[AMDGPU] Add SIPreEmitPeephole pass.

[cdevadas]

Summary

This pass will handle all the optimization
opportunities found just before code emission.
Presently it includes the handling of vcc branch
optimization that was handled earlier in SIInsertSkips.

[arsenm]

LGTM

llvm/lib/Target/AMDGPU/SIPreEmitPeephole.cpp

163

Extra blank line

llvm/lib/Target/AMDGPU/AMDGPU.h

 extern char &SIWholeQuadModeID;
 
 void initializeSILowerControlFlowPass(PassRegistry &);
 extern char &SILowerControlFlowID;
 
 void initializeSIRemoveShortExecBranchesPass(PassRegistry &);
 extern char &SIRemoveShortExecBranchesID;
 
+void initializeSIPreEmitPeepholePass(PassRegistry &);
+extern char &SIPreEmitPeepholeID;
+
 void initializeSIInsertSkipsPass(PassRegistry &);
 extern char &SIInsertSkipsPassID;
 
 void initializeSIOptimizeExecMaskingPass(PassRegistry &);
 extern char &SIOptimizeExecMaskingID;
 
 void initializeSIPreAllocateWWMRegsPass(PassRegistry &);
 extern char &SIPreAllocateWWMRegsID;

llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeAMDGPUTarget() {
   initializeAMDGPURewriteOutArgumentsPass(*PR);
   initializeAMDGPUUnifyMetadataPass(*PR);
   initializeSIAnnotateControlFlowPass(*PR);
   initializeSIInsertWaitcntsPass(*PR);
   initializeSIModeRegisterPass(*PR);
   initializeSIWholeQuadModePass(*PR);
   initializeSILowerControlFlowPass(*PR);
   initializeSIRemoveShortExecBranchesPass(*PR);
+  initializeSIPreEmitPeepholePass(*PR);
   initializeSIInsertSkipsPass(*PR);
   initializeSIMemoryLegalizerPass(*PR);
   initializeSIOptimizeExecMaskingPass(*PR);
   initializeSIPreAllocateWWMRegsPass(*PR);
   initializeSIFormMemoryClausesPass(*PR);
   initializeSIPostRABundlerPass(*PR);
   initializeAMDGPUUnifyDivergentExitNodesPass(*PR);
   initializeAMDGPUAAWrapperPassPass(*PR);
void GCNPassConfig::addPreEmitPass() {
   // Here we add a stand-alone hazard recognizer pass which can handle all
   // cases.
   //
   // FIXME: This stand-alone pass will emit indiv. S_NOP 0, as needed. It would
   // be better for it to emit S_NOP <N> when possible.
   addPass(&PostRAHazardRecognizerID);
 
   addPass(&SIRemoveShortExecBranchesID);
+  addPass(&SIPreEmitPeepholeID);
   addPass(&SIInsertSkipsPassID);
   addPass(&BranchRelaxationPassID);
 }
 
 TargetPassConfig *GCNTargetMachine::createPassConfig(PassManagerBase &PM) {
   return new GCNPassConfig(*this, PM);
 }
 

llvm/lib/Target/AMDGPU/CMakeLists.txt

add_llvm_target(AMDGPUCodeGen
   SILowerSGPRSpills.cpp
   SIMachineFunctionInfo.cpp
   SIMachineScheduler.cpp
   SIMemoryLegalizer.cpp
   SIOptimizeExecMasking.cpp
   SIOptimizeExecMaskingPreRA.cpp
   SIPeepholeSDWA.cpp
   SIPostRABundler.cpp
+  SIPreEmitPeephole.cpp
   SIRegisterInfo.cpp
   SIRemoveShortExecBranches.cpp
   SIShrinkInstructions.cpp
   SIWholeQuadMode.cpp
   GCNILPSched.cpp
   GCNRegBankReassign.cpp
   GCNNSAReassign.cpp
   GCNDPPCombine.cpp
   SIModeRegister.cpp
   )
 
 add_subdirectory(AsmParser)
 add_subdirectory(Disassembler)
 add_subdirectory(MCTargetDesc)
 add_subdirectory(TargetInfo)
 add_subdirectory(Utils)

llvm/lib/Target/AMDGPU/SIInsertSkips.cpp

private:
   bool dominatesAllReachable(MachineBasicBlock &MBB);
   void skipIfDead(MachineBasicBlock &MBB, MachineBasicBlock::iterator I,
                   DebugLoc DL);
 
   bool kill(MachineInstr &MI);
 
   bool skipMaskBranch(MachineInstr &MI, MachineBasicBlock &MBB);
 
-  bool optimizeVccBranch(MachineInstr &MI) const;
-
 public:
   static char ID;
 
   SIInsertSkips() : MachineFunctionPass(ID) {}
 
   bool runOnMachineFunction(MachineFunction &MF) override;
 
   StringRef getPassName() const override {
bool SIInsertSkips::skipMaskBranch(MachineInstr &MI,
   MachineBasicBlock::iterator InsPt = std::next(MI.getIterator());
 
   BuildMI(SrcMBB, InsPt, DL, TII->get(AMDGPU::S_CBRANCH_EXECZ))
     .addMBB(DestBB);
 
   return true;
 }
 
-bool SIInsertSkips::optimizeVccBranch(MachineInstr &MI) const {
-  // Match:
-  // sreg = -1
-  // vcc = S_AND_B64 exec, sreg
-  // S_CBRANCH_VCC[N]Z
-  // =>
-  // S_CBRANCH_EXEC[N]Z
-  bool Changed = false;
-  MachineBasicBlock &MBB = *MI.getParent();
-  const GCNSubtarget &ST = MBB.getParent()->getSubtarget<GCNSubtarget>();
-  const bool IsWave32 = ST.isWave32();
-  const unsigned CondReg = TRI->getVCC();
-  const unsigned ExecReg = IsWave32 ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
-  const unsigned And = IsWave32 ? AMDGPU::S_AND_B32 : AMDGPU::S_AND_B64;
-
-  MachineBasicBlock::reverse_iterator A = MI.getReverseIterator(),
-                                      E = MBB.rend();
-  bool ReadsCond = false;
-  unsigned Threshold = 5;
-  for (++A ; A != E ; ++A) {
-    if (!--Threshold)
-      return false;
-    if (A->modifiesRegister(ExecReg, TRI))
-      return false;
-    if (A->modifiesRegister(CondReg, TRI)) {
-      if (!A->definesRegister(CondReg, TRI) || A->getOpcode() != And)
-        return false;
-      break;
-    }
-    ReadsCond |= A->readsRegister(CondReg, TRI);
-  }
-  if (A == E)
-    return false;
-
-  MachineOperand &Op1 = A->getOperand(1);
-  MachineOperand &Op2 = A->getOperand(2);
-  if (Op1.getReg() != ExecReg && Op2.isReg() && Op2.getReg() == ExecReg) {
-    TII->commuteInstruction(*A);
-    Changed = true;
-  }
-  if (Op1.getReg() != ExecReg)
-    return Changed;
-  if (Op2.isImm() && Op2.getImm() != -1)
-    return Changed;
-
-  unsigned SReg = AMDGPU::NoRegister;
-  if (Op2.isReg()) {
-    SReg = Op2.getReg();
-    auto M = std::next(A);
-    bool ReadsSreg = false;
-    for ( ; M != E ; ++M) {
-      if (M->definesRegister(SReg, TRI))
-        break;
-      if (M->modifiesRegister(SReg, TRI))
-        return Changed;
-      ReadsSreg |= M->readsRegister(SReg, TRI);
-    }
-    if (M == E ||
-        !M->isMoveImmediate() ||
-        !M->getOperand(1).isImm() ||
-        M->getOperand(1).getImm() != -1)
-      return Changed;
-    // First if sreg is only used in and instruction fold the immediate
-    // into that and.
-    if (!ReadsSreg && Op2.isKill()) {
-      A->getOperand(2).ChangeToImmediate(-1);
-      M->eraseFromParent();
-    }
-  }
-
-  if (!ReadsCond && A->registerDefIsDead(AMDGPU::SCC) &&
-      MI.killsRegister(CondReg, TRI))
-    A->eraseFromParent();
-
-  bool IsVCCZ = MI.getOpcode() == AMDGPU::S_CBRANCH_VCCZ;
-  if (SReg == ExecReg) {
-    if (IsVCCZ) {
-      MI.eraseFromParent();
-      return true;
-    }
-    MI.setDesc(TII->get(AMDGPU::S_BRANCH));
-  } else {
-    MI.setDesc(TII->get(IsVCCZ ? AMDGPU::S_CBRANCH_EXECZ
-                               : AMDGPU::S_CBRANCH_EXECNZ));
-  }
-
-  MI.RemoveOperand(MI.findRegisterUseOperandIdx(CondReg, false /*Kill*/, TRI));
-  MI.addImplicitDefUseOperands(*MBB.getParent());
-
-  return true;
-}
-
 bool SIInsertSkips::runOnMachineFunction(MachineFunction &MF) {
   const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
   TII = ST.getInstrInfo();
   TRI = &TII->getRegisterInfo();
   MDT = &getAnalysis<MachineDominatorTree>();
   SkipThreshold = SkipThresholdFlag;
 
   MachineBasicBlock *EmptyMBBAtEnd = nullptr;
for (I = MBB.begin(); I != MBB.end(); I = Next) {
           BuildMI(MBB, &MI, MI.getDebugLoc(), TII->get(AMDGPU::S_BRANCH))
             .addMBB(EmptyMBBAtEnd);
           MI.eraseFromParent();
 
           MDT->getBase().insertEdge(&MBB, EmptyMBBAtEnd);
         }
         break;
 
-      case AMDGPU::S_CBRANCH_VCCZ:
-      case AMDGPU::S_CBRANCH_VCCNZ:
-        MadeChange |= optimizeVccBranch(MI);
-        break;
-
       default:
         break;
       }
     }
   }
 
   for (MachineInstr *Kill : KillInstrs) {
     skipIfDead(*Kill->getParent(), std::next(Kill->getIterator()),
                Kill->getDebugLoc());
     Kill->eraseFromParent();
   }
   KillInstrs.clear();
 
   return MadeChange;
 }

llvm/lib/Target/AMDGPU/SIPreEmitPeephole.cpp

+//===-- SIPreEmitPeephole.cpp ------------------------------------===//
+//
+// 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
+/// This pass performs the peephole optimizations before code emission.
+///
+//===----------------------------------------------------------------------===//
+
+#include "AMDGPU.h"
+#include "AMDGPUSubtarget.h"
+#include "MCTargetDesc/AMDGPUMCTargetDesc.h"
+#include "SIInstrInfo.h"
+#include "llvm/CodeGen/MachineFunctionPass.h"
+#include "llvm/Support/CommandLine.h"
+
+using namespace llvm;
+
+#define DEBUG_TYPE "si-pre-emit-peephole"
+
+namespace {
+
+class SIPreEmitPeephole : public MachineFunctionPass {
+private:
+  const SIInstrInfo *TII = nullptr;
+  const SIRegisterInfo *TRI = nullptr;
+
+  bool optimizeVccBranch(MachineInstr &MI) const;
+
+public:
+  static char ID;
+
+  SIPreEmitPeephole() : MachineFunctionPass(ID) {
+    initializeSIPreEmitPeepholePass(*PassRegistry::getPassRegistry());
+  }
+
+  bool runOnMachineFunction(MachineFunction &MF) override;
+};
+
+} // End anonymous namespace.
+
+INITIALIZE_PASS(SIPreEmitPeephole, DEBUG_TYPE,
+                "SI peephole optimizations", false, false)
+
+char SIPreEmitPeephole::ID = 0;
+
+char &llvm::SIPreEmitPeepholeID = SIPreEmitPeephole::ID;
+
+bool SIPreEmitPeephole::optimizeVccBranch(MachineInstr &MI) const {
+  // Match:
+  // sreg = -1
+  // vcc = S_AND_B64 exec, sreg
+  // S_CBRANCH_VCC[N]Z
+  // =>
+  // S_CBRANCH_EXEC[N]Z
+  // We end up with this pattern sometimes after basic block placement.
+  // It happens while combining a block which assigns -1 to a saved mask and
+  // another block which consumes that saved mask and then a branch.
+  bool Changed = false;
+  MachineBasicBlock &MBB = *MI.getParent();
+  const GCNSubtarget &ST = MBB.getParent()->getSubtarget<GCNSubtarget>();
+  const bool IsWave32 = ST.isWave32();
+  const unsigned CondReg = TRI->getVCC();
+  const unsigned ExecReg = IsWave32 ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+  const unsigned And = IsWave32 ? AMDGPU::S_AND_B32 : AMDGPU::S_AND_B64;
+
+  MachineBasicBlock::reverse_iterator A = MI.getReverseIterator(),
+                                      E = MBB.rend();
+  bool ReadsCond = false;
+  unsigned Threshold = 5;
+  for (++A; A != E; ++A) {
+    if (!--Threshold)
+      return false;
+    if (A->modifiesRegister(ExecReg, TRI))
+      return false;
+    if (A->modifiesRegister(CondReg, TRI)) {
+      if (!A->definesRegister(CondReg, TRI) || A->getOpcode() != And)
+        return false;
+      break;
+    }
+    ReadsCond |= A->readsRegister(CondReg, TRI);
+  }
+  if (A == E)
+    return false;
+
+  MachineOperand &Op1 = A->getOperand(1);
+  MachineOperand &Op2 = A->getOperand(2);
+  if (Op1.getReg() != ExecReg && Op2.isReg() && Op2.getReg() == ExecReg) {
+    TII->commuteInstruction(*A);
+    Changed = true;
+  }
+  if (Op1.getReg() != ExecReg)
+    return Changed;
+  if (Op2.isImm() && Op2.getImm() != -1)
+    return Changed;
+
+  Register SReg;
+  if (Op2.isReg()) {
+    SReg = Op2.getReg();
+    auto M = std::next(A);
+    bool ReadsSreg = false;
+    for (; M != E; ++M) {
+      if (M->definesRegister(SReg, TRI))
+        break;
+      if (M->modifiesRegister(SReg, TRI))
+        return Changed;
+      ReadsSreg |= M->readsRegister(SReg, TRI);
+    }
+    if (M == E || !M->isMoveImmediate() || !M->getOperand(1).isImm() ||
+        M->getOperand(1).getImm() != -1)
+      return Changed;
+    // First if sreg is only used in and instruction fold the immediate
+    // into that and.
+    if (!ReadsSreg && Op2.isKill()) {
+      A->getOperand(2).ChangeToImmediate(-1);
+      M->eraseFromParent();
+    }
+  }
+
+  if (!ReadsCond && A->registerDefIsDead(AMDGPU::SCC) &&
+      MI.killsRegister(CondReg, TRI))
+    A->eraseFromParent();
+
+  bool IsVCCZ = MI.getOpcode() == AMDGPU::S_CBRANCH_VCCZ;
+  if (SReg == ExecReg) {
+    if (IsVCCZ) {
+      MI.eraseFromParent();
+      return true;
+    }
+    MI.setDesc(TII->get(AMDGPU::S_BRANCH));
+  } else {
+    MI.setDesc(
+        TII->get(IsVCCZ ? AMDGPU::S_CBRANCH_EXECZ : AMDGPU::S_CBRANCH_EXECNZ));
+  }
+
+  MI.RemoveOperand(MI.findRegisterUseOperandIdx(CondReg, false /*Kill*/, TRI));
+  MI.addImplicitDefUseOperands(*MBB.getParent());
+
+  return true;
+}
+
+bool SIPreEmitPeephole::runOnMachineFunction(MachineFunction &MF) {
+  const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
+  TII = ST.getInstrInfo();
+  TRI = &TII->getRegisterInfo();
+  bool Changed = false;
+
+  for (MachineBasicBlock &MBB : MF) {
+    MachineBasicBlock::iterator MBBI = MBB.getFirstTerminator();
+    if (MBBI == MBB.end())
+      continue;
+
+    MachineInstr &MI = *MBBI;
+    switch (MI.getOpcode()) {
+    case AMDGPU::S_CBRANCH_VCCZ:
+    case AMDGPU::S_CBRANCH_VCCNZ:
+      Changed |= optimizeVccBranch(MI);
+      break;
+
+    default:
+      break;
+    }
+  }
+
+  return Changed;
+}

llvm/test/CodeGen/AMDGPU/insert-skip-from-vcc.mir

-# RUN: llc -march=amdgcn -mcpu=fiji -run-pass si-insert-skips -verify-machineinstrs -o - %s | FileCheck -check-prefix=GCN %s
-# RUN: llc -march=amdgcn -mcpu=gfx1010 -run-pass si-insert-skips -verify-machineinstrs -o - %s | FileCheck -check-prefix=W32 %s
+# RUN: llc -march=amdgcn -mcpu=fiji -run-pass si-pre-emit-peephole -verify-machineinstrs -o - %s | FileCheck -check-prefix=GCN %s
+# RUN: llc -march=amdgcn -mcpu=gfx1010 -run-pass si-pre-emit-peephole -verify-machineinstrs -o - %s | FileCheck -check-prefix=W32 %s
 
 ---
 # GCN-LABEL: name: and_execz_mov_vccz
 # GCN-NOT: S_MOV_
 # GCN-NOT: S_AND_
 # GCN: S_CBRANCH_EXECZ %bb.1, implicit $exec
 name:            and_execz_mov_vccz
 body:             |