1
; GCN-SCRATCH-NEXT: global_load_short_d16 v6, v[0:1], off ; GCN-SCRATCH-NEXT: global_load_short_d16 v5, v[0:1], off
; GCN-SCRATCH-NEXT: global_load_short_d16 v2, v[0:1], off offset:64 ; GCN-SCRATCH-NEXT: global_load_short_d16 v2, v[0:1], off offset:64
; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(1) ; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(1)
; GCN-SCRATCH-NEXT: global_store_dword v[3:4], v6, off ; GCN-SCRATCH-NEXT: global_store_dword v[3:4], v5, off
; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(0) ; GCN-SCRATCH-NEXT: s_waitcnt vmcnt(0)
; GCN-SCRATCH-NEXT: global_store_dword v[3:4], v2, off offset:128 ; GCN-SCRATCH-NEXT: global_store_dword v[3:4], v2, off offset:128
; GCN-SCRATCH-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-SCRATCH-NEXT: s_waitcnt_vscnt null, 0x0
; GCN-SCRATCH-NEXT: s_setpc_b64 s[30:31] ; GCN-SCRATCH-NEXT: s_setpc_b64 s[30:31]
entry: entry:
%gep = getelementptr inbounds i16, i16 addrspace(1)* %in, i64 32 %gep = getelementptr inbounds i16, i16 addrspace(1)* %in, i64 32
%reg.bc1 = bitcast i32 %reg to <2 x i16> %reg.bc1 = bitcast i32 %reg to <2 x i16>
%reg.bc2 = bitcast i32 %reg to <2 x i16> %reg.bc2 = bitcast i32 %reg to <2 x i16>
▲ Show 20 Lines • Show All 148 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/regbank-reassign-split.mir

  • This file was deleted.
# RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs -run-pass greedy,amdgpu-regbanks-reassign,virtregrewriter -o - %s | FileCheck -check-prefix=GCN %s
--- |
define amdgpu_kernel void @do_not_reassign_spill() #0 { ret void }
attributes #0 = { "amdgpu-num-vgpr"="8" }
...
# GCN-LABEL: do_not_reassign_spill{{$}}
# GCN: V_AND_B32_e32 killed $vgpr1, killed $vgpr5,
---
name: do_not_reassign_spill
tracksRegLiveness: true
machineFunctionInfo:
stackPtrOffsetReg: $sgpr32
stack:
- { id: 0, type: default, offset: 0, size: 4, alignment: 4 }
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr0' }
- { id: 1, class: vgpr_32, preferred-register: '$vgpr1' }
- { id: 2, class: vgpr_32, preferred-register: '$vgpr2' }
- { id: 3, class: vgpr_32, preferred-register: '$vgpr3' }
- { id: 4, class: vgpr_32, preferred-register: '$vgpr4' }
- { id: 5, class: vgpr_32, preferred-register: '$vgpr5' }
- { id: 6, class: vgpr_32 }
body: |
bb.0:
%0 = SI_SPILL_V32_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load 4 from %stack.0, align 4, addrspace 5)
%1 = SI_SPILL_V32_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load 4 from %stack.0, align 4, addrspace 5)
%2 = SI_SPILL_V32_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load 4 from %stack.0, align 4, addrspace 5)
%3 = SI_SPILL_V32_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load 4 from %stack.0, align 4, addrspace 5)
%4 = SI_SPILL_V32_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load 4 from %stack.0, align 4, addrspace 5)
%5 = SI_SPILL_V32_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load 4 from %stack.0, align 4, addrspace 5)
S_NOP 0, implicit-def dead $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
S_NOP 0, implicit %0, implicit %1, implicit %2, implicit %3, implicit %4, implicit %5
%6 = V_AND_B32_e32 %1, %5, implicit $exec
S_ENDPGM 0, implicit %6
...

llvm/test/CodeGen/AMDGPU/regbank-reassign-wave64.mir

  • This file was deleted.
# RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -verify-machineinstrs -run-pass greedy,amdgpu-regbanks-reassign,virtregrewriter -o - %s | FileCheck -check-prefix=GCN %s
# Test that subreg reassignments are correctly handled when whole register also
# conflicts. If this is mishandled stall counts will be incorrect and cause an
# infinite loop.
# GCN-LABEL: vgpr64_mixed_use{{$}}
# GCN: $vgpr0_vgpr1 = IMPLICIT_DEF
# GCN: $vgpr4_vgpr5 = IMPLICIT_DEF
# GCN: $vcc = IMPLICIT_DEF
# GCN: $vgpr2_vgpr3 = IMPLICIT_DEF
# GCN: $vgpr6_vgpr7 = IMPLICIT_DEF
# GCN: $vgpr8_vgpr9_vgpr10_vgpr11 = IMPLICIT_DEF
# GCN: $vgpr12_vgpr13_vgpr14_vgpr15 = IMPLICIT_DEF
# GCN: $vgpr16_vgpr17_vgpr18_vgpr19 = IMPLICIT_DEF
# GCN: $vgpr20_vgpr21_vgpr22_vgpr23 = IMPLICIT_DEF
# GCN: $vgpr24_vgpr25_vgpr26_vgpr27 = IMPLICIT_DEF
# GCN: $vgpr28_vgpr29_vgpr30_vgpr31 = IMPLICIT_DEF
# GCN: $vgpr32_vgpr33_vgpr34_vgpr35 = IMPLICIT_DEF
# GCN: $vgpr36_vgpr37_vgpr38_vgpr39 = IMPLICIT_DEF
# GCN: $vgpr40_vgpr41_vgpr42_vgpr43 = IMPLICIT_DEF
# GCN: $vgpr44_vgpr45_vgpr46_vgpr47 = IMPLICIT_DEF
# GCN: $vgpr2 = V_CNDMASK_B32_e64 0, $vgpr1, 0, $vgpr5, $vcc, implicit $exec
# GCN: $vgpr2 = V_CNDMASK_B32_e64 0, $vgpr0, 0, $vgpr4, killed $vcc, implicit $exec
# GCN: $sgpr0_sgpr1 = V_CMP_LT_U64_e64 killed $vgpr4_vgpr5, killed $vgpr0_vgpr1, implicit $exec
---
name: vgpr64_mixed_use
tracksRegLiveness: true
registers:
- { id: 0, class: vreg_64, preferred-register: '$vgpr0_vgpr1' }
- { id: 1, class: vreg_64, preferred-register: '$vgpr4_vgpr5' }
- { id: 2, class: sreg_64_xexec, preferred-register: '$vcc' }
- { id: 3, class: vgpr_32 }
- { id: 4, class: vgpr_32 }
- { id: 5, class: sreg_64_xexec }
- { id: 6, class: vreg_64, preferred-register: '$vgpr2_vgpr3' }
- { id: 7, class: vreg_64, preferred-register: '$vgpr6_vgpr7' }
- { id: 8, class: vreg_128, preferred-register: '$vgpr8_vgpr9_vgpr10_vgpr11' }
- { id: 9, class: vreg_128, preferred-register: '$vgpr12_vgpr13_vgpr14_vgpr15' }
- { id: 10, class: vreg_128, preferred-register: '$vgpr16_vgpr17_vgpr18_vgpr19' }
- { id: 11, class: vreg_128, preferred-register: '$vgpr20_vgpr21_vgpr22_vgpr23' }
- { id: 12, class: vreg_128, preferred-register: '$vgpr24_vgpr25_vgpr26_vgpr27' }
- { id: 13, class: vreg_128, preferred-register: '$vgpr28_vgpr29_vgpr30_vgpr31' }
- { id: 14, class: vreg_128, preferred-register: '$vgpr32_vgpr33_vgpr34_vgpr35' }
- { id: 15, class: vreg_128, preferred-register: '$vgpr36_vgpr37_vgpr38_vgpr39' }
- { id: 16, class: vreg_128, preferred-register: '$vgpr40_vgpr41_vgpr42_vgpr43' }
- { id: 17, class: vreg_128, preferred-register: '$vgpr44_vgpr45_vgpr46_vgpr47' }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
%2 = IMPLICIT_DEF
%6 = IMPLICIT_DEF
%7 = IMPLICIT_DEF
%8 = IMPLICIT_DEF
%9 = IMPLICIT_DEF
%10 = IMPLICIT_DEF
%11 = IMPLICIT_DEF
%12 = IMPLICIT_DEF
%13 = IMPLICIT_DEF
%14 = IMPLICIT_DEF
%15 = IMPLICIT_DEF
%16 = IMPLICIT_DEF
%17 = IMPLICIT_DEF
%3 = V_CNDMASK_B32_e64 0, %0.sub1, 0, %1.sub1, %2, implicit $exec
%4 = V_CNDMASK_B32_e64 0, %0.sub0, 0, %1.sub0, %2, implicit $exec
%5 = V_CMP_LT_U64_e64 %1, %0, implicit $exec
S_ENDPGM 0
...

llvm/test/CodeGen/AMDGPU/regbank-reassign.mir

  • This file was deleted.
# RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs -run-pass greedy,amdgpu-regbanks-reassign,virtregrewriter -o - %s | FileCheck -check-prefix=GCN %s
# GCN-LABEL: v1_vs_v5{{$}}
# GCN: V_AND_B32_e32 killed $vgpr3, killed $vgpr1,
---
name: v1_vs_v5
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr1' }
- { id: 1, class: vgpr_32, preferred-register: '$vgpr5' }
- { id: 2, class: vgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
%2 = V_AND_B32_e32 %1, %0, implicit $exec
S_ENDPGM 0
...
# GCN-LABEL: v0_1_vs_v4{{$}}
# GCN: GLOBAL_STORE_DWORD killed renamable $vgpr0_vgpr1, killed renamable $vgpr3,
---
name: v0_1_vs_v4
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr4' }
- { id: 1, class: vreg_64, preferred-register: '$vgpr0_vgpr1' }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
GLOBAL_STORE_DWORD %1, %0, 0, 0, implicit $exec
S_ENDPGM 0
...
# GCN-LABEL: v1_2_vs_v4_5{{$}}
# GCN: GLOBAL_STORE_DWORDX2 killed renamable $vgpr2_vgpr3, killed renamable $vgpr4_vgpr5,
---
name: v1_2_vs_v4_5
tracksRegLiveness: true
registers:
- { id: 0, class: vreg_64, preferred-register: '$vgpr4_vgpr5' }
- { id: 1, class: vreg_64, preferred-register: '$vgpr1_vgpr2' }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
GLOBAL_STORE_DWORDX2 %1, %0, 0, 0, implicit $exec
S_ENDPGM 0
...
# GCN-LABEL: s11_vs_vcc{{$}}
# GCN: $vgpr0, $vcc_lo = V_ADDC_U32_e64 killed $sgpr14, killed $vgpr0, killed $vcc_lo, 0
---
name: s11_vs_vcc
tracksRegLiveness: true
registers:
- { id: 0, class: sgpr_32, preferred-register: '$sgpr11' }
- { id: 1, class: vgpr_32 }
- { id: 2, class: vgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
$vcc_lo = IMPLICIT_DEF
%2, $vcc_lo = V_ADDC_U32_e64 killed %0, killed %1, killed $vcc_lo, 0, implicit $exec
S_ENDPGM 0
...
# GCN-LABEL: s0_vs_s16{{$}}
# GCN: S_AND_B32 killed renamable $sgpr14, $sgpr0,
---
name: s0_vs_s16
tracksRegLiveness: true
registers:
- { id: 0, class: sgpr_32, preferred-register: '$sgpr16' }
- { id: 1, class: sgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
$sgpr0 = IMPLICIT_DEF
%1 = S_AND_B32 %0, $sgpr0, implicit-def $scc
S_ENDPGM 0
...
# GCN-LABEL: s1_vs_s16{{$}}
# GCN: S_AND_B32 killed renamable $sgpr14, $sgpr1,
---
name: s1_vs_s16
tracksRegLiveness: true
registers:
- { id: 0, class: sgpr_32, preferred-register: '$sgpr16' }
- { id: 1, class: sgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
$sgpr1 = IMPLICIT_DEF
%1 = S_AND_B32 %0, $sgpr1, implicit-def $scc
S_ENDPGM 0
...
# GCN-LABEL: s12_vs_null{{$}}
# GCN: S_AND_B32 $sgpr_null, killed renamable $sgpr14,
---
name: s12_vs_null
tracksRegLiveness: true
registers:
- { id: 0, class: sgpr_32, preferred-register: '$sgpr12' }
- { id: 1, class: sgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = S_AND_B32 $sgpr_null, %0, implicit-def $scc
S_ENDPGM 0
...
# GCN-LABEL: s13_vs_m0{{$}}
# GCN: S_AND_B32 $m0, killed renamable $sgpr14,
---
name: s13_vs_m0
tracksRegLiveness: true
registers:
- { id: 0, class: sgpr_32, preferred-register: '$sgpr13' }
- { id: 1, class: sgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = S_AND_B32 $m0, %0, implicit-def $scc
S_ENDPGM 0
...
# GCN-LABEL: s12_13_vs_s28_s29{{$}}
# GCN: S_AND_B64 $sgpr28_sgpr29, killed renamable $sgpr14_sgpr15,
---
name: s12_13_vs_s28_s29
tracksRegLiveness: true
registers:
- { id: 0, class: sreg_64, preferred-register: '$sgpr12_sgpr13' }
- { id: 1, class: sreg_64 }
body: |
bb.0:
%0 = IMPLICIT_DEF
$sgpr28_sgpr29 = IMPLICIT_DEF
%1 = S_AND_B64 $sgpr28_sgpr29, %0, implicit-def $scc
S_ENDPGM 0
...
# GCN-LABEL: livein{{$}}
# GCN: V_AND_B32_e32 killed $vgpr4, killed $vgpr0,
---
name: livein
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr0' }
- { id: 1, class: vgpr_32, preferred-register: '$vgpr4' }
- { id: 2, class: vgpr_32 }
liveins:
- { reg: '$vgpr0', virtual-reg: '' }
- { reg: '$vgpr4', virtual-reg: '' }
body: |
bb.0:
liveins: $vgpr0, $vgpr4
%0 = COPY $vgpr0
%1 = COPY $vgpr4
%2 = V_AND_B32_e32 %1, %0, implicit $exec
S_ENDPGM 0
...
# GCN-LABEL: liveout{{$}}
# GCN: V_AND_B32_e32 $vgpr4, $vgpr0,
---
name: liveout
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr0' }
- { id: 1, class: vgpr_32, preferred-register: '$vgpr4' }
- { id: 2, class: vgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
%2 = V_AND_B32_e32 %1, %0, implicit $exec
$vgpr0 = COPY %0
$vgpr4 = COPY %1
S_ENDPGM 0
...
# GCN-LABEL: implicit{{$}}
# GCN: V_MOV_B32_indirect undef $vgpr4, undef $vgpr0, implicit $exec, implicit-def dead renamable $vgpr0_vgpr1_vgpr2_vgpr3, implicit killed $vgpr4_vgpr5_vgpr6_vgpr7, implicit $m0
---
name: implicit
tracksRegLiveness: true
registers:
- { id: 0, class: vreg_128 }
- { id: 1, class: vreg_128, preferred-register: '$vgpr4_vgpr5_vgpr6_vgpr7' }
body: |
bb.0:
%1 = IMPLICIT_DEF
V_MOV_B32_indirect undef %1.sub0:vreg_128, undef $vgpr0, implicit $exec, implicit-def %0:vreg_128, implicit %1:vreg_128, implicit $m0
S_ENDPGM 0
...
# GCN-LABEL: occupancy_limit{{$}}
# GCN: V_AND_B32_e32 $vgpr4, $vgpr0,
---
name: occupancy_limit
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr0' }
- { id: 1, class: vgpr_32, preferred-register: '$vgpr4' }
- { id: 2, class: vgpr_32, preferred-register: '$vgpr1' }
- { id: 3, class: vreg_64, preferred-register: '$vgpr2_vgpr3' }
- { id: 4, class: vgpr_32, preferred-register: '$vgpr5' }
- { id: 5, class: vreg_64, preferred-register: '$vgpr6_vgpr7' }
- { id: 6, class: vreg_128, preferred-register: '$vgpr8_vgpr9_vgpr10_vgpr11' }
- { id: 7, class: vreg_128, preferred-register: '$vgpr12_vgpr13_vgpr14_vgpr15' }
- { id: 8, class: vreg_128, preferred-register: '$vgpr16_vgpr17_vgpr18_vgpr19' }
- { id: 9, class: vreg_128, preferred-register: '$vgpr20_vgpr21_vgpr22_vgpr23' }
- { id: 10, class: vreg_128, preferred-register: '$vgpr24_vgpr25_vgpr26_vgpr27' }
- { id: 11, class: vreg_128, preferred-register: '$vgpr28_vgpr29_vgpr30_vgpr31' }
- { id: 12, class: vreg_128, preferred-register: '$vgpr32_vgpr33_vgpr34_vgpr35' }
- { id: 13, class: vreg_128, preferred-register: '$vgpr36_vgpr37_vgpr38_vgpr39' }
- { id: 14, class: vreg_128, preferred-register: '$vgpr40_vgpr41_vgpr42_vgpr43' }
- { id: 15, class: vreg_128, preferred-register: '$vgpr44_vgpr45_vgpr46_vgpr47' }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
%3 = IMPLICIT_DEF
%4 = IMPLICIT_DEF
%5 = IMPLICIT_DEF
%6 = IMPLICIT_DEF
%7 = IMPLICIT_DEF
%8 = IMPLICIT_DEF
%9 = IMPLICIT_DEF
%10 = IMPLICIT_DEF
%11 = IMPLICIT_DEF
%12 = IMPLICIT_DEF
%13 = IMPLICIT_DEF
%14 = IMPLICIT_DEF
%15 = IMPLICIT_DEF
%2 = V_AND_B32_e32 %1, %0, implicit $exec
GLOBAL_STORE_DWORD %3, %0, 0, 0, implicit $exec
GLOBAL_STORE_DWORD %3, %1, 0, 0, implicit $exec
GLOBAL_STORE_DWORD %3, %2, 0, 0, implicit $exec
GLOBAL_STORE_DWORD %3, %4, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX2 %3, %5, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX4 %3, %6, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX4 %3, %7, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX4 %3, %8, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX4 %3, %9, 0, 0, implicit $exec
S_ENDPGM 0
...
# GCN-LABEL: csr{{$}}
# GCN: V_AND_B32_e32 $vgpr37, $vgpr0,
---
name: csr
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr0' }
- { id: 1, class: vgpr_32, preferred-register: '$vgpr4' }
- { id: 2, class: vgpr_32, preferred-register: '$vgpr1' }
- { id: 3, class: vreg_64, preferred-register: '$vgpr2_vgpr3' }
- { id: 4, class: vgpr_32, preferred-register: '$vgpr5' }
- { id: 5, class: vreg_64, preferred-register: '$vgpr6_vgpr7' }
- { id: 6, class: vreg_128, preferred-register: '$vgpr8_vgpr9_vgpr10_vgpr11' }
- { id: 7, class: vreg_128, preferred-register: '$vgpr12_vgpr13_vgpr14_vgpr15' }
- { id: 8, class: vreg_128, preferred-register: '$vgpr16_vgpr17_vgpr18_vgpr19' }
- { id: 9, class: vreg_128, preferred-register: '$vgpr20_vgpr21_vgpr22_vgpr23' }
- { id: 10, class: vreg_128, preferred-register: '$vgpr24_vgpr25_vgpr26_vgpr27' }
- { id: 11, class: vreg_128, preferred-register: '$vgpr28_vgpr29_vgpr30_vgpr31' }
- { id: 12, class: vgpr_32, preferred-register: '$vgpr33' }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
%3 = IMPLICIT_DEF
%4 = IMPLICIT_DEF
%5 = IMPLICIT_DEF
%6 = IMPLICIT_DEF
%7 = IMPLICIT_DEF
%8 = IMPLICIT_DEF
%9 = IMPLICIT_DEF
%10 = IMPLICIT_DEF
%11 = IMPLICIT_DEF
%12 = IMPLICIT_DEF
%2 = V_AND_B32_e32 %1, %0, implicit $exec
GLOBAL_STORE_DWORD %3, %0, 0, 0, implicit $exec
GLOBAL_STORE_DWORD %3, %1, 0, 0, implicit $exec
GLOBAL_STORE_DWORD %3, %2, 0, 0, implicit $exec
GLOBAL_STORE_DWORD %3, %4, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX2 %3, %5, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX4 %3, %6, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX4 %3, %7, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX4 %3, %8, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX4 %3, %9, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX4 %3, %10, 0, 0, implicit $exec
GLOBAL_STORE_DWORDX4 %3, %11, 0, 0, implicit $exec
GLOBAL_STORE_DWORD %3, %12, 0, 0, implicit $exec
S_ENDPGM 0
...
# Do not touch undefs
# GCN-LABEL: s0_vs_s16_undef{{$}}
# GCN: S_AND_B32 killed renamable $sgpr16, undef $sgpr0,
---
name: s0_vs_s16_undef
tracksRegLiveness: true
registers:
- { id: 0, class: sgpr_32, preferred-register: '$sgpr16' }
- { id: 1, class: sgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = S_AND_B32 %0, undef $sgpr0, implicit-def $scc
S_ENDPGM 0
...
# GCN-LABEL: smem_bundle{{$}}
# GCN: S_BUFFER_LOAD_DWORD_SGPR renamable $sgpr0_sgpr1_sgpr2_sgpr3, renamable $sgpr14, 0
# GCN: S_BUFFER_LOAD_DWORD_SGPR renamable $sgpr0_sgpr1_sgpr2_sgpr3, renamable $sgpr15, 0
---
name: smem_bundle
tracksRegLiveness: true
registers:
- { id: 0, class: sgpr_128, preferred-register: '$sgpr0_sgpr1_sgpr2_sgpr3' }
- { id: 1, class: sreg_32_xm0_xexec, preferred-register: '$sgpr16' }
- { id: 2, class: sreg_32_xm0_xexec, preferred-register: '$sgpr17' }
- { id: 3, class: sreg_32_xm0_xexec, preferred-register: '$sgpr4' }
- { id: 4, class: sreg_32_xm0_xexec, preferred-register: '$sgpr5' }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
%2 = IMPLICIT_DEF
early-clobber %3, early-clobber %4 = BUNDLE %0, %1, %2 {
%3 = S_BUFFER_LOAD_DWORD_SGPR %0, %1, 0
%4 = S_BUFFER_LOAD_DWORD_SGPR %0, %2, 0
}
S_ENDPGM 0
...
# GCN-LABEL: vreg_512_subs{{$}}
# don't care about the assignment: this used to trigger an infinite loop
---
name: vreg_512_subs
tracksRegLiveness: true
registers:
- { id: 1, class: vreg_512, preferred-register: '$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15' }
- { id: 2, class: vgpr_32, preferred-register: '$vgpr28' }
body: |
bb.0:
%1 = IMPLICIT_DEF
%2 = IMPLICIT_DEF
DS_WRITE2_B32_gfx9 %2, %1.sub0, %1.sub1, 0, 1, 0, implicit $exec
DS_WRITE2_B32_gfx9 %2, %1.sub2, %1.sub3, 2, 3, 0, implicit $exec
DS_WRITE2_B32_gfx9 %2, %1.sub4, %1.sub5, 4, 5, 0, implicit $exec
DS_WRITE2_B32_gfx9 %2, %1.sub6, %1.sub7, 6, 7, 0, implicit $exec
DS_WRITE2_B32_gfx9 %2, %1.sub8, %1.sub9, 8, 9, 0, implicit $exec
DS_WRITE2_B32_gfx9 %2, %1.sub10, %1.sub11, 10, 11, 0, implicit $exec
DS_WRITE2_B32_gfx9 %2, %1.sub12, %1.sub13, 12, 13, 0, implicit $exec
DS_WRITE2_B32_gfx9 %2, %1.sub14, %1.sub15, 14, 15, 0, implicit $exec
S_ENDPGM 0
...
# GCN-LABEL: vgpr_lo16_sub{{$}}
# GCN: renamable $vgpr0 = V_AND_B32_e32 killed $vgpr3, killed $vgpr1, implicit $exec
# GCN: renamable $vgpr1_lo16 = COPY killed renamable $vgpr0_lo16
---
name: vgpr_lo16_sub
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr1' }
- { id: 1, class: vgpr_32, preferred-register: '$vgpr5' }
- { id: 2, class: vgpr_32 }
- { id: 3, class: vgpr_lo16 }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
%2 = V_AND_B32_e32 %1, %0, implicit $exec
%3 = COPY %2.lo16
$vgpr1_lo16 = COPY %3
SI_RETURN_TO_EPILOG $vgpr1_lo16
...
# GCN-LABEL: vgpr_lo16{{$}}
# GCN: $vgpr1_lo16 = COPY killed renamable $vgpr0_lo16
---
name: vgpr_lo16
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_lo16, preferred-register: '$vgpr4_lo16' }
body: |
bb.0:
liveins: $vgpr0_lo16
%0 = COPY $vgpr0_lo16
$vgpr1_lo16 = COPY %0
SI_RETURN_TO_EPILOG $vgpr1_lo16
...
# GCN-LABEL: vgpr_hi16_sub{{$}}
# GCN: renamable $vgpr0 = V_AND_B32_e32 killed $vgpr3, killed $vgpr1, implicit $exec
# GCN: renamable $vgpr1_hi16 = COPY killed renamable $vgpr0_hi16
---
name: vgpr_hi16_sub
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr1' }
- { id: 1, class: vgpr_32, preferred-register: '$vgpr5' }
- { id: 2, class: vgpr_32 }
- { id: 3, class: vgpr_hi16 }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
%2 = V_AND_B32_e32 %1, %0, implicit $exec
%3 = COPY %2.hi16
$vgpr1_hi16 = COPY %3
SI_RETURN_TO_EPILOG $vgpr1_hi16
...
# GCN-LABEL: vgpr_hi16{{$}}
# GCN: $vgpr1_hi16 = COPY killed renamable $vgpr0_hi16
---
name: vgpr_hi16
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_hi16, preferred-register: '$vgpr4_hi16' }
body: |
bb.0:
liveins: $vgpr0_hi16
%0 = COPY $vgpr0_hi16
$vgpr1_hi16 = COPY %0
SI_RETURN_TO_EPILOG $vgpr1_hi16
...
# GCN-LABEL: sgpr_lo16_sub{{$}}
# GCN: renamable $sgpr0 = S_AND_B32 killed renamable $sgpr14, $sgpr0, implicit-def $scc
# GCN: renamable $sgpr1_lo16 = COPY killed renamable $sgpr0_lo16
---
name: sgpr_lo16_sub
tracksRegLiveness: true
registers:
- { id: 0, class: sgpr_32, preferred-register: '$sgpr16' }
- { id: 1, class: sgpr_32 }
- { id: 2, class: sgpr_lo16 }
body: |
bb.0:
%0 = IMPLICIT_DEF
$sgpr0 = IMPLICIT_DEF
%1 = S_AND_B32 %0, $sgpr0, implicit-def $scc
%2 = COPY %1.lo16
$sgpr1_lo16 = COPY %2
SI_RETURN_TO_EPILOG $sgpr1_lo16
...
# GCN-LABEL: sgpr_lo16{{$}}
# GCN: $sgpr1_lo16 = COPY killed renamable $sgpr0_lo16
---
name: sgpr_lo16
tracksRegLiveness: true
registers:
- { id: 0, class: sgpr_lo16, preferred-register: '$sgpr4_lo16' }
body: |
bb.0:
liveins: $sgpr0_lo16
%0 = COPY $sgpr0_lo16
$sgpr1_lo16 = COPY %0
SI_RETURN_TO_EPILOG $sgpr1_lo16
...
# Check that we do not use VGPR3 which we would use otherwise.
# We cannot use it because of interference with VGPR3_LO16.
# GCN-LABEL: v1_vs_v5_src_interence{{$}}
# GCN: V_AND_B32_e32 killed $vgpr7, killed $vgpr1,
---
name: v1_vs_v5_src_interence
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr1' }
- { id: 1, class: vgpr_32, preferred-register: '$vgpr5' }
- { id: 2, class: vgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
$vgpr3_lo16 = IMPLICIT_DEF
%2 = V_AND_B32_e32 %1, %0, implicit $exec
S_ENDPGM 0
...
# Test that bank of subreg is considered during scavenging.
# If handled incorrectly an infinite loop occurs.
# GCN-LABEL: s0_vs_s15_16_17_sub1{{$}}
# GCN: S_AND_B32 killed renamable $sgpr13, $sgpr0,
---
name: s0_vs_s15_16_17_sub1
tracksRegLiveness: true
registers:
- { id: 0, class: sgpr_96, preferred-register: '$sgpr15_sgpr16_sgpr17' }
- { id: 1, class: sgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
$sgpr0 = IMPLICIT_DEF
%1 = S_AND_B32 %0.sub1, $sgpr0, implicit-def $scc
S_ENDPGM 0
...
# Test that the size of subreg is correctly handled in bank calculation.
# If handled incorrectly an infinite loop occurs.
# GCN-LABEL: vgpr_sub_dependence{{$}}
# GCN: $vgpr9_vgpr10_vgpr11_vgpr12 = IMPLICIT_DEF
# GCN: $vgpr16_vgpr17 = IMPLICIT_DEF
# GCN: $vgpr14_vgpr15 = IMPLICIT_DEF
# GCN: $vgpr0_vgpr1 = IMPLICIT_DEF
# GCN: $vgpr7_vgpr8 = IMPLICIT_DEF
# GCN: $vgpr3_vgpr4_vgpr5_vgpr6 = IMPLICIT_DEF
# GCN: $vgpr18_vgpr19 = IMPLICIT_DEF
# GCN: $vgpr20_vgpr21_vgpr22_vgpr23 = IMPLICIT_DEF
# GCN: $vgpr24_vgpr25_vgpr26_vgpr27 = IMPLICIT_DEF
# GCN: $vgpr28_vgpr29_vgpr30_vgpr31 = IMPLICIT_DEF
# GCN: $vgpr32_vgpr33_vgpr34_vgpr35 = IMPLICIT_DEF
# GCN: $vgpr36_vgpr37_vgpr38_vgpr39 = IMPLICIT_DEF
# GCN: $vgpr40_vgpr41_vgpr42_vgpr43 = IMPLICIT_DEF
# GCN: $vgpr44_vgpr45_vgpr46_vgpr47 = IMPLICIT_DEF
# GCN: $vgpr0_vgpr1 = V_ADD_F64_e64 0, $vgpr11_vgpr12, 0, killed $vgpr16_vgpr17, 0, 0, implicit $mode, implicit $exec
# GCN: $vgpr0_vgpr1 = V_ADD_F64_e64 0, killed $vgpr9_vgpr10, 0, killed $vgpr14_vgpr15, 0, 0, implicit $mode, implicit $exec
---
name: vgpr_sub_dependence
tracksRegLiveness: true
registers:
- { id: 0, class: vreg_128, preferred-register: '$vgpr10_vgpr11_vgpr12_vgpr13' }
- { id: 1, class: vreg_64, preferred-register: '$vgpr16_vgpr17' }
- { id: 2, class: vreg_64, preferred-register: '$vgpr14_vgpr15' }
- { id: 3, class: vreg_64 }
- { id: 4, class: vreg_64 }
- { id: 5, class: vreg_64, preferred-register: '$vgpr0_vgpr1' }
- { id: 6, class: vreg_64, preferred-register: '$vgpr7_vgpr8' }
- { id: 7, class: vreg_128, preferred-register: '$vgpr3_vgpr4_vgpr5_vgpr6' }
- { id: 8, class: vreg_64, preferred-register: '$vgpr18_vgpr19' }
- { id: 9, class: vreg_128, preferred-register: '$vgpr20_vgpr21_vgpr22_vgpr23' }
- { id: 10, class: vreg_128, preferred-register: '$vgpr24_vgpr25_vgpr26_vgpr27' }
- { id: 11, class: vreg_128, preferred-register: '$vgpr28_vgpr29_vgpr30_vgpr31' }
- { id: 12, class: vreg_128, preferred-register: '$vgpr32_vgpr33_vgpr34_vgpr35' }
- { id: 13, class: vreg_128, preferred-register: '$vgpr36_vgpr37_vgpr38_vgpr39' }
- { id: 14, class: vreg_128, preferred-register: '$vgpr40_vgpr41_vgpr42_vgpr43' }
- { id: 15, class: vreg_128, preferred-register: '$vgpr44_vgpr45_vgpr46_vgpr47' }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
%2 = IMPLICIT_DEF
%5 = IMPLICIT_DEF
%6 = IMPLICIT_DEF
%7 = IMPLICIT_DEF
%8 = IMPLICIT_DEF
%9 = IMPLICIT_DEF
%10 = IMPLICIT_DEF
%11 = IMPLICIT_DEF
%12 = IMPLICIT_DEF
%13 = IMPLICIT_DEF
%14 = IMPLICIT_DEF
%15 = IMPLICIT_DEF
%3 = V_ADD_F64_e64 0, %0.sub2_sub3:vreg_128, 0, %1:vreg_64, 0, 0, implicit $mode, implicit $exec
%4 = V_ADD_F64_e64 0, %0.sub0_sub1:vreg_128, 0, %2:vreg_64, 0, 0, implicit $mode, implicit $exec
S_ENDPGM 0
...
# GCN-LABEL: dbg_value_v1_v5{{$}}
# GCN: renamable $vgpr1 = IMPLICIT_DEF
# GCN: renamable $vgpr5 = IMPLICIT_DEF
---
name: dbg_value_v1_v5
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr1' }
- { id: 1, class: vgpr_32, preferred-register: '$vgpr5' }
- { id: 2, class: vgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
DBG_VALUE debug-use %1, debug-use %0
S_ENDPGM 0, implicit %0, implicit %1
...
# GCN-LABEL: kill_v1_v5{{$}}
# GCN: renamable $vgpr1 = IMPLICIT_DEF
# GCN: renamable $vgpr5 = IMPLICIT_DEF
# GCN: KILL killed renamable $vgpr5, killed renamable $vgpr1
---
name: kill_v1_v5
tracksRegLiveness: true
registers:
- { id: 0, class: vgpr_32, preferred-register: '$vgpr1' }
- { id: 1, class: vgpr_32, preferred-register: '$vgpr5' }
- { id: 2, class: vgpr_32 }
body: |
bb.0:
%0 = IMPLICIT_DEF
%1 = IMPLICIT_DEF
KILL %1, %0
S_ENDPGM 0
...

llvm/test/CodeGen/AMDGPU/saddo.ll

Show First 20 Lines • Show All 452 Lines • ▼ Show 20 Lines
; GFX9-NEXT: s_endpgm ; GFX9-NEXT: s_endpgm
; ;
; GFX10-LABEL: v_saddo_i64: ; GFX10-LABEL: v_saddo_i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx8 s[4:11], s[0:1], 0x24 ; GFX10-NEXT: s_load_dwordx8 s[4:11], s[0:1], 0x24
; GFX10-NEXT: v_mov_b32_e32 v6, 0 ; GFX10-NEXT: v_mov_b32_e32 v6, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx2 v[9:10], v6, s[8:9] ; GFX10-NEXT: global_load_dwordx2 v[0:1], v6, s[8:9]
; GFX10-NEXT: global_load_dwordx2 v[2:3], v6, s[10:11] ; GFX10-NEXT: global_load_dwordx2 v[2:3], v6, s[10:11]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_add_co_u32 v7, vcc_lo, v9, v2 ; GFX10-NEXT: v_add_co_u32 v4, vcc_lo, v0, v2
; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, v10, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, 0, v[2:3] ; GFX10-NEXT: v_cmp_gt_i64_e32 vcc_lo, 0, v[2:3]
; GFX10-NEXT: v_cmp_lt_i64_e64 s0, v[7:8], v[9:10] ; GFX10-NEXT: v_cmp_lt_i64_e64 s0, v[4:5], v[0:1]
; GFX10-NEXT: s_xor_b32 s0, vcc_lo, s0 ; GFX10-NEXT: s_xor_b32 s0, vcc_lo, s0
; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0 ; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0
; GFX10-NEXT: global_store_dwordx2 v6, v[7:8], s[4:5] ; GFX10-NEXT: global_store_dwordx2 v6, v[4:5], s[4:5]
; GFX10-NEXT: global_store_byte v6, v0, s[6:7] ; GFX10-NEXT: global_store_byte v6, v0, s[6:7]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%a = load i64, i64 addrspace(1)* %aptr, align 4 %a = load i64, i64 addrspace(1)* %aptr, align 4
%b = load i64, i64 addrspace(1)* %bptr, align 4 %b = load i64, i64 addrspace(1)* %bptr, align 4
%sadd = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 %a, i64 %b) nounwind %sadd = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 %a, i64 %b) nounwind
%val = extractvalue { i64, i1 } %sadd, 0 %val = extractvalue { i64, i1 } %sadd, 0
%carry = extractvalue { i64, i1 } %sadd, 1 %carry = extractvalue { i64, i1 } %sadd, 1
store i64 %val, i64 addrspace(1)* %out, align 8 store i64 %val, i64 addrspace(1)* %out, align 8
▲ Show 20 Lines • Show All 91 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_load_dwordx8 s[0:7], s[0:1], 0x24 ; GFX10-NEXT: s_load_dwordx8 s[0:7], s[0:1], 0x24
; GFX10-NEXT: v_mov_b32_e32 v4, 0 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[4:5] ; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[4:5]
; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[6:7] ; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[6:7]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_add_nc_i32 v5, v1, v3 clamp ; GFX10-NEXT: v_add_nc_i32 v5, v1, v3 clamp
; GFX10-NEXT: v_add_nc_u32_e32 v10, v1, v3 ; GFX10-NEXT: v_add_nc_u32_e32 v1, v1, v3
; GFX10-NEXT: v_add_nc_i32 v6, v0, v2 clamp ; GFX10-NEXT: v_add_nc_i32 v6, v0, v2 clamp
; GFX10-NEXT: v_add_nc_u32_e32 v9, v0, v2 ; GFX10-NEXT: v_add_nc_u32_e32 v0, v0, v2
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, v10, v5 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, v9, v6 ; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, v0, v6
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX10-NEXT: global_store_dwordx2 v4, v[9:10], s[0:1] ; GFX10-NEXT: global_store_dwordx2 v4, v[0:1], s[0:1]
; GFX10-NEXT: global_store_dwordx2 v4, v[2:3], s[2:3] ; GFX10-NEXT: global_store_dwordx2 v4, v[2:3], s[2:3]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%a = load <2 x i32>, <2 x i32> addrspace(1)* %aptr, align 4 %a = load <2 x i32>, <2 x i32> addrspace(1)* %aptr, align 4
%b = load <2 x i32>, <2 x i32> addrspace(1)* %bptr, align 4 %b = load <2 x i32>, <2 x i32> addrspace(1)* %bptr, align 4
%sadd = call { <2 x i32>, <2 x i1> } @llvm.sadd.with.overflow.v2i32(<2 x i32> %a, <2 x i32> %b) nounwind %sadd = call { <2 x i32>, <2 x i1> } @llvm.sadd.with.overflow.v2i32(<2 x i32> %a, <2 x i32> %b) nounwind
%val = extractvalue { <2 x i32>, <2 x i1> } %sadd, 0 %val = extractvalue { <2 x i32>, <2 x i1> } %sadd, 0
%carry = extractvalue { <2 x i32>, <2 x i1> } %sadd, 1 %carry = extractvalue { <2 x i32>, <2 x i1> } %sadd, 1
store <2 x i32> %val, <2 x i32> addrspace(1)* %out, align 4 store <2 x i32> %val, <2 x i32> addrspace(1)* %out, align 4
%carry.ext = zext <2 x i1> %carry to <2 x i32> %carry.ext = zext <2 x i1> %carry to <2 x i32>
store <2 x i32> %carry.ext, <2 x i32> addrspace(1)* %carryout store <2 x i32> %carry.ext, <2 x i32> addrspace(1)* %carryout
ret void ret void
} }

llvm/test/CodeGen/AMDGPU/saddsat.ll

Show First 20 Lines • Show All 480 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_saddsat_i64: ; GFX10-LABEL: v_saddsat_i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_add_co_u32 v10, vcc_lo, v0, v2 ; GFX10-NEXT: v_add_co_u32 v4, vcc_lo, v0, v2
; GFX10-NEXT: v_bfrev_b32_e32 v6, -2 ; GFX10-NEXT: v_bfrev_b32_e32 v6, -2
; GFX10-NEXT: v_add_co_ci_u32_e32 v11, vcc_lo, v1, v3, vcc_lo ; GFX10-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_gt_i64_e64 s4, 0, v[2:3] ; GFX10-NEXT: v_cmp_gt_i64_e64 s4, 0, v[2:3]
; GFX10-NEXT: v_cmp_gt_i64_e64 s5, 0, v[10:11] ; GFX10-NEXT: v_cmp_gt_i64_e64 s5, 0, v[4:5]
; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[10:11], v[0:1] ; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[4:5], v[0:1]
; GFX10-NEXT: v_ashrrev_i32_e32 v0, 31, v11 ; GFX10-NEXT: v_ashrrev_i32_e32 v0, 31, v5
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0x80000000, v6, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v1, 0x80000000, v6, s5
; GFX10-NEXT: s_xor_b32 vcc_lo, s4, vcc_lo ; GFX10-NEXT: s_xor_b32 vcc_lo, s4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v0, v10, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v11, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i64 @llvm.sadd.sat.i64(i64 %lhs, i64 %rhs) %result = call i64 @llvm.sadd.sat.i64(i64 %lhs, i64 %rhs)
ret i64 %result ret i64 %result
} }
declare i8 @llvm.sadd.sat.i8(i8, i8) #0 declare i8 @llvm.sadd.sat.i8(i8, i8) #0
declare i16 @llvm.sadd.sat.i16(i16, i16) #0 declare i16 @llvm.sadd.sat.i16(i16, i16) #0
declare <2 x i16> @llvm.sadd.sat.v2i16(<2 x i16>, <2 x i16>) #0 declare <2 x i16> @llvm.sadd.sat.v2i16(<2 x i16>, <2 x i16>) #0
declare <3 x i16> @llvm.sadd.sat.v3i16(<3 x i16>, <3 x i16>) #0 declare <3 x i16> @llvm.sadd.sat.v3i16(<3 x i16>, <3 x i16>) #0
declare <4 x i16> @llvm.sadd.sat.v4i16(<4 x i16>, <4 x i16>) #0 declare <4 x i16> @llvm.sadd.sat.v4i16(<4 x i16>, <4 x i16>) #0
declare i32 @llvm.sadd.sat.i32(i32, i32) #0 declare i32 @llvm.sadd.sat.i32(i32, i32) #0
declare <2 x i32> @llvm.sadd.sat.v2i32(<2 x i32>, <2 x i32>) #0 declare <2 x i32> @llvm.sadd.sat.v2i32(<2 x i32>, <2 x i32>) #0
declare i64 @llvm.sadd.sat.i64(i64, i64) #0 declare i64 @llvm.sadd.sat.i64(i64, i64) #0

llvm/test/CodeGen/AMDGPU/shl.v2i16.ll

Show First 20 Lines • Show All 538 Lines • ▼ Show 20 Lines
; CI-NEXT: v_or_b32_e32 v3, v3, v5 ; CI-NEXT: v_or_b32_e32 v3, v3, v5
; CI-NEXT: v_or_b32_e32 v2, v2, v4 ; CI-NEXT: v_or_b32_e32 v2, v2, v4
; CI-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64 ; CI-NEXT: buffer_store_dwordx2 v[2:3], v[0:1], s[4:7], 0 addr64
; CI-NEXT: s_endpgm ; CI-NEXT: s_endpgm
; ;
; GFX10-LABEL: v_shl_v4i16: ; GFX10-LABEL: v_shl_v4i16:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24 ; GFX10-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX10-NEXT: v_lshlrev_b32_e32 v7, 3, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v4, 3, v0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: global_load_dwordx2 v[0:1], v7, s[2:3] ; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[2:3]
; GFX10-NEXT: global_load_dwordx2 v[2:3], v7, s[2:3] offset:8 ; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[2:3] offset:8
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_pk_lshlrev_b16 v1, v3, v1 ; GFX10-NEXT: v_pk_lshlrev_b16 v1, v3, v1
; GFX10-NEXT: v_pk_lshlrev_b16 v0, v2, v0 ; GFX10-NEXT: v_pk_lshlrev_b16 v0, v2, v0
; GFX10-NEXT: global_store_dwordx2 v7, v[0:1], s[0:1] ; GFX10-NEXT: global_store_dwordx2 v4, v[0:1], s[0:1]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x() %tid = call i32 @llvm.amdgcn.workitem.id.x()
%tid.ext = sext i32 %tid to i64 %tid.ext = sext i32 %tid to i64
%in.gep = getelementptr inbounds <4 x i16>, <4 x i16> addrspace(1)* %in, i64 %tid.ext %in.gep = getelementptr inbounds <4 x i16>, <4 x i16> addrspace(1)* %in, i64 %tid.ext
%out.gep = getelementptr inbounds <4 x i16>, <4 x i16> addrspace(1)* %out, i64 %tid.ext %out.gep = getelementptr inbounds <4 x i16>, <4 x i16> addrspace(1)* %out, i64 %tid.ext
%b_ptr = getelementptr <4 x i16>, <4 x i16> addrspace(1)* %in.gep, i32 1 %b_ptr = getelementptr <4 x i16>, <4 x i16> addrspace(1)* %in.gep, i32 1
%a = load <4 x i16>, <4 x i16> addrspace(1)* %in.gep %a = load <4 x i16>, <4 x i16> addrspace(1)* %in.gep
%b = load <4 x i16>, <4 x i16> addrspace(1)* %b_ptr %b = load <4 x i16>, <4 x i16> addrspace(1)* %b_ptr
▲ Show 20 Lines • Show All 92 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/ssubsat.ll

Show First 20 Lines • Show All 1,094 Lines • ▼ Show 20 Lines
; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc ; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_ssubsat_i64: ; GFX10-LABEL: v_ssubsat_i64:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_sub_co_u32 v10, vcc_lo, v0, v2 ; GFX10-NEXT: v_sub_co_u32 v4, vcc_lo, v0, v2
; GFX10-NEXT: v_bfrev_b32_e32 v6, -2 ; GFX10-NEXT: v_bfrev_b32_e32 v6, -2
; GFX10-NEXT: v_sub_co_ci_u32_e32 v11, vcc_lo, v1, v3, vcc_lo ; GFX10-NEXT: v_sub_co_ci_u32_e32 v5, vcc_lo, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_lt_i64_e64 s4, 0, v[2:3] ; GFX10-NEXT: v_cmp_lt_i64_e64 s4, 0, v[2:3]
; GFX10-NEXT: v_cmp_gt_i64_e64 s5, 0, v[10:11] ; GFX10-NEXT: v_cmp_gt_i64_e64 s5, 0, v[4:5]
; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[10:11], v[0:1] ; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, v[4:5], v[0:1]
; GFX10-NEXT: v_ashrrev_i32_e32 v0, 31, v11 ; GFX10-NEXT: v_ashrrev_i32_e32 v0, 31, v5
; GFX10-NEXT: v_cndmask_b32_e64 v1, 0x80000000, v6, s5 ; GFX10-NEXT: v_cndmask_b32_e64 v1, 0x80000000, v6, s5
; GFX10-NEXT: s_xor_b32 vcc_lo, s4, vcc_lo ; GFX10-NEXT: s_xor_b32 vcc_lo, s4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v0, v10, v0, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v11, v1, vcc_lo ; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i64 @llvm.ssub.sat.i64(i64 %lhs, i64 %rhs) %result = call i64 @llvm.ssub.sat.i64(i64 %lhs, i64 %rhs)
ret i64 %result ret i64 %result
} }
declare i8 @llvm.ssub.sat.i8(i8, i8) #0 declare i8 @llvm.ssub.sat.i8(i8, i8) #0
declare i16 @llvm.ssub.sat.i16(i16, i16) #0 declare i16 @llvm.ssub.sat.i16(i16, i16) #0
declare <2 x i16> @llvm.ssub.sat.v2i16(<2 x i16>, <2 x i16>) #0 declare <2 x i16> @llvm.ssub.sat.v2i16(<2 x i16>, <2 x i16>) #0
Show All 11 Lines

llvm/test/CodeGen/AMDGPU/store-local.128.ll

Show First 20 Lines • Show All 224 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_lshr_b32 s3, s6, 24 ; GFX10-NEXT: s_lshr_b32 s3, s6, 24
; GFX10-NEXT: s_lshr_b32 s0, s7, 8 ; GFX10-NEXT: s_lshr_b32 s0, s7, 8
; GFX10-NEXT: s_lshr_b32 s2, s6, 8 ; GFX10-NEXT: s_lshr_b32 s2, s6, 8
; GFX10-NEXT: v_mov_b32_e32 v2, s6 ; GFX10-NEXT: v_mov_b32_e32 v2, s6
; GFX10-NEXT: s_lshr_b32 s6, s5, 8 ; GFX10-NEXT: s_lshr_b32 s6, s5, 8
; GFX10-NEXT: v_mov_b32_e32 v3, s5 ; GFX10-NEXT: v_mov_b32_e32 v3, s5
; GFX10-NEXT: s_lshr_b32 s1, s7, 24 ; GFX10-NEXT: s_lshr_b32 s1, s7, 24
; GFX10-NEXT: s_lshr_b32 s5, s5, 24 ; GFX10-NEXT: s_lshr_b32 s5, s5, 24
; GFX10-NEXT: v_mov_b32_e32 v15, s3 ; GFX10-NEXT: v_mov_b32_e32 v8, s3
; GFX10-NEXT: v_mov_b32_e32 v5, s0 ; GFX10-NEXT: v_mov_b32_e32 v5, s0
; GFX10-NEXT: v_mov_b32_e32 v9, s6 ; GFX10-NEXT: v_mov_b32_e32 v9, s6
; GFX10-NEXT: s_lshr_b32 s0, s4, 8 ; GFX10-NEXT: s_lshr_b32 s0, s4, 8
; GFX10-NEXT: v_mov_b32_e32 v6, s1 ; GFX10-NEXT: v_mov_b32_e32 v6, s1
; GFX10-NEXT: v_mov_b32_e32 v11, s4 ; GFX10-NEXT: v_mov_b32_e32 v4, s4
; GFX10-NEXT: v_mov_b32_e32 v7, s2 ; GFX10-NEXT: v_mov_b32_e32 v7, s2
; GFX10-NEXT: ds_write_b8 v0, v1 offset:12 ; GFX10-NEXT: ds_write_b8 v0, v1 offset:12
; GFX10-NEXT: ds_write_b8_d16_hi v0, v1 offset:14 ; GFX10-NEXT: ds_write_b8_d16_hi v0, v1 offset:14
; GFX10-NEXT: ds_write_b8 v0, v2 offset:8 ; GFX10-NEXT: ds_write_b8 v0, v2 offset:8
; GFX10-NEXT: ds_write_b8_d16_hi v0, v2 offset:10 ; GFX10-NEXT: ds_write_b8_d16_hi v0, v2 offset:10
; GFX10-NEXT: ds_write_b8 v0, v3 offset:4 ; GFX10-NEXT: ds_write_b8 v0, v3 offset:4
; GFX10-NEXT: ds_write_b8_d16_hi v0, v3 offset:6 ; GFX10-NEXT: ds_write_b8_d16_hi v0, v3 offset:6
; GFX10-NEXT: ds_write_b8 v0, v11 ; GFX10-NEXT: ds_write_b8 v0, v4
; GFX10-NEXT: ds_write_b8_d16_hi v0, v11 offset:2 ; GFX10-NEXT: ds_write_b8_d16_hi v0, v4 offset:2
; GFX10-NEXT: ds_write_b8 v0, v5 offset:13 ; GFX10-NEXT: ds_write_b8 v0, v5 offset:13
; GFX10-NEXT: ds_write_b8 v0, v6 offset:15 ; GFX10-NEXT: ds_write_b8 v0, v6 offset:15
; GFX10-NEXT: ds_write_b8 v0, v7 offset:9 ; GFX10-NEXT: ds_write_b8 v0, v7 offset:9
; GFX10-NEXT: s_lshr_b32 s1, s4, 24 ; GFX10-NEXT: s_lshr_b32 s1, s4, 24
; GFX10-NEXT: v_mov_b32_e32 v1, s5 ; GFX10-NEXT: v_mov_b32_e32 v1, s5
; GFX10-NEXT: v_mov_b32_e32 v2, s0 ; GFX10-NEXT: v_mov_b32_e32 v2, s0
; GFX10-NEXT: v_mov_b32_e32 v3, s1 ; GFX10-NEXT: v_mov_b32_e32 v3, s1
; GFX10-NEXT: ds_write_b8 v0, v15 offset:11 ; GFX10-NEXT: ds_write_b8 v0, v8 offset:11
; GFX10-NEXT: ds_write_b8 v0, v9 offset:5 ; GFX10-NEXT: ds_write_b8 v0, v9 offset:5
; GFX10-NEXT: ds_write_b8 v0, v1 offset:7 ; GFX10-NEXT: ds_write_b8 v0, v1 offset:7
; GFX10-NEXT: ds_write_b8 v0, v2 offset:1 ; GFX10-NEXT: ds_write_b8 v0, v2 offset:1
; GFX10-NEXT: ds_write_b8 v0, v3 offset:3 ; GFX10-NEXT: ds_write_b8 v0, v3 offset:3
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
store <4 x i32> %x, <4 x i32> addrspace(3)* %out, align 1 store <4 x i32> %x, <4 x i32> addrspace(3)* %out, align 1
ret void ret void
} }
▲ Show 20 Lines • Show All 82 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: s_load_dword s2, s[0:1], 0x24 ; GFX10-NEXT: s_load_dword s2, s[0:1], 0x24
; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x34 ; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x34
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, s2 ; GFX10-NEXT: v_mov_b32_e32 v0, s2
; GFX10-NEXT: v_mov_b32_e32 v1, s7 ; GFX10-NEXT: v_mov_b32_e32 v1, s7
; GFX10-NEXT: v_mov_b32_e32 v2, s6 ; GFX10-NEXT: v_mov_b32_e32 v2, s6
; GFX10-NEXT: v_mov_b32_e32 v3, s5 ; GFX10-NEXT: v_mov_b32_e32 v3, s5
; GFX10-NEXT: v_mov_b32_e32 v7, s4 ; GFX10-NEXT: v_mov_b32_e32 v4, s4
; GFX10-NEXT: ds_write_b16 v0, v1 offset:12 ; GFX10-NEXT: ds_write_b16 v0, v1 offset:12
; GFX10-NEXT: ds_write_b16_d16_hi v0, v1 offset:14 ; GFX10-NEXT: ds_write_b16_d16_hi v0, v1 offset:14
; GFX10-NEXT: ds_write_b16 v0, v2 offset:8 ; GFX10-NEXT: ds_write_b16 v0, v2 offset:8
; GFX10-NEXT: ds_write_b16_d16_hi v0, v2 offset:10 ; GFX10-NEXT: ds_write_b16_d16_hi v0, v2 offset:10
; GFX10-NEXT: ds_write_b16 v0, v3 offset:4 ; GFX10-NEXT: ds_write_b16 v0, v3 offset:4
; GFX10-NEXT: ds_write_b16_d16_hi v0, v3 offset:6 ; GFX10-NEXT: ds_write_b16_d16_hi v0, v3 offset:6
; GFX10-NEXT: ds_write_b16 v0, v7 ; GFX10-NEXT: ds_write_b16 v0, v4
; GFX10-NEXT: ds_write_b16_d16_hi v0, v7 offset:2 ; GFX10-NEXT: ds_write_b16_d16_hi v0, v4 offset:2
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
store <4 x i32> %x, <4 x i32> addrspace(3)* %out, align 2 store <4 x i32> %x, <4 x i32> addrspace(3)* %out, align 2
ret void ret void
} }
define amdgpu_kernel void @store_lds_v4i32_align4(<4 x i32> addrspace(3)* %out, <4 x i32> %x) { define amdgpu_kernel void @store_lds_v4i32_align4(<4 x i32> addrspace(3)* %out, <4 x i32> %x) {
; GFX9-LABEL: store_lds_v4i32_align4: ; GFX9-LABEL: store_lds_v4i32_align4:
; GFX9: ; %bb.0: ; GFX9: ; %bb.0:
▲ Show 20 Lines • Show All 44 Lines • ▼ Show 20 Lines
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: s_load_dword s2, s[0:1], 0x24 ; GFX10-NEXT: s_load_dword s2, s[0:1], 0x24
; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x34 ; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x34
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, s2 ; GFX10-NEXT: v_mov_b32_e32 v0, s2
; GFX10-NEXT: v_mov_b32_e32 v1, s4 ; GFX10-NEXT: v_mov_b32_e32 v1, s4
; GFX10-NEXT: v_mov_b32_e32 v2, s5 ; GFX10-NEXT: v_mov_b32_e32 v2, s5
; GFX10-NEXT: v_mov_b32_e32 v3, s6 ; GFX10-NEXT: v_mov_b32_e32 v3, s6
; GFX10-NEXT: v_mov_b32_e32 v6, s7 ; GFX10-NEXT: v_mov_b32_e32 v4, s7
; GFX10-NEXT: ds_write2_b32 v0, v1, v2 offset1:1 ; GFX10-NEXT: ds_write2_b32 v0, v1, v2 offset1:1
; GFX10-NEXT: ds_write2_b32 v0, v3, v6 offset0:2 offset1:3 ; GFX10-NEXT: ds_write2_b32 v0, v3, v4 offset0:2 offset1:3
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
store <4 x i32> %x, <4 x i32> addrspace(3)* %out, align 4 store <4 x i32> %x, <4 x i32> addrspace(3)* %out, align 4
ret void ret void
} }
define amdgpu_kernel void @store_lds_v4i32_align8(<4 x i32> addrspace(3)* %out, <4 x i32> %x) { define amdgpu_kernel void @store_lds_v4i32_align8(<4 x i32> addrspace(3)* %out, <4 x i32> %x) {
; GFX9-LABEL: store_lds_v4i32_align8: ; GFX9-LABEL: store_lds_v4i32_align8:
; GFX9: ; %bb.0: ; GFX9: ; %bb.0:
▲ Show 20 Lines • Show All 114 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/store-local.96.ll

Show First 20 Lines • Show All 190 Lines • ▼ Show 20 Lines
; GFX10-NEXT: v_mov_b32_e32 v2, s5 ; GFX10-NEXT: v_mov_b32_e32 v2, s5
; GFX10-NEXT: s_lshr_b32 s0, s6, 8 ; GFX10-NEXT: s_lshr_b32 s0, s6, 8
; GFX10-NEXT: s_lshr_b32 s1, s6, 24 ; GFX10-NEXT: s_lshr_b32 s1, s6, 24
; GFX10-NEXT: s_lshr_b32 s2, s5, 8 ; GFX10-NEXT: s_lshr_b32 s2, s5, 8
; GFX10-NEXT: s_lshr_b32 s3, s5, 24 ; GFX10-NEXT: s_lshr_b32 s3, s5, 24
; GFX10-NEXT: s_lshr_b32 s5, s4, 8 ; GFX10-NEXT: s_lshr_b32 s5, s4, 8
; GFX10-NEXT: v_mov_b32_e32 v3, s4 ; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_lshr_b32 s4, s4, 24 ; GFX10-NEXT: s_lshr_b32 s4, s4, 24
; GFX10-NEXT: v_mov_b32_e32 v11, s0 ; GFX10-NEXT: v_mov_b32_e32 v4, s0
; GFX10-NEXT: v_mov_b32_e32 v5, s1 ; GFX10-NEXT: v_mov_b32_e32 v5, s1
; GFX10-NEXT: v_mov_b32_e32 v6, s2 ; GFX10-NEXT: v_mov_b32_e32 v6, s2
; GFX10-NEXT: v_mov_b32_e32 v7, s3 ; GFX10-NEXT: v_mov_b32_e32 v7, s3
; GFX10-NEXT: v_mov_b32_e32 v15, s5 ; GFX10-NEXT: v_mov_b32_e32 v8, s5
; GFX10-NEXT: v_mov_b32_e32 v9, s4 ; GFX10-NEXT: v_mov_b32_e32 v9, s4
; GFX10-NEXT: ds_write_b8 v0, v1 offset:8 ; GFX10-NEXT: ds_write_b8 v0, v1 offset:8
; GFX10-NEXT: ds_write_b8_d16_hi v0, v1 offset:10 ; GFX10-NEXT: ds_write_b8_d16_hi v0, v1 offset:10
; GFX10-NEXT: ds_write_b8 v0, v2 offset:4 ; GFX10-NEXT: ds_write_b8 v0, v2 offset:4
; GFX10-NEXT: ds_write_b8_d16_hi v0, v2 offset:6 ; GFX10-NEXT: ds_write_b8_d16_hi v0, v2 offset:6
; GFX10-NEXT: ds_write_b8 v0, v3 ; GFX10-NEXT: ds_write_b8 v0, v3
; GFX10-NEXT: ds_write_b8_d16_hi v0, v3 offset:2 ; GFX10-NEXT: ds_write_b8_d16_hi v0, v3 offset:2
; GFX10-NEXT: ds_write_b8 v0, v11 offset:9 ; GFX10-NEXT: ds_write_b8 v0, v4 offset:9
; GFX10-NEXT: ds_write_b8 v0, v5 offset:11 ; GFX10-NEXT: ds_write_b8 v0, v5 offset:11
; GFX10-NEXT: ds_write_b8 v0, v6 offset:5 ; GFX10-NEXT: ds_write_b8 v0, v6 offset:5
; GFX10-NEXT: ds_write_b8 v0, v7 offset:7 ; GFX10-NEXT: ds_write_b8 v0, v7 offset:7
; GFX10-NEXT: ds_write_b8 v0, v15 offset:1 ; GFX10-NEXT: ds_write_b8 v0, v8 offset:1
; GFX10-NEXT: ds_write_b8 v0, v9 offset:3 ; GFX10-NEXT: ds_write_b8 v0, v9 offset:3
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
store <3 x i32> %x, <3 x i32> addrspace(3)* %out, align 1 store <3 x i32> %x, <3 x i32> addrspace(3)* %out, align 1
ret void ret void
} }
define amdgpu_kernel void @store_lds_v3i32_align2(<3 x i32> addrspace(3)* %out, <3 x i32> %x) { define amdgpu_kernel void @store_lds_v3i32_align2(<3 x i32> addrspace(3)* %out, <3 x i32> %x) {
; GFX9-LABEL: store_lds_v3i32_align2: ; GFX9-LABEL: store_lds_v3i32_align2:
▲ Show 20 Lines • Show All 258 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/strict_fadd.f64.ll

Show First 20 Lines • Show All 59 Lines • ▼ Show 20 Lines
; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5] ; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5]
; GCN-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7] ; GCN-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fadd_v2f64_fpexcept_strict: ; GFX10-LABEL: v_constained_fadd_v2f64_fpexcept_strict:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v9, v3 ; GFX10-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v8, v2 ; GFX10-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v11, v1
; GFX10-NEXT: v_mov_b32_e32 v10, v0
; GFX10-NEXT: v_add_f64 v[2:3], v[8:9], v[6:7]
; GFX10-NEXT: v_add_f64 v[0:1], v[10:11], v[4:5]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.strict") %val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
ret <2 x double> %val ret <2 x double> %val
} }
define <2 x double> @v_constained_fadd_v2f64_fpexcept_ignore(<2 x double> %x, <2 x double> %y) #0 { define <2 x double> @v_constained_fadd_v2f64_fpexcept_ignore(<2 x double> %x, <2 x double> %y) #0 {
; GCN-LABEL: v_constained_fadd_v2f64_fpexcept_ignore: ; GCN-LABEL: v_constained_fadd_v2f64_fpexcept_ignore:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5] ; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5]
; GCN-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7] ; GCN-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fadd_v2f64_fpexcept_ignore: ; GFX10-LABEL: v_constained_fadd_v2f64_fpexcept_ignore:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v9, v3 ; GFX10-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v8, v2 ; GFX10-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v11, v1
; GFX10-NEXT: v_mov_b32_e32 v10, v0
; GFX10-NEXT: v_add_f64 v[2:3], v[8:9], v[6:7]
; GFX10-NEXT: v_add_f64 v[0:1], v[10:11], v[4:5]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.ignore") %val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.ignore")
ret <2 x double> %val ret <2 x double> %val
} }
define <2 x double> @v_constained_fadd_v2f64_fpexcept_maytrap(<2 x double> %x, <2 x double> %y) #0 { define <2 x double> @v_constained_fadd_v2f64_fpexcept_maytrap(<2 x double> %x, <2 x double> %y) #0 {
; GCN-LABEL: v_constained_fadd_v2f64_fpexcept_maytrap: ; GCN-LABEL: v_constained_fadd_v2f64_fpexcept_maytrap:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5] ; GCN-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5]
; GCN-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7] ; GCN-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fadd_v2f64_fpexcept_maytrap: ; GFX10-LABEL: v_constained_fadd_v2f64_fpexcept_maytrap:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v9, v3 ; GFX10-NEXT: v_add_f64 v[0:1], v[0:1], v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v8, v2 ; GFX10-NEXT: v_add_f64 v[2:3], v[2:3], v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v11, v1
; GFX10-NEXT: v_mov_b32_e32 v10, v0
; GFX10-NEXT: v_add_f64 v[2:3], v[8:9], v[6:7]
; GFX10-NEXT: v_add_f64 v[0:1], v[10:11], v[4:5]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap") %val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap")
ret <2 x double> %val ret <2 x double> %val
} }
define <3 x double> @v_constained_fadd_v3f64_fpexcept_strict(<3 x double> %x, <3 x double> %y) #0 { define <3 x double> @v_constained_fadd_v3f64_fpexcept_strict(<3 x double> %x, <3 x double> %y) #0 {
; GCN-LABEL: v_constained_fadd_v3f64_fpexcept_strict: ; GCN-LABEL: v_constained_fadd_v3f64_fpexcept_strict:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
▲ Show 20 Lines • Show All 41 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/strict_fma.f16.ll

Show First 20 Lines • Show All 69 Lines • ▼ Show 20 Lines
; GFX8-NEXT: v_or_b32_sdwa v0, v0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD ; GFX8-NEXT: v_or_b32_sdwa v0, v0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX8-NEXT: v_fma_f16 v1, v1, v3, v5 ; GFX8-NEXT: v_fma_f16 v1, v1, v3, v5
; GFX8-NEXT: s_setpc_b64 s[30:31] ; GFX8-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fma_v3f16_fpexcept_strict: ; GFX10-LABEL: v_constained_fma_v3f16_fpexcept_strict:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v6, v5 ; GFX10-NEXT: v_fmac_f16_e32 v5, v1, v3
; GFX10-NEXT: v_pk_fma_f16 v0, v0, v2, v4 ; GFX10-NEXT: v_pk_fma_f16 v0, v0, v2, v4
; GFX10-NEXT: v_fmac_f16_e32 v6, v1, v3 ; GFX10-NEXT: v_mov_b32_e32 v1, v5
; GFX10-NEXT: v_mov_b32_e32 v1, v6
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <3 x half> @llvm.experimental.constrained.fma.v3f16(<3 x half> %x, <3 x half> %y, <3 x half> %z, metadata !"round.tonearest", metadata !"fpexcept.strict") %val = call <3 x half> @llvm.experimental.constrained.fma.v3f16(<3 x half> %x, <3 x half> %y, <3 x half> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
ret <3 x half> %val ret <3 x half> %val
} }
define <4 x half> @v_constained_fma_v4f16_fpexcept_strict(<4 x half> %x, <4 x half> %y, <4 x half> %z) #0 { define <4 x half> @v_constained_fma_v4f16_fpexcept_strict(<4 x half> %x, <4 x half> %y, <4 x half> %z) #0 {
; GFX9-LABEL: v_constained_fma_v4f16_fpexcept_strict: ; GFX9-LABEL: v_constained_fma_v4f16_fpexcept_strict:
; GFX9: ; %bb.0: ; GFX9: ; %bb.0:
Show All 33 Lines
; GFX8-NEXT: v_lshlrev_b32_e32 v2, 16, v6 ; GFX8-NEXT: v_lshlrev_b32_e32 v2, 16, v6
; GFX8-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD ; GFX8-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX8-NEXT: s_setpc_b64 s[30:31] ; GFX8-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fma_v4f16_fpexcept_strict: ; GFX10-LABEL: v_constained_fma_v4f16_fpexcept_strict:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v14, v5 ; GFX10-NEXT: v_lshrrev_b32_e32 v6, 16, v5
; GFX10-NEXT: v_mov_b32_e32 v15, v4 ; GFX10-NEXT: v_lshrrev_b32_e32 v9, 16, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 16, v3 ; GFX10-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v8, 16, v1 ; GFX10-NEXT: v_lshrrev_b32_e32 v8, 16, v1
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 16, v2 ; GFX10-NEXT: v_lshrrev_b32_e32 v10, 16, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 16, v14
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 16, v15
; GFX10-NEXT: v_lshrrev_b32_e32 v11, 16, v0 ; GFX10-NEXT: v_lshrrev_b32_e32 v11, 16, v0
; GFX10-NEXT: v_fmac_f16_e32 v15, v0, v2 ; GFX10-NEXT: v_fmac_f16_e32 v4, v0, v2
; GFX10-NEXT: v_mov_b32_e32 v0, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v0, 0xffff
; GFX10-NEXT: v_fmac_f16_e32 v14, v1, v3 ; GFX10-NEXT: v_fmac_f16_e32 v5, v1, v3
; GFX10-NEXT: v_fmac_f16_e32 v5, v8, v7 ; GFX10-NEXT: v_fmac_f16_e32 v6, v8, v7
; GFX10-NEXT: v_fmac_f16_e32 v4, v11, v10 ; GFX10-NEXT: v_fmac_f16_e32 v9, v11, v10
; GFX10-NEXT: v_and_b32_e32 v1, v0, v15 ; GFX10-NEXT: v_and_b32_e32 v1, v0, v4
; GFX10-NEXT: v_and_b32_e32 v2, v0, v14 ; GFX10-NEXT: v_and_b32_e32 v2, v0, v5
; GFX10-NEXT: v_lshl_or_b32 v0, v4, 16, v1 ; GFX10-NEXT: v_lshl_or_b32 v0, v9, 16, v1
; GFX10-NEXT: v_lshl_or_b32 v1, v5, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v1, v6, 16, v2
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <4 x half> @llvm.experimental.constrained.fma.v4f16(<4 x half> %x, <4 x half> %y, <4 x half> %z, metadata !"round.tonearest", metadata !"fpexcept.strict") %val = call <4 x half> @llvm.experimental.constrained.fma.v4f16(<4 x half> %x, <4 x half> %y, <4 x half> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
ret <4 x half> %val ret <4 x half> %val
} }
define half @v_constained_fma_f16_fpexcept_strict_fneg(half %x, half %y, half %z) #0 { define half @v_constained_fma_f16_fpexcept_strict_fneg(half %x, half %y, half %z) #0 {
; GCN-LABEL: v_constained_fma_f16_fpexcept_strict_fneg: ; GCN-LABEL: v_constained_fma_f16_fpexcept_strict_fneg:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
▲ Show 20 Lines • Show All 94 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/strict_fma.f64.ll

Show All 25 Lines
; GCN-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9] ; GCN-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9]
; GCN-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11] ; GCN-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fma_v2f64_fpexcept_strict: ; GFX10-LABEL: v_constained_fma_v2f64_fpexcept_strict:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v13, v3 ; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9]
; GFX10-NEXT: v_mov_b32_e32 v12, v2 ; GFX10-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11]
; GFX10-NEXT: v_mov_b32_e32 v15, v1
; GFX10-NEXT: v_mov_b32_e32 v14, v0
; GFX10-NEXT: v_fma_f64 v[2:3], v[12:13], v[6:7], v[10:11]
; GFX10-NEXT: v_fma_f64 v[0:1], v[14:15], v[4:5], v[8:9]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double> %x, <2 x double> %y, <2 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict") %val = call <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double> %x, <2 x double> %y, <2 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
ret <2 x double> %val ret <2 x double> %val
} }
define <3 x double> @v_constained_fma_v3f64_fpexcept_strict(<3 x double> %x, <3 x double> %y, <3 x double> %z) #0 { define <3 x double> @v_constained_fma_v3f64_fpexcept_strict(<3 x double> %x, <3 x double> %y, <3 x double> %z) #0 {
; GCN-LABEL: v_constained_fma_v3f64_fpexcept_strict: ; GCN-LABEL: v_constained_fma_v3f64_fpexcept_strict:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
Show All 24 Lines
; GCN-NEXT: v_fma_f64 v[4:5], v[4:5], v[12:13], v[20:21] ; GCN-NEXT: v_fma_f64 v[4:5], v[4:5], v[12:13], v[20:21]
; GCN-NEXT: v_fma_f64 v[6:7], v[6:7], v[14:15], v[22:23] ; GCN-NEXT: v_fma_f64 v[6:7], v[6:7], v[14:15], v[22:23]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fma_v4f64_fpexcept_strict: ; GFX10-LABEL: v_constained_fma_v4f64_fpexcept_strict:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v29, v7 ; GFX10-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[16:17]
; GFX10-NEXT: v_mov_b32_e32 v28, v6 ; GFX10-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[18:19]
; GFX10-NEXT: v_mov_b32_e32 v31, v5 ; GFX10-NEXT: v_fma_f64 v[4:5], v[4:5], v[12:13], v[20:21]
; GFX10-NEXT: v_mov_b32_e32 v30, v4 ; GFX10-NEXT: v_fma_f64 v[6:7], v[6:7], v[14:15], v[22:23]
; GFX10-NEXT: v_mov_b32_e32 v25, v3
; GFX10-NEXT: v_mov_b32_e32 v24, v2
; GFX10-NEXT: v_mov_b32_e32 v27, v1
; GFX10-NEXT: v_mov_b32_e32 v26, v0
; GFX10-NEXT: v_fma_f64 v[4:5], v[30:31], v[12:13], v[20:21]
; GFX10-NEXT: v_fma_f64 v[6:7], v[28:29], v[14:15], v[22:23]
; GFX10-NEXT: v_fma_f64 v[2:3], v[24:25], v[10:11], v[18:19]
; GFX10-NEXT: v_fma_f64 v[0:1], v[26:27], v[8:9], v[16:17]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <4 x double> @llvm.experimental.constrained.fma.v4f64(<4 x double> %x, <4 x double> %y, <4 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict") %val = call <4 x double> @llvm.experimental.constrained.fma.v4f64(<4 x double> %x, <4 x double> %y, <4 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
ret <4 x double> %val ret <4 x double> %val
} }
define double @v_constained_fma_f64_fpexcept_strict_fneg(double %x, double %y, double %z) #0 { define double @v_constained_fma_f64_fpexcept_strict_fneg(double %x, double %y, double %z) #0 {
; GCN-LABEL: v_constained_fma_f64_fpexcept_strict_fneg: ; GCN-LABEL: v_constained_fma_f64_fpexcept_strict_fneg:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
▲ Show 20 Lines • Show All 57 Lines • ▼ Show 20 Lines
; GCN-NEXT: v_fma_f64 v[0:1], -v[0:1], -v[4:5], v[8:9] ; GCN-NEXT: v_fma_f64 v[0:1], -v[0:1], -v[4:5], v[8:9]
; GCN-NEXT: v_fma_f64 v[2:3], -v[2:3], -v[6:7], v[10:11] ; GCN-NEXT: v_fma_f64 v[2:3], -v[2:3], -v[6:7], v[10:11]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fma_v2f64_fpexcept_strict_fneg_fneg: ; GFX10-LABEL: v_constained_fma_v2f64_fpexcept_strict_fneg_fneg:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v13, v3 ; GFX10-NEXT: v_fma_f64 v[0:1], -v[0:1], -v[4:5], v[8:9]
; GFX10-NEXT: v_mov_b32_e32 v12, v2 ; GFX10-NEXT: v_fma_f64 v[2:3], -v[2:3], -v[6:7], v[10:11]
; GFX10-NEXT: v_mov_b32_e32 v15, v1
; GFX10-NEXT: v_mov_b32_e32 v14, v0
; GFX10-NEXT: v_fma_f64 v[2:3], -v[12:13], -v[6:7], v[10:11]
; GFX10-NEXT: v_fma_f64 v[0:1], -v[14:15], -v[4:5], v[8:9]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%neg.x = fneg <2 x double> %x %neg.x = fneg <2 x double> %x
%neg.y = fneg <2 x double> %y %neg.y = fneg <2 x double> %y
%val = call <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double> %neg.x, <2 x double> %neg.y, <2 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict") %val = call <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double> %neg.x, <2 x double> %neg.y, <2 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
ret <2 x double> %val ret <2 x double> %val
} }
declare double @llvm.fabs.f64(double) #1 declare double @llvm.fabs.f64(double) #1
declare double @llvm.experimental.constrained.fma.f64(double, double, double, metadata, metadata) #1 declare double @llvm.experimental.constrained.fma.f64(double, double, double, metadata, metadata) #1
declare <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double>, <2 x double>, <2 x double>, metadata, metadata) #1 declare <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double>, <2 x double>, <2 x double>, metadata, metadata) #1
declare <3 x double> @llvm.experimental.constrained.fma.v3f64(<3 x double>, <3 x double>, <3 x double>, metadata, metadata) #1 declare <3 x double> @llvm.experimental.constrained.fma.v3f64(<3 x double>, <3 x double>, <3 x double>, metadata, metadata) #1
declare <4 x double> @llvm.experimental.constrained.fma.v4f64(<4 x double>, <4 x double>, <4 x double>, metadata, metadata) #1 declare <4 x double> @llvm.experimental.constrained.fma.v4f64(<4 x double>, <4 x double>, <4 x double>, metadata, metadata) #1
attributes #0 = { strictfp } attributes #0 = { strictfp }
attributes #1 = { inaccessiblememonly nounwind willreturn } attributes #1 = { inaccessiblememonly nounwind willreturn }

llvm/test/CodeGen/AMDGPU/strict_fmul.f64.ll

Show First 20 Lines • Show All 59 Lines • ▼ Show 20 Lines
; GCN-NEXT: v_mul_f64 v[0:1], v[0:1], v[4:5] ; GCN-NEXT: v_mul_f64 v[0:1], v[0:1], v[4:5]
; GCN-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7] ; GCN-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fmul_v2f64_fpexcept_strict: ; GFX10-LABEL: v_constained_fmul_v2f64_fpexcept_strict:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v9, v3 ; GFX10-NEXT: v_mul_f64 v[0:1], v[0:1], v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v8, v2 ; GFX10-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v11, v1
; GFX10-NEXT: v_mov_b32_e32 v10, v0
; GFX10-NEXT: v_mul_f64 v[2:3], v[8:9], v[6:7]
; GFX10-NEXT: v_mul_f64 v[0:1], v[10:11], v[4:5]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.strict") %val = call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
ret <2 x double> %val ret <2 x double> %val
} }
define <2 x double> @v_constained_fmul_v2f64_fpexcept_ignore(<2 x double> %x, <2 x double> %y) #0 { define <2 x double> @v_constained_fmul_v2f64_fpexcept_ignore(<2 x double> %x, <2 x double> %y) #0 {
; GCN-LABEL: v_constained_fmul_v2f64_fpexcept_ignore: ; GCN-LABEL: v_constained_fmul_v2f64_fpexcept_ignore:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_mul_f64 v[0:1], v[0:1], v[4:5] ; GCN-NEXT: v_mul_f64 v[0:1], v[0:1], v[4:5]
; GCN-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7] ; GCN-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fmul_v2f64_fpexcept_ignore: ; GFX10-LABEL: v_constained_fmul_v2f64_fpexcept_ignore:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v9, v3 ; GFX10-NEXT: v_mul_f64 v[0:1], v[0:1], v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v8, v2 ; GFX10-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v11, v1
; GFX10-NEXT: v_mov_b32_e32 v10, v0
; GFX10-NEXT: v_mul_f64 v[2:3], v[8:9], v[6:7]
; GFX10-NEXT: v_mul_f64 v[0:1], v[10:11], v[4:5]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.ignore") %val = call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.ignore")
ret <2 x double> %val ret <2 x double> %val
} }
define <2 x double> @v_constained_fmul_v2f64_fpexcept_maytrap(<2 x double> %x, <2 x double> %y) #0 { define <2 x double> @v_constained_fmul_v2f64_fpexcept_maytrap(<2 x double> %x, <2 x double> %y) #0 {
; GCN-LABEL: v_constained_fmul_v2f64_fpexcept_maytrap: ; GCN-LABEL: v_constained_fmul_v2f64_fpexcept_maytrap:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_mul_f64 v[0:1], v[0:1], v[4:5] ; GCN-NEXT: v_mul_f64 v[0:1], v[0:1], v[4:5]
; GCN-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7] ; GCN-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fmul_v2f64_fpexcept_maytrap: ; GFX10-LABEL: v_constained_fmul_v2f64_fpexcept_maytrap:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v9, v3 ; GFX10-NEXT: v_mul_f64 v[0:1], v[0:1], v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v8, v2 ; GFX10-NEXT: v_mul_f64 v[2:3], v[2:3], v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v11, v1
; GFX10-NEXT: v_mov_b32_e32 v10, v0
; GFX10-NEXT: v_mul_f64 v[2:3], v[8:9], v[6:7]
; GFX10-NEXT: v_mul_f64 v[0:1], v[10:11], v[4:5]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap") %val = call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap")
ret <2 x double> %val ret <2 x double> %val
} }
define <3 x double> @v_constained_fmul_v3f64_fpexcept_strict(<3 x double> %x, <3 x double> %y) #0 { define <3 x double> @v_constained_fmul_v3f64_fpexcept_strict(<3 x double> %x, <3 x double> %y) #0 {
; GCN-LABEL: v_constained_fmul_v3f64_fpexcept_strict: ; GCN-LABEL: v_constained_fmul_v3f64_fpexcept_strict:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
▲ Show 20 Lines • Show All 41 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/strict_fsub.f64.ll

Show First 20 Lines • Show All 59 Lines • ▼ Show 20 Lines
; GCN-NEXT: v_add_f64 v[0:1], v[0:1], -v[4:5] ; GCN-NEXT: v_add_f64 v[0:1], v[0:1], -v[4:5]
; GCN-NEXT: v_add_f64 v[2:3], v[2:3], -v[6:7] ; GCN-NEXT: v_add_f64 v[2:3], v[2:3], -v[6:7]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fsub_v2f64_fpexcept_strict: ; GFX10-LABEL: v_constained_fsub_v2f64_fpexcept_strict:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v9, v3 ; GFX10-NEXT: v_add_f64 v[0:1], v[0:1], -v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v8, v2 ; GFX10-NEXT: v_add_f64 v[2:3], v[2:3], -v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v11, v1
; GFX10-NEXT: v_mov_b32_e32 v10, v0
; GFX10-NEXT: v_add_f64 v[2:3], v[8:9], -v[6:7]
; GFX10-NEXT: v_add_f64 v[0:1], v[10:11], -v[4:5]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <2 x double> @llvm.experimental.constrained.fsub.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.strict") %val = call <2 x double> @llvm.experimental.constrained.fsub.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
ret <2 x double> %val ret <2 x double> %val
} }
define <2 x double> @v_constained_fsub_v2f64_fpexcept_ignore(<2 x double> %x, <2 x double> %y) #0 { define <2 x double> @v_constained_fsub_v2f64_fpexcept_ignore(<2 x double> %x, <2 x double> %y) #0 {
; GCN-LABEL: v_constained_fsub_v2f64_fpexcept_ignore: ; GCN-LABEL: v_constained_fsub_v2f64_fpexcept_ignore:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_add_f64 v[0:1], v[0:1], -v[4:5] ; GCN-NEXT: v_add_f64 v[0:1], v[0:1], -v[4:5]
; GCN-NEXT: v_add_f64 v[2:3], v[2:3], -v[6:7] ; GCN-NEXT: v_add_f64 v[2:3], v[2:3], -v[6:7]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fsub_v2f64_fpexcept_ignore: ; GFX10-LABEL: v_constained_fsub_v2f64_fpexcept_ignore:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v9, v3 ; GFX10-NEXT: v_add_f64 v[0:1], v[0:1], -v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v8, v2 ; GFX10-NEXT: v_add_f64 v[2:3], v[2:3], -v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v11, v1
; GFX10-NEXT: v_mov_b32_e32 v10, v0
; GFX10-NEXT: v_add_f64 v[2:3], v[8:9], -v[6:7]
; GFX10-NEXT: v_add_f64 v[0:1], v[10:11], -v[4:5]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <2 x double> @llvm.experimental.constrained.fsub.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.ignore") %val = call <2 x double> @llvm.experimental.constrained.fsub.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.ignore")
ret <2 x double> %val ret <2 x double> %val
} }
define <2 x double> @v_constained_fsub_v2f64_fpexcept_maytrap(<2 x double> %x, <2 x double> %y) #0 { define <2 x double> @v_constained_fsub_v2f64_fpexcept_maytrap(<2 x double> %x, <2 x double> %y) #0 {
; GCN-LABEL: v_constained_fsub_v2f64_fpexcept_maytrap: ; GCN-LABEL: v_constained_fsub_v2f64_fpexcept_maytrap:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_add_f64 v[0:1], v[0:1], -v[4:5] ; GCN-NEXT: v_add_f64 v[0:1], v[0:1], -v[4:5]
; GCN-NEXT: v_add_f64 v[2:3], v[2:3], -v[6:7] ; GCN-NEXT: v_add_f64 v[2:3], v[2:3], -v[6:7]
; GCN-NEXT: s_setpc_b64 s[30:31] ; GCN-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: v_constained_fsub_v2f64_fpexcept_maytrap: ; GFX10-LABEL: v_constained_fsub_v2f64_fpexcept_maytrap:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v9, v3 ; GFX10-NEXT: v_add_f64 v[0:1], v[0:1], -v[4:5]
; GFX10-NEXT: v_mov_b32_e32 v8, v2 ; GFX10-NEXT: v_add_f64 v[2:3], v[2:3], -v[6:7]
; GFX10-NEXT: v_mov_b32_e32 v11, v1
; GFX10-NEXT: v_mov_b32_e32 v10, v0
; GFX10-NEXT: v_add_f64 v[2:3], v[8:9], -v[6:7]
; GFX10-NEXT: v_add_f64 v[0:1], v[10:11], -v[4:5]
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val = call <2 x double> @llvm.experimental.constrained.fsub.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap") %val = call <2 x double> @llvm.experimental.constrained.fsub.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap")
ret <2 x double> %val ret <2 x double> %val
} }
define <3 x double> @v_constained_fsub_v3f64_fpexcept_strict(<3 x double> %x, <3 x double> %y) #0 { define <3 x double> @v_constained_fsub_v3f64_fpexcept_strict(<3 x double> %x, <3 x double> %y) #0 {
; GCN-LABEL: v_constained_fsub_v3f64_fpexcept_strict: ; GCN-LABEL: v_constained_fsub_v3f64_fpexcept_strict:
; GCN: ; %bb.0: ; GCN: ; %bb.0:
▲ Show 20 Lines • Show All 41 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/vector_shuffle.packed.ll

Show First 20 Lines • Show All 740 Lines • ▼ Show 20 Lines
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: shuffle_v4f16_3456: ; GFX10-LABEL: shuffle_v4f16_3456:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: global_load_dword v6, v[0:1], off offset:4 ; GFX10-NEXT: global_load_dword v6, v[0:1], off offset:4
; GFX10-NEXT: global_load_dwordx2 v[4:5], v[2:3], off ; GFX10-NEXT: global_load_dwordx2 v[4:5], v[2:3], off
; GFX10-NEXT: v_mov_b32_e32 v3, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v0, 0xffff
; GFX10-NEXT: s_waitcnt vmcnt(1) ; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_and_b32_sdwa v1, v3, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; GFX10-NEXT: v_and_b32_sdwa v1, v0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_and_b32_sdwa v2, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; GFX10-NEXT: v_and_b32_sdwa v2, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX10-NEXT: v_lshl_or_b32 v0, v4, 16, v1 ; GFX10-NEXT: v_lshl_or_b32 v0, v4, 16, v1
; GFX10-NEXT: v_lshl_or_b32 v1, v5, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v1, v5, 16, v2
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0 %val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1 %val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
%shuffle = shufflevector <4 x half> %val0, <4 x half> %val1, <4 x i32> <i32 3, i32 4, i32 5, i32 6> %shuffle = shufflevector <4 x half> %val0, <4 x half> %val1, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
ret <4 x half> %shuffle ret <4 x half> %shuffle
} }
Show All 11 Lines
; GFX9-NEXT: v_lshl_or_b32 v1, v4, 16, v1 ; GFX9-NEXT: v_lshl_or_b32 v1, v4, 16, v1
; GFX9-NEXT: v_lshl_or_b32 v0, v5, 16, v0 ; GFX9-NEXT: v_lshl_or_b32 v0, v5, 16, v0
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: shuffle_v4f16_5634: ; GFX10-LABEL: shuffle_v4f16_5634:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: global_load_dwordx2 v[9:10], v[2:3], off ; GFX10-NEXT: global_load_dwordx2 v[4:5], v[2:3], off
; GFX10-NEXT: global_load_dword v6, v[0:1], off offset:4 ; GFX10-NEXT: global_load_dword v6, v[0:1], off offset:4
; GFX10-NEXT: v_mov_b32_e32 v3, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v0, 0xffff
; GFX10-NEXT: s_waitcnt vmcnt(1) ; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_and_b32_sdwa v1, v3, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; GFX10-NEXT: v_and_b32_sdwa v1, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_and_b32_sdwa v2, v3, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; GFX10-NEXT: v_and_b32_sdwa v2, v0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX10-NEXT: v_lshl_or_b32 v0, v10, 16, v1 ; GFX10-NEXT: v_lshl_or_b32 v0, v5, 16, v1
; GFX10-NEXT: v_lshl_or_b32 v1, v9, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v1, v4, 16, v2
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0 %val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1 %val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
%shuffle = shufflevector <4 x half> %val0, <4 x half> %val1, <4 x i32> <i32 5, i32 6, i32 3, i32 4> %shuffle = shufflevector <4 x half> %val0, <4 x half> %val1, <4 x i32> <i32 5, i32 6, i32 3, i32 4>
ret <4 x half> %shuffle ret <4 x half> %shuffle
} }
define <4 x half> @shuffle_v4f16_5734(<4 x half> addrspace(1)* %arg0, <4 x half> addrspace(1)* %arg1) { define <4 x half> @shuffle_v4f16_5734(<4 x half> addrspace(1)* %arg0, <4 x half> addrspace(1)* %arg1) {
Show All 13 Lines
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: shuffle_v4f16_5734: ; GFX10-LABEL: shuffle_v4f16_5734:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: global_load_dwordx2 v[4:5], v[2:3], off ; GFX10-NEXT: global_load_dwordx2 v[4:5], v[2:3], off
; GFX10-NEXT: global_load_dword v6, v[0:1], off offset:4 ; GFX10-NEXT: global_load_dword v6, v[0:1], off offset:4
; GFX10-NEXT: v_mov_b32_e32 v3, 0xffff ; GFX10-NEXT: v_mov_b32_e32 v0, 0xffff
; GFX10-NEXT: s_waitcnt vmcnt(1) ; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_and_b32_sdwa v2, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 16, v5 ; GFX10-NEXT: v_lshrrev_b32_e32 v1, 16, v5
; GFX10-NEXT: v_and_b32_sdwa v2, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_and_b32_sdwa v3, v3, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; GFX10-NEXT: v_and_b32_sdwa v3, v0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v2 ; GFX10-NEXT: v_lshl_or_b32 v0, v1, 16, v2
; GFX10-NEXT: v_lshl_or_b32 v1, v4, 16, v3 ; GFX10-NEXT: v_lshl_or_b32 v1, v4, 16, v3
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0 %val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1 %val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
%shuffle = shufflevector <4 x half> %val0, <4 x half> %val1, <4 x i32> <i32 5, i32 7, i32 3, i32 4> %shuffle = shufflevector <4 x half> %val0, <4 x half> %val1, <4 x i32> <i32 5, i32 7, i32 3, i32 4>
ret <4 x half> %shuffle ret <4 x half> %shuffle
} }
▲ Show 20 Lines • Show All 481 Lines • ▼ Show 20 Lines
; GFX10-LABEL: fma_shuffle: ; GFX10-LABEL: fma_shuffle:
; GFX10: ; %bb.0: ; %entry ; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_clause 0x1 ; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0 ; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
; GFX10-NEXT: s_load_dwordx4 s[8:11], s[4:5], 0x10 ; GFX10-NEXT: s_load_dwordx4 s[8:11], s[4:5], 0x10
; GFX10-NEXT: v_lshlrev_b32_e32 v6, 3, v0 ; GFX10-NEXT: v_lshlrev_b32_e32 v6, 3, v0
; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_clause 0x2 ; GFX10-NEXT: s_clause 0x2
; GFX10-NEXT: global_load_dwordx2 v[7:8], v6, s[0:1] ; GFX10-NEXT: global_load_dwordx2 v[0:1], v6, s[0:1]
; GFX10-NEXT: global_load_dwordx2 v[2:3], v6, s[2:3] ; GFX10-NEXT: global_load_dwordx2 v[2:3], v6, s[2:3]
; GFX10-NEXT: global_load_dwordx2 v[4:5], v6, s[8:9] ; GFX10-NEXT: global_load_dwordx2 v[4:5], v6, s[8:9]
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_pk_fma_f16 v4, v7, v2, v4 op_sel_hi:[0,1,1] ; GFX10-NEXT: v_pk_fma_f16 v4, v0, v2, v4 op_sel_hi:[0,1,1]
; GFX10-NEXT: v_pk_fma_f16 v2, v8, v2, v5 op_sel_hi:[0,1,1] ; GFX10-NEXT: v_pk_fma_f16 v2, v1, v2, v5 op_sel_hi:[0,1,1]
; GFX10-NEXT: v_pk_fma_f16 v0, v7, v3, v4 op_sel:[1,0,0] ; GFX10-NEXT: v_pk_fma_f16 v0, v0, v3, v4 op_sel:[1,0,0]
; GFX10-NEXT: v_pk_fma_f16 v1, v8, v3, v2 op_sel:[1,0,0] ; GFX10-NEXT: v_pk_fma_f16 v1, v1, v3, v2 op_sel:[1,0,0]
; GFX10-NEXT: global_store_dwordx2 v6, v[0:1], s[8:9] ; GFX10-NEXT: global_store_dwordx2 v6, v[0:1], s[8:9]
; GFX10-NEXT: s_endpgm ; GFX10-NEXT: s_endpgm
entry: entry:
%tmp1 = tail call i32 @llvm.amdgcn.workitem.id.x() %tmp1 = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp12 = zext i32 %tmp1 to i64 %tmp12 = zext i32 %tmp1 to i64
%arrayidx = getelementptr inbounds <4 x half>, <4 x half> addrspace(1)* %A, i64 %tmp12 %arrayidx = getelementptr inbounds <4 x half>, <4 x half> addrspace(1)* %A, i64 %tmp12
%tmp14 = load <4 x half>, <4 x half> addrspace(1)* %arrayidx, align 8 %tmp14 = load <4 x half>, <4 x half> addrspace(1)* %arrayidx, align 8
%arrayidx1 = getelementptr inbounds <4 x half>, <4 x half> addrspace(1)* %B, i64 %tmp12 %arrayidx1 = getelementptr inbounds <4 x half>, <4 x half> addrspace(1)* %B, i64 %tmp12
Show All 37 Lines
; GFX9-NEXT: v_lshl_or_b32 v1, v6, 16, v2 ; GFX9-NEXT: v_lshl_or_b32 v1, v6, 16, v2
; GFX9-NEXT: s_setpc_b64 s[30:31] ; GFX9-NEXT: s_setpc_b64 s[30:31]
; ;
; GFX10-LABEL: shuffle_v4f16_0456: ; GFX10-LABEL: shuffle_v4f16_0456:
; GFX10: ; %bb.0: ; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 ; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: global_load_dwordx2 v[4:5], v[0:1], off ; GFX10-NEXT: global_load_dwordx2 v[4:5], v[0:1], off
; GFX10-NEXT: global_load_dwordx2 v[6:7], v[2:3], off
; GFX10-NEXT: v_mov_b32_e32 v3, 0xffff
; GFX10-NEXT: s_waitcnt vmcnt(1)
; GFX10-NEXT: v_and_b32_e32 v1, v3, v4
; GFX10-NEXT: s_waitcnt vmcnt(0) ; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_and_b32_sdwa v2, v3, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; GFX10-NEXT: global_load_dwordx2 v[5:6], v[2:3], off
; GFX10-NEXT: v_lshl_or_b32 v0, v6, 16, v1 ; GFX10-NEXT: ; kill: killed $vgpr0 killed $vgpr1
; GFX10-NEXT: v_lshl_or_b32 v1, v7, 16, v2 ; GFX10-NEXT: v_mov_b32_e32 v0, 0xffff
; GFX10-NEXT: ; kill: killed $vgpr2 killed $vgpr3
; GFX10-NEXT: v_and_b32_e32 v1, v0, v4
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_and_b32_sdwa v2, v0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX10-NEXT: v_lshl_or_b32 v0, v5, 16, v1
; GFX10-NEXT: v_lshl_or_b32 v1, v6, 16, v2
; GFX10-NEXT: s_setpc_b64 s[30:31] ; GFX10-NEXT: s_setpc_b64 s[30:31]
%val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0 %val0 = load <4 x half>, <4 x half> addrspace(1)* %arg0
%val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1 %val1 = load <4 x half>, <4 x half> addrspace(1)* %arg1
%shuffle = shufflevector <4 x half> %val0, <4 x half> %val1, <4 x i32> <i32 0, i32 4, i32 5, i32 6> %shuffle = shufflevector <4 x half> %val0, <4 x half> %val1, <4 x i32> <i32 0, i32 4, i32 5, i32 6>
ret <4 x half> %shuffle ret <4 x half> %shuffle
} }
define amdgpu_kernel void @shuffle_scalar_load_v8i32_0123(<8 x i32> addrspace(4)* %in, <4 x i32> addrspace(1)* %out) { define amdgpu_kernel void @shuffle_scalar_load_v8i32_0123(<8 x i32> addrspace(4)* %in, <4 x i32> addrspace(1)* %out) {
Show All 37 Lines
llvm/test/CodeGen/AMDGPU/memory_clause.ll