This is an archive of the discontinued LLVM Phabricator instance.

Differential D34407

Switch scalarize global loads ON by default
ClosedPublic

Authored by alex-t on Jun 20 2017, 9:28 AM.

Details

Reviewers
rampitec
Commits
rGea7f08bee53c: [AMDGPU] Switch scalarize global loads ON by default
rL307097: [AMDGPU] Switch scalarize global loads ON by default
rL307026: [AMDGPU] Switch scalarize global loads ON by default
Summary

This change switches ON scalarization for global loads.
It finalizes series of changes that fix bugs in AMDGPUAnnotateUniformValues and DivergenceAnalysis.

Diff Detail

Repository
rL LLVM

Event Timeline

alex-t created this revision.Jun 20 2017, 9:28 AM
Herald added subscribers: tpr, nhaehnle, arsenm. · View Herald TranscriptJun 20 2017, 9:28 AM
alex-t added a reviewer: rampitec.Jun 20 2017, 9:29 AM
rampitec edited edge metadata.Jun 20 2017, 9:31 AM

Are there any lit tests affected?

alex-t updated this revision to Diff 104850.Jun 30 2017, 6:01 AM

There are most of the LIT tests affected.
The reason is that most of the AMDGPU tests are initially incorrect.
AMDGPU instruction selection select vector form of the instruction by default.
The reason was related with the SelectionDAG interface restriction. There was no legal way to pass divergence flag into ISel.
That's why any operation that operates on the scalar values is selected to vector form.
Since we're able to select scalar loads for uniform address, most of the tests need to be changed.
In fact there should be separate change that would split tests to vector and scalar parts.
Currently I changed 30 of 140 failed tests and switch scalarizer OFF for the rest of them.
Since SDNode flags was recently changed from byte to short and we have 5 unoccupied bits it become possible to pass divergence to ISel.
I'm going to submit the change that use SDNode flag for divergence information soon.
That's why I switched OFF scalarizer for most of the affected tests. I'm going to change them after we'll have full support for divergence.

Herald added a subscriber: qcolombet. · View Herald TranscriptJun 30 2017, 6:01 AM
rampitec accepted this revision.Jun 30 2017, 11:47 AM
rampitec added a subscriber: llvm-commits.

LGTM

This revision is now accepted and ready to land.Jun 30 2017, 11:48 AM
Closed by commit rL307026: [AMDGPU] Switch scalarize global loads ON by default (authored by alex-t). · Explain WhyJul 3 2017, 7:55 AM
This revision was automatically updated to reflect the committed changes.
chapuni reopened this revision.Jul 3 2017, 7:20 PM
chapuni added a subscriber: chapuni.

Excuse me, reverted in rL307054.

Seems a new test, llvm/test/CodeGen/AMDGPU/alignbit-pat.ll wouldn't be up-to-date.
Could you add it, please?

This revision is now accepted and ready to land.Jul 3 2017, 7:20 PM
alex-t closed this revision.Jul 4 2017, 9:07 AM
In D34407#798650, @chapuni wrote:

Excuse me, reverted in rL307054.

Seems a new test, llvm/test/CodeGen/AMDGPU/alignbit-pat.ll wouldn't be up-to-date.
Could you add it, please?

Sure I will. Thank you for handling this.

Revision Contents

PathSize
llvm/
trunk/
lib/
Target/
AMDGPU/
2 lines
test/
CodeGen/
AMDGPU/
18 lines
16 lines
8 lines
3 lines
55 lines
6 lines
20 lines
2 lines
8 lines
6 lines
2 lines
62 lines
48 lines
74 lines
91 lines
29 lines
19 lines
88 lines
2 lines
2 lines
9 lines
14 lines
4 lines
40 lines
12 lines
18 lines
2 lines
4 lines
12 lines
9 lines
91 lines
6 lines
4 lines
4 lines
4 lines
4 lines
18 lines
12 lines
18 lines
6 lines
6 lines
6 lines
6 lines
4 lines
4 lines
6 lines
10 lines
10 lines
6 lines
4 lines
6 lines
6 lines
6 lines
4 lines
6 lines
4 lines
4 lines
4 lines
6 lines
8 lines
2 lines
2 lines
2 lines
2 lines
2 lines
4 lines
4 lines
4 lines
4 lines
4 lines
4 lines
4 lines
4 lines
4 lines
8 lines
4 lines
4 lines
4 lines
6 lines
4 lines
4 lines
4 lines
10 lines
6 lines
10 lines
8 lines
10 lines
10 lines
10 lines
4 lines
4 lines
2 lines
6 lines
2 lines
6 lines
2 lines
6 lines
4 lines
4 lines
4 lines
4 lines
4 lines
6 lines
6 lines
4 lines
2 lines
4 lines
6 lines
6 lines
6 lines
4 lines
8 lines
4 lines
4 lines
2 lines
2 lines
4 lines
4 lines
2 lines
4 lines
4 lines
6 lines
6 lines
2 lines
6 lines
6 lines
4 lines
6 lines
4 lines
4 lines
6 lines
6 lines
8 lines
4 lines
6 lines
6 lines
6 lines
4 lines
2 lines
6 lines

Diff 105075

llvm/trunk/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

Show First 20 Lines • Show All 79 Lines • ▼ Show 20 Linesstatic cl::opt<bool> EnableLoadStoreVectorizer(
cl::desc("Enable load store vectorizer"), cl::desc("Enable load store vectorizer"),
cl::init(true), cl::init(true),
cl::Hidden); cl::Hidden);
// Option to to control global loads scalarization // Option to to control global loads scalarization
static cl::opt<bool> ScalarizeGlobal( static cl::opt<bool> ScalarizeGlobal(
"amdgpu-scalarize-global-loads", "amdgpu-scalarize-global-loads",
cl::desc("Enable global load scalarization"), cl::desc("Enable global load scalarization"),
cl::init(false), cl::init(true),
cl::Hidden); cl::Hidden);
// Option to run internalize pass. // Option to run internalize pass.
static cl::opt<bool> InternalizeSymbols( static cl::opt<bool> InternalizeSymbols(
"amdgpu-internalize-symbols", "amdgpu-internalize-symbols",
cl::desc("Enable elimination of non-kernel functions and unused globals"), cl::desc("Enable elimination of non-kernel functions and unused globals"),
cl::init(false), cl::init(false),
cl::Hidden); cl::Hidden);
▲ Show 20 Lines • Show All 761 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/add.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
;FUNC-LABEL: {{^}}test1: ;FUNC-LABEL: {{^}}test1:
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
;SI: v_add_i32_e32 [[REG:v[0-9]+]], vcc, {{v[0-9]+, v[0-9]+}} ;SI: s_add_i32 s[[REG:[0-9]+]], {{s[0-9]+, s[0-9]+}}
;SI-NOT: [[REG]] ;SI: v_mov_b32_e32 v[[REG]], s[[REG]]
;SI: buffer_store_dword [[REG]], ;SI: buffer_store_dword v[[REG]],
define amdgpu_kernel void @test1(i32 addrspace(1)* %out, i32 addrspace(1)* %in) { define amdgpu_kernel void @test1(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
%b_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1 %b_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
%a = load i32, i32 addrspace(1)* %in %a = load i32, i32 addrspace(1)* %in
%b = load i32, i32 addrspace(1)* %b_ptr %b = load i32, i32 addrspace(1)* %b_ptr
%result = add i32 %a, %b %result = add i32 %a, %b
store i32 %result, i32 addrspace(1)* %out store i32 %result, i32 addrspace(1)* %out
ret void ret void
} }
;FUNC-LABEL: {{^}}test2: ;FUNC-LABEL: {{^}}test2:
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}} ;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}} ;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
define amdgpu_kernel void @test2(<2 x i32> addrspace(1)* %out, <2 x i32> addrspace(1)* %in) { define amdgpu_kernel void @test2(<2 x i32> addrspace(1)* %out, <2 x i32> addrspace(1)* %in) {
%b_ptr = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %in, i32 1 %b_ptr = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %in, i32 1
%a = load <2 x i32>, <2 x i32> addrspace(1)* %in %a = load <2 x i32>, <2 x i32> addrspace(1)* %in
%b = load <2 x i32>, <2 x i32> addrspace(1)* %b_ptr %b = load <2 x i32>, <2 x i32> addrspace(1)* %b_ptr
%result = add <2 x i32> %a, %b %result = add <2 x i32> %a, %b
store <2 x i32> %result, <2 x i32> addrspace(1)* %out store <2 x i32> %result, <2 x i32> addrspace(1)* %out
ret void ret void
} }
;FUNC-LABEL: {{^}}test4: ;FUNC-LABEL: {{^}}test4:
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}} ;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}} ;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}} ;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}} ;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
define amdgpu_kernel void @test4(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) { define amdgpu_kernel void @test4(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
%b_ptr = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %in, i32 1 %b_ptr = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %in, i32 1
%a = load <4 x i32>, <4 x i32> addrspace(1)* %in %a = load <4 x i32>, <4 x i32> addrspace(1)* %in
%b = load <4 x i32>, <4 x i32> addrspace(1)* %b_ptr %b = load <4 x i32>, <4 x i32> addrspace(1)* %b_ptr
%result = add <4 x i32> %a, %b %result = add <4 x i32> %a, %b
store <4 x i32> %result, <4 x i32> addrspace(1)* %out store <4 x i32> %result, <4 x i32> addrspace(1)* %out
ret void ret void
▲ Show 20 Lines • Show All 136 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/add_i128.ll

Show All 13 Linesdefine amdgpu_kernel void @test_i128_vreg(i128 addrspace(1)* noalias %out, i128 addrspace(1)* noalias %inA, i128 addrspace(1)* noalias %inB) {
%b = load i128, i128 addrspace(1)* %b_ptr %b = load i128, i128 addrspace(1)* %b_ptr
%result = add i128 %a, %b %result = add i128 %a, %b
store i128 %result, i128 addrspace(1)* %out store i128 %result, i128 addrspace(1)* %out
ret void ret void
} }
; Check that the SGPR add operand is correctly moved to a VGPR. ; Check that the SGPR add operand is correctly moved to a VGPR.
; GCN-LABEL: {{^}}sgpr_operand: ; GCN-LABEL: {{^}}sgpr_operand:
; GCN: v_add_i32 ; GCN: s_add_u32
; GCN: v_addc_u32 ; GCN: s_addc_u32
; GCN: v_addc_u32 ; GCN: s_addc_u32
; GCN: v_addc_u32 ; GCN: s_addc_u32
define amdgpu_kernel void @sgpr_operand(i128 addrspace(1)* noalias %out, i128 addrspace(1)* noalias %in, i128 %a) { define amdgpu_kernel void @sgpr_operand(i128 addrspace(1)* noalias %out, i128 addrspace(1)* noalias %in, i128 %a) {
%foo = load i128, i128 addrspace(1)* %in, align 8 %foo = load i128, i128 addrspace(1)* %in, align 8
%result = add i128 %foo, %a %result = add i128 %foo, %a
store i128 %result, i128 addrspace(1)* %out store i128 %result, i128 addrspace(1)* %out
ret void ret void
} }
; GCN-LABEL: {{^}}sgpr_operand_reversed: ; GCN-LABEL: {{^}}sgpr_operand_reversed:
; GCN: v_add_i32 ; GCN: s_add_u32
; GCN: v_addc_u32 ; GCN: s_addc_u32
; GCN: v_addc_u32 ; GCN: s_addc_u32
; GCN: v_addc_u32 ; GCN: s_addc_u32
define amdgpu_kernel void @sgpr_operand_reversed(i128 addrspace(1)* noalias %out, i128 addrspace(1)* noalias %in, i128 %a) { define amdgpu_kernel void @sgpr_operand_reversed(i128 addrspace(1)* noalias %out, i128 addrspace(1)* noalias %in, i128 %a) {
%foo = load i128, i128 addrspace(1)* %in, align 8 %foo = load i128, i128 addrspace(1)* %in, align 8
%result = add i128 %a, %foo %result = add i128 %a, %foo
store i128 %result, i128 addrspace(1)* %out store i128 %result, i128 addrspace(1)* %out
ret void ret void
} }
; GCN-LABEL: {{^}}test_sreg: ; GCN-LABEL: {{^}}test_sreg:
Show All 11 Lines

llvm/trunk/test/CodeGen/AMDGPU/add_i64.ll

Show All 13 Linesdefine amdgpu_kernel void @test_i64_vreg(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %inA, i64 addrspace(1)* noalias %inB) {
%b = load i64, i64 addrspace(1)* %b_ptr %b = load i64, i64 addrspace(1)* %b_ptr
%result = add i64 %a, %b %result = add i64 %a, %b
store i64 %result, i64 addrspace(1)* %out store i64 %result, i64 addrspace(1)* %out
ret void ret void
} }
; Check that the SGPR add operand is correctly moved to a VGPR. ; Check that the SGPR add operand is correctly moved to a VGPR.
; SI-LABEL: {{^}}sgpr_operand: ; SI-LABEL: {{^}}sgpr_operand:
; SI: v_add_i32 ; SI: s_add_u32
; SI: v_addc_u32 ; SI: s_addc_u32
define amdgpu_kernel void @sgpr_operand(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in, i64 addrspace(1)* noalias %in_bar, i64 %a) { define amdgpu_kernel void @sgpr_operand(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in, i64 addrspace(1)* noalias %in_bar, i64 %a) {
%foo = load i64, i64 addrspace(1)* %in, align 8 %foo = load i64, i64 addrspace(1)* %in, align 8
%result = add i64 %foo, %a %result = add i64 %foo, %a
store i64 %result, i64 addrspace(1)* %out store i64 %result, i64 addrspace(1)* %out
ret void ret void
} }
; Swap the arguments. Check that the SGPR -> VGPR copy works with the ; Swap the arguments. Check that the SGPR -> VGPR copy works with the
; SGPR as other operand. ; SGPR as other operand.
; ;
; SI-LABEL: {{^}}sgpr_operand_reversed: ; SI-LABEL: {{^}}sgpr_operand_reversed:
; SI: v_add_i32 ; SI: s_add_u32
; SI: v_addc_u32 ; SI: s_addc_u32
define amdgpu_kernel void @sgpr_operand_reversed(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in, i64 %a) { define amdgpu_kernel void @sgpr_operand_reversed(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in, i64 %a) {
%foo = load i64, i64 addrspace(1)* %in, align 8 %foo = load i64, i64 addrspace(1)* %in, align 8
%result = add i64 %a, %foo %result = add i64 %a, %foo
store i64 %result, i64 addrspace(1)* %out store i64 %result, i64 addrspace(1)* %out
ret void ret void
} }
Show All 40 Lines

llvm/trunk/test/CodeGen/AMDGPU/and-gcn.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}v_and_i64_br: ; FUNC-LABEL: {{^}}v_and_i64_br:
; SI: v_and_b32 ; SI: s_and_b64
; SI: v_and_b32
define amdgpu_kernel void @v_and_i64_br(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr, i64 addrspace(1)* %bptr) { define amdgpu_kernel void @v_and_i64_br(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr, i64 addrspace(1)* %bptr) {
entry: entry:
%tid = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #0 %tid = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #0
%tmp0 = icmp eq i32 %tid, 0 %tmp0 = icmp eq i32 %tid, 0
br i1 %tmp0, label %if, label %endif br i1 %tmp0, label %if, label %endif
if: if:
%a = load i64, i64 addrspace(1)* %aptr, align 8 %a = load i64, i64 addrspace(1)* %aptr, align 8
Show All 13 Lines

llvm/trunk/test/CodeGen/AMDGPU/and.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i32 @llvm.r600.read.tidig.x() #0 declare i32 @llvm.r600.read.tidig.x() #0
; FUNC-LABEL: {{^}}test2: ; FUNC-LABEL: {{^}}test2:
; EG: AND_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: AND_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; EG: AND_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: AND_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
define amdgpu_kernel void @test2(<2 x i32> addrspace(1)* %out, <2 x i32> addrspace(1)* %in) { define amdgpu_kernel void @test2(<2 x i32> addrspace(1)* %out, <2 x i32> addrspace(1)* %in) {
%b_ptr = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %in, i32 1 %b_ptr = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %in, i32 1
%a = load <2 x i32>, <2 x i32> addrspace(1) * %in %a = load <2 x i32>, <2 x i32> addrspace(1) * %in
%b = load <2 x i32>, <2 x i32> addrspace(1) * %b_ptr %b = load <2 x i32>, <2 x i32> addrspace(1) * %b_ptr
%result = and <2 x i32> %a, %b %result = and <2 x i32> %a, %b
store <2 x i32> %result, <2 x i32> addrspace(1)* %out store <2 x i32> %result, <2 x i32> addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}test4: ; FUNC-LABEL: {{^}}test4:
; EG: AND_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: AND_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; EG: AND_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: AND_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; EG: AND_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: AND_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; EG: AND_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: AND_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
define amdgpu_kernel void @test4(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) { define amdgpu_kernel void @test4(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
%b_ptr = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %in, i32 1 %b_ptr = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %in, i32 1
%a = load <4 x i32>, <4 x i32> addrspace(1) * %in %a = load <4 x i32>, <4 x i32> addrspace(1) * %in
%b = load <4 x i32>, <4 x i32> addrspace(1) * %b_ptr %b = load <4 x i32>, <4 x i32> addrspace(1) * %b_ptr
%result = and <4 x i32> %a, %b %result = and <4 x i32> %a, %b
store <4 x i32> %result, <4 x i32> addrspace(1)* %out store <4 x i32> %result, <4 x i32> addrspace(1)* %out
ret void ret void
▲ Show 20 Lines • Show All 90 Lines • ▼ Show 20 Linesdefine amdgpu_kernel void @v_and_i32_vgpr_sgpr(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr, i32 %b) {
%and = and i32 %a, %b %and = and i32 %a, %b
store i32 %and, i32 addrspace(1)* %gep.out store i32 %and, i32 addrspace(1)* %gep.out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_and_constant_i32 ; FUNC-LABEL: {{^}}v_and_constant_i32
; SI: v_and_b32_e32 v{{[0-9]+}}, 0x12d687, v{{[0-9]+}} ; SI: v_and_b32_e32 v{{[0-9]+}}, 0x12d687, v{{[0-9]+}}
define amdgpu_kernel void @v_and_constant_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr) { define amdgpu_kernel void @v_and_constant_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr) {
%a = load i32, i32 addrspace(1)* %aptr, align 4 %tid = call i32 @llvm.r600.read.tidig.x() #0
%gep = getelementptr i32, i32 addrspace(1)* %aptr, i32 %tid
%a = load i32, i32 addrspace(1)* %gep, align 4
%and = and i32 %a, 1234567 %and = and i32 %a, 1234567
store i32 %and, i32 addrspace(1)* %out, align 4 store i32 %and, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_and_inline_imm_64_i32 ; FUNC-LABEL: {{^}}v_and_inline_imm_64_i32
; SI: v_and_b32_e32 v{{[0-9]+}}, 64, v{{[0-9]+}} ; SI: v_and_b32_e32 v{{[0-9]+}}, 64, v{{[0-9]+}}
define amdgpu_kernel void @v_and_inline_imm_64_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr) { define amdgpu_kernel void @v_and_inline_imm_64_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr) {
%a = load i32, i32 addrspace(1)* %aptr, align 4 %tid = call i32 @llvm.r600.read.tidig.x() #0
%gep = getelementptr i32, i32 addrspace(1)* %aptr, i32 %tid
%a = load i32, i32 addrspace(1)* %gep, align 4
%and = and i32 %a, 64 %and = and i32 %a, 64
store i32 %and, i32 addrspace(1)* %out, align 4 store i32 %and, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_and_inline_imm_neg_16_i32 ; FUNC-LABEL: {{^}}v_and_inline_imm_neg_16_i32
; SI: v_and_b32_e32 v{{[0-9]+}}, -16, v{{[0-9]+}} ; SI: v_and_b32_e32 v{{[0-9]+}}, -16, v{{[0-9]+}}
define amdgpu_kernel void @v_and_inline_imm_neg_16_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr) { define amdgpu_kernel void @v_and_inline_imm_neg_16_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr) {
%a = load i32, i32 addrspace(1)* %aptr, align 4 %tid = call i32 @llvm.r600.read.tidig.x() #0
%gep = getelementptr i32, i32 addrspace(1)* %aptr, i32 %tid
%a = load i32, i32 addrspace(1)* %gep, align 4
%and = and i32 %a, -16 %and = and i32 %a, -16
store i32 %and, i32 addrspace(1)* %out, align 4 store i32 %and, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}s_and_i64 ; FUNC-LABEL: {{^}}s_and_i64
; SI: s_and_b64 ; SI: s_and_b64
define amdgpu_kernel void @s_and_i64(i64 addrspace(1)* %out, i64 %a, i64 %b) { define amdgpu_kernel void @s_and_i64(i64 addrspace(1)* %out, i64 %a, i64 %b) {
▲ Show 20 Lines • Show All 68 Lines • ▼ Show 20 Linesdefine amdgpu_kernel void @s_and_multi_use_inline_imm_i64(i64 addrspace(1)* %out, i64 %a, i64 %b, i64 %c) {
store volatile i64 %add1, i64 addrspace(1)* %out store volatile i64 %add1, i64 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_and_i64: ; FUNC-LABEL: {{^}}v_and_i64:
; SI: v_and_b32 ; SI: v_and_b32
; SI: v_and_b32 ; SI: v_and_b32
define amdgpu_kernel void @v_and_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr, i64 addrspace(1)* %bptr) { define amdgpu_kernel void @v_and_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr, i64 addrspace(1)* %bptr) {
%a = load i64, i64 addrspace(1)* %aptr, align 8 %tid = call i32 @llvm.r600.read.tidig.x() #0
%b = load i64, i64 addrspace(1)* %bptr, align 8 %gep.a = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%a = load i64, i64 addrspace(1)* %gep.a, align 8
%gep.b = getelementptr i64, i64 addrspace(1)* %bptr, i32 %tid
%b = load i64, i64 addrspace(1)* %gep.b, align 8
%and = and i64 %a, %b %and = and i64 %a, %b
store i64 %and, i64 addrspace(1)* %out, align 8 store i64 %and, i64 addrspace(1)* %out, align 8
ret void ret void
} }
; FUNC-LABEL: {{^}}v_and_constant_i64: ; FUNC-LABEL: {{^}}v_and_constant_i64:
; SI-DAG: v_and_b32_e32 {{v[0-9]+}}, 0xab19b207, {{v[0-9]+}} ; SI-DAG: v_and_b32_e32 {{v[0-9]+}}, 0xab19b207, {{v[0-9]+}}
; SI-DAG: v_and_b32_e32 {{v[0-9]+}}, 0x11e, {{v[0-9]+}} ; SI-DAG: v_and_b32_e32 {{v[0-9]+}}, 0x11e, {{v[0-9]+}}
; SI: buffer_store_dwordx2 ; SI: buffer_store_dwordx2
define amdgpu_kernel void @v_and_constant_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) { define amdgpu_kernel void @v_and_constant_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) {
%a = load i64, i64 addrspace(1)* %aptr, align 8 %tid = call i32 @llvm.r600.read.tidig.x() #0
%gep.a = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%a = load i64, i64 addrspace(1)* %gep.a, align 8
%and = and i64 %a, 1231231234567 %and = and i64 %a, 1231231234567
store i64 %and, i64 addrspace(1)* %out, align 8 store i64 %and, i64 addrspace(1)* %out, align 8
ret void ret void
} }
; FUNC-LABEL: {{^}}v_and_multi_use_constant_i64: ; FUNC-LABEL: {{^}}v_and_multi_use_constant_i64:
; SI-DAG: buffer_load_dwordx2 v{{\[}}[[LO0:[0-9]+]]:[[HI0:[0-9]+]]{{\]}} ; SI-DAG: buffer_load_dwordx2 v{{\[}}[[LO0:[0-9]+]]:[[HI0:[0-9]+]]{{\]}}
; SI-DAG: buffer_load_dwordx2 v{{\[}}[[LO1:[0-9]+]]:[[HI1:[0-9]+]]{{\]}} ; SI-DAG: buffer_load_dwordx2 v{{\[}}[[LO1:[0-9]+]]:[[HI1:[0-9]+]]{{\]}}
Show All 31 Linesdefine amdgpu_kernel void @v_and_multi_use_inline_imm_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) {
%and0 = and i64 %a, 63 %and0 = and i64 %a, 63
%and1 = and i64 %b, 63 %and1 = and i64 %b, 63
store volatile i64 %and0, i64 addrspace(1)* %out store volatile i64 %and0, i64 addrspace(1)* %out
store volatile i64 %and1, i64 addrspace(1)* %out store volatile i64 %and1, i64 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_and_i64_32_bit_constant: ; FUNC-LABEL: {{^}}v_and_i64_32_bit_constant:
; SI: buffer_load_dword [[VAL:v[0-9]+]] ; SI: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; SI-NOT: and ; SI-NOT: and
; SI: v_and_b32_e32 {{v[0-9]+}}, 0x12d687, [[VAL]] ; SI: v_and_b32_e32 {{v[0-9]+}}, 0x12d687, [[VAL]]
; SI-NOT: and ; SI-NOT: and
; SI: buffer_store_dwordx2 ; SI: buffer_store_dwordx2
define amdgpu_kernel void @v_and_i64_32_bit_constant(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) { define amdgpu_kernel void @v_and_i64_32_bit_constant(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) {
%a = load i64, i64 addrspace(1)* %aptr, align 8 %tid = call i32 @llvm.r600.read.tidig.x() #0
%gep.a = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%a = load i64, i64 addrspace(1)* %gep.a, align 8
%and = and i64 %a, 1234567 %and = and i64 %a, 1234567
store i64 %and, i64 addrspace(1)* %out, align 8 store i64 %and, i64 addrspace(1)* %out, align 8
ret void ret void
} }
; FUNC-LABEL: {{^}}v_and_inline_imm_i64: ; FUNC-LABEL: {{^}}v_and_inline_imm_i64:
; SI: buffer_load_dword v{{[0-9]+}} ; SI: {{buffer|flat}}_load_dword v{{[0-9]+}}
; SI-NOT: and ; SI-NOT: and
; SI: v_and_b32_e32 {{v[0-9]+}}, 64, {{v[0-9]+}} ; SI: v_and_b32_e32 {{v[0-9]+}}, 64, {{v[0-9]+}}
; SI-NOT: and ; SI-NOT: and
; SI: buffer_store_dwordx2 ; SI: buffer_store_dwordx2
define amdgpu_kernel void @v_and_inline_imm_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) { define amdgpu_kernel void @v_and_inline_imm_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) {
%a = load i64, i64 addrspace(1)* %aptr, align 8 %tid = call i32 @llvm.r600.read.tidig.x() #0
%gep.a = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%a = load i64, i64 addrspace(1)* %gep.a, align 8
%and = and i64 %a, 64 %and = and i64 %a, 64
store i64 %and, i64 addrspace(1)* %out, align 8 store i64 %and, i64 addrspace(1)* %out, align 8
ret void ret void
} }
; FIXME: Should be able to reduce load width ; FIXME: Should be able to reduce load width
; FUNC-LABEL: {{^}}v_and_inline_neg_imm_i64: ; FUNC-LABEL: {{^}}v_and_inline_neg_imm_i64:
; SI: buffer_load_dwordx2 v{{\[}}[[VAL_LO:[0-9]+]]:[[VAL_HI:[0-9]+]]{{\]}} ; SI: {{buffer|flat}}_load_dwordx2 v{{\[}}[[VAL_LO:[0-9]+]]:[[VAL_HI:[0-9]+]]{{\]}}
; SI-NOT: and ; SI-NOT: and
; SI: v_and_b32_e32 v[[VAL_LO]], -8, v[[VAL_LO]] ; SI: v_and_b32_e32 v[[VAL_LO]], -8, v[[VAL_LO]]
; SI-NOT: and ; SI-NOT: and
; SI: buffer_store_dwordx2 v{{\[}}[[VAL_LO]]:[[VAL_HI]]{{\]}} ; SI: buffer_store_dwordx2 v{{\[}}[[VAL_LO]]:[[VAL_HI]]{{\]}}
define amdgpu_kernel void @v_and_inline_neg_imm_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) { define amdgpu_kernel void @v_and_inline_neg_imm_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) {
%a = load i64, i64 addrspace(1)* %aptr, align 8 %tid = call i32 @llvm.r600.read.tidig.x() #0
%gep.a = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%a = load i64, i64 addrspace(1)* %gep.a, align 8
%and = and i64 %a, -8 %and = and i64 %a, -8
store i64 %and, i64 addrspace(1)* %out, align 8 store i64 %and, i64 addrspace(1)* %out, align 8
ret void ret void
} }
; FUNC-LABEL: {{^}}s_and_inline_imm_64_i64 ; FUNC-LABEL: {{^}}s_and_inline_imm_64_i64
; SI: s_load_dword ; SI: s_load_dword
; SI-NOT: and ; SI-NOT: and
▲ Show 20 Lines • Show All 200 Lines • ▼ Show 20 Lines
; SI: s_and_b32 s[[K_HI:[0-9]+]], s{{[0-9]+}}, -4.0 ; SI: s_and_b32 s[[K_HI:[0-9]+]], s{{[0-9]+}}, -4.0
; SI-NOT: and ; SI-NOT: and
; SI: buffer_store_dwordx2 ; SI: buffer_store_dwordx2
define amdgpu_kernel void @s_and_inline_high_imm_f32_neg_4.0_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr, i64 %a) { define amdgpu_kernel void @s_and_inline_high_imm_f32_neg_4.0_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr, i64 %a) {
%and = and i64 %a, 13871086852301127680 %and = and i64 %a, 13871086852301127680
store i64 %and, i64 addrspace(1)* %out, align 8 store i64 %and, i64 addrspace(1)* %out, align 8
ret void ret void
} }
attributes #0 = { nounwind readnone } attributes #0 = { nounwind readnone }

llvm/trunk/test/CodeGen/AMDGPU/any_extend_vector_inreg.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; GCN-LABEL: {{^}}any_extend_vector_inreg_v16i8_to_v4i32: ; GCN-LABEL: {{^}}any_extend_vector_inreg_v16i8_to_v4i32:
; GCN: {{buffer|flat}}_load_dwordx4 ; GCN: s_load_dwordx4
; GCN-DAG: {{buffer|flat}}_load_dwordx4 ; GCN-DAG: s_load_dwordx4
; GCN-DAG: {{buffer|flat}}_load_dword ; GCN-DAG: s_load_dword
; GCN: {{buffer|flat}}_store_byte ; GCN: {{buffer|flat}}_store_byte
; GCN: {{buffer|flat}}_store_byte ; GCN: {{buffer|flat}}_store_byte
; GCN: {{buffer|flat}}_store_byte ; GCN: {{buffer|flat}}_store_byte
; GCN: {{buffer|flat}}_store_byte ; GCN: {{buffer|flat}}_store_byte
; GCN: {{buffer|flat}}_store_byte ; GCN: {{buffer|flat}}_store_byte
; GCN: {{buffer|flat}}_store_byte ; GCN: {{buffer|flat}}_store_byte
▲ Show 20 Lines • Show All 43 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/bitreverse.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=FUNC %s
declare i32 @llvm.amdgcn.workitem.id.x() #1
declare i16 @llvm.bitreverse.i16(i16) #1 declare i16 @llvm.bitreverse.i16(i16) #1
declare i32 @llvm.bitreverse.i32(i32) #1 declare i32 @llvm.bitreverse.i32(i32) #1
declare i64 @llvm.bitreverse.i64(i64) #1 declare i64 @llvm.bitreverse.i64(i64) #1
declare <2 x i32> @llvm.bitreverse.v2i32(<2 x i32>) #1 declare <2 x i32> @llvm.bitreverse.v2i32(<2 x i32>) #1
declare <4 x i32> @llvm.bitreverse.v4i32(<4 x i32>) #1 declare <4 x i32> @llvm.bitreverse.v4i32(<4 x i32>) #1
declare <2 x i64> @llvm.bitreverse.v2i64(<2 x i64>) #1 declare <2 x i64> @llvm.bitreverse.v2i64(<2 x i64>) #1
Show All 24 Lines
; SI: s_endpgm ; SI: s_endpgm
define amdgpu_kernel void @s_brev_i32(i32 addrspace(1)* noalias %out, i32 %val) #0 { define amdgpu_kernel void @s_brev_i32(i32 addrspace(1)* noalias %out, i32 %val) #0 {
%brev = call i32 @llvm.bitreverse.i32(i32 %val) #1 %brev = call i32 @llvm.bitreverse.i32(i32 %val) #1
store i32 %brev, i32 addrspace(1)* %out store i32 %brev, i32 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_brev_i32: ; FUNC-LABEL: {{^}}v_brev_i32:
; SI: buffer_load_dword [[VAL:v[0-9]+]], ; SI: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; SI: v_bfrev_b32_e32 [[RESULT:v[0-9]+]], [[VAL]] ; SI: v_bfrev_b32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; SI: buffer_store_dword [[RESULT]], ; SI: buffer_store_dword [[RESULT]],
; SI: s_endpgm ; SI: s_endpgm
define amdgpu_kernel void @v_brev_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) #0 { define amdgpu_kernel void @v_brev_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) #0 {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %gep
%brev = call i32 @llvm.bitreverse.i32(i32 %val) #1 %brev = call i32 @llvm.bitreverse.i32(i32 %val) #1
store i32 %brev, i32 addrspace(1)* %out store i32 %brev, i32 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}s_brev_v2i32: ; FUNC-LABEL: {{^}}s_brev_v2i32:
; SI: s_brev_b32 ; SI: s_brev_b32
; SI: s_brev_b32 ; SI: s_brev_b32
define amdgpu_kernel void @s_brev_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> %val) #0 { define amdgpu_kernel void @s_brev_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> %val) #0 {
%brev = call <2 x i32> @llvm.bitreverse.v2i32(<2 x i32> %val) #1 %brev = call <2 x i32> @llvm.bitreverse.v2i32(<2 x i32> %val) #1
store <2 x i32> %brev, <2 x i32> addrspace(1)* %out store <2 x i32> %brev, <2 x i32> addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_brev_v2i32: ; FUNC-LABEL: {{^}}v_brev_v2i32:
; SI: v_bfrev_b32_e32 ; SI: v_bfrev_b32_e32
; SI: v_bfrev_b32_e32 ; SI: v_bfrev_b32_e32
define amdgpu_kernel void @v_brev_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) #0 { define amdgpu_kernel void @v_brev_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) #0 {
%val = load <2 x i32>, <2 x i32> addrspace(1)* %valptr %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %valptr, i32 %tid
%val = load <2 x i32>, <2 x i32> addrspace(1)* %gep
%brev = call <2 x i32> @llvm.bitreverse.v2i32(<2 x i32> %val) #1 %brev = call <2 x i32> @llvm.bitreverse.v2i32(<2 x i32> %val) #1
store <2 x i32> %brev, <2 x i32> addrspace(1)* %out store <2 x i32> %brev, <2 x i32> addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}s_brev_i64: ; FUNC-LABEL: {{^}}s_brev_i64:
define amdgpu_kernel void @s_brev_i64(i64 addrspace(1)* noalias %out, i64 %val) #0 { define amdgpu_kernel void @s_brev_i64(i64 addrspace(1)* noalias %out, i64 %val) #0 {
%brev = call i64 @llvm.bitreverse.i64(i64 %val) #1 %brev = call i64 @llvm.bitreverse.i64(i64 %val) #1
store i64 %brev, i64 addrspace(1)* %out store i64 %brev, i64 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_brev_i64: ; FUNC-LABEL: {{^}}v_brev_i64:
; SI-NOT: v_or_b32_e64 v{{[0-9]+}}, 0, 0 ; SI-NOT: v_or_b32_e64 v{{[0-9]+}}, 0, 0
define amdgpu_kernel void @v_brev_i64(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %valptr) #0 { define amdgpu_kernel void @v_brev_i64(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %valptr) #0 {
%val = load i64, i64 addrspace(1)* %valptr %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i64, i64 addrspace(1)* %valptr, i32 %tid
%val = load i64, i64 addrspace(1)* %gep
%brev = call i64 @llvm.bitreverse.i64(i64 %val) #1 %brev = call i64 @llvm.bitreverse.i64(i64 %val) #1
store i64 %brev, i64 addrspace(1)* %out store i64 %brev, i64 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}s_brev_v2i64: ; FUNC-LABEL: {{^}}s_brev_v2i64:
define amdgpu_kernel void @s_brev_v2i64(<2 x i64> addrspace(1)* noalias %out, <2 x i64> %val) #0 { define amdgpu_kernel void @s_brev_v2i64(<2 x i64> addrspace(1)* noalias %out, <2 x i64> %val) #0 {
%brev = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %val) #1 %brev = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %val) #1
store <2 x i64> %brev, <2 x i64> addrspace(1)* %out store <2 x i64> %brev, <2 x i64> addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_brev_v2i64: ; FUNC-LABEL: {{^}}v_brev_v2i64:
define amdgpu_kernel void @v_brev_v2i64(<2 x i64> addrspace(1)* noalias %out, <2 x i64> addrspace(1)* noalias %valptr) #0 { define amdgpu_kernel void @v_brev_v2i64(<2 x i64> addrspace(1)* noalias %out, <2 x i64> addrspace(1)* noalias %valptr) #0 {
%val = load <2 x i64>, <2 x i64> addrspace(1)* %valptr %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <2 x i64> , <2 x i64> addrspace(1)* %valptr, i32 %tid
%val = load <2 x i64>, <2 x i64> addrspace(1)* %gep
%brev = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %val) #1 %brev = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %val) #1
store <2 x i64> %brev, <2 x i64> addrspace(1)* %out store <2 x i64> %brev, <2 x i64> addrspace(1)* %out
ret void ret void
} }
attributes #0 = { nounwind } attributes #0 = { nounwind }
attributes #1 = { nounwind readnone } attributes #1 = { nounwind readnone }

llvm/trunk/test/CodeGen/AMDGPU/bswap.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
declare i32 @llvm.bswap.i32(i32) nounwind readnone declare i32 @llvm.bswap.i32(i32) nounwind readnone
declare <2 x i32> @llvm.bswap.v2i32(<2 x i32>) nounwind readnone declare <2 x i32> @llvm.bswap.v2i32(<2 x i32>) nounwind readnone
declare <4 x i32> @llvm.bswap.v4i32(<4 x i32>) nounwind readnone declare <4 x i32> @llvm.bswap.v4i32(<4 x i32>) nounwind readnone
declare <8 x i32> @llvm.bswap.v8i32(<8 x i32>) nounwind readnone declare <8 x i32> @llvm.bswap.v8i32(<8 x i32>) nounwind readnone
declare i64 @llvm.bswap.i64(i64) nounwind readnone declare i64 @llvm.bswap.i64(i64) nounwind readnone
declare <2 x i64> @llvm.bswap.v2i64(<2 x i64>) nounwind readnone declare <2 x i64> @llvm.bswap.v2i64(<2 x i64>) nounwind readnone
declare <4 x i64> @llvm.bswap.v4i64(<4 x i64>) nounwind readnone declare <4 x i64> @llvm.bswap.v4i64(<4 x i64>) nounwind readnone
; FUNC-LABEL: @test_bswap_i32 ; FUNC-LABEL: @test_bswap_i32
; SI: buffer_load_dword [[VAL:v[0-9]+]] ; SI: s_load_dword [[VAL:s[0-9]+]]
; SI-DAG: v_alignbit_b32 [[TMP0:v[0-9]+]], [[VAL]], [[VAL]], 8 ; SI-DAG: v_alignbit_b32 [[TMP0:v[0-9]+]], [[VAL]], [[VAL]], 8
; SI-DAG: v_alignbit_b32 [[TMP1:v[0-9]+]], [[VAL]], [[VAL]], 24 ; SI-DAG: v_alignbit_b32 [[TMP1:v[0-9]+]], [[VAL]], [[VAL]], 24
; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0xff00ff ; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0xff00ff
; SI: v_bfi_b32 [[RESULT:v[0-9]+]], [[K]], [[TMP1]], [[TMP0]] ; SI: v_bfi_b32 [[RESULT:v[0-9]+]], [[K]], [[TMP1]], [[TMP0]]
; SI: buffer_store_dword [[RESULT]] ; SI: buffer_store_dword [[RESULT]]
; SI: s_endpgm ; SI: s_endpgm
define amdgpu_kernel void @test_bswap_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) nounwind { define amdgpu_kernel void @test_bswap_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) nounwind {
%val = load i32, i32 addrspace(1)* %in, align 4 %val = load i32, i32 addrspace(1)* %in, align 4
▲ Show 20 Lines • Show All 96 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/cgp-addressing-modes-flat.ll

; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=bonaire < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-CI -check-prefix=OPT-CIVI %s ; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=bonaire < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-CI -check-prefix=OPT-CIVI %s
; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-VI -check-prefix=OPT-CIVI %s ; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-VI -check-prefix=OPT-CIVI %s
; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=gfx900 -mattr=-flat-for-global < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-GFX9 %s ; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=gfx900 -mattr=-flat-for-global < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-GFX9 %s
; RUN: llc -march=amdgcn -mcpu=bonaire -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=CIVI %s ; RUN: llc -march=amdgcn -amdgpu-scalarize-global-loads=false -mcpu=bonaire -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=CIVI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=CIVI %s ; RUN: llc -march=amdgcn -amdgpu-scalarize-global-loads=false -mcpu=tonga -mattr=-flat-for-global -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=CIVI %s
; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s ; RUN: llc -march=amdgcn -amdgpu-scalarize-global-loads=false -mcpu=gfx900 -mattr=-flat-for-global -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
; OPT-LABEL: @test_no_sink_flat_small_offset_i32( ; OPT-LABEL: @test_no_sink_flat_small_offset_i32(
; OPT-CIVI: getelementptr i32, i32 addrspace(4)* %in ; OPT-CIVI: getelementptr i32, i32 addrspace(4)* %in
; OPT-CIVI: br i1 ; OPT-CIVI: br i1
; OPT-CIVI-NOT: ptrtoint ; OPT-CIVI-NOT: ptrtoint
; OPT-GFX9: br ; OPT-GFX9: br
; OPT-GFX9: %sunkaddr = getelementptr i8, i8 addrspace(4)* %0, i64 28 ; OPT-GFX9: %sunkaddr = getelementptr i8, i8 addrspace(4)* %0, i64 28
Show All 20 Linesendif:
br label %done br label %done
done: done:
ret void ret void
} }
; OPT-LABEL: @test_sink_noop_addrspacecast_flat_to_global_i32( ; OPT-LABEL: @test_sink_noop_addrspacecast_flat_to_global_i32(
; OPT: getelementptr i32, i32 addrspace(4)* %out, ; OPT: getelementptr i32, i32 addrspace(4)* %out,
; OPT-CI-NOT: getelementptr ; rOPT-CI-NOT: getelementptr
; OPT: br i1 ; OPT: br i1
; OPT-CI: addrspacecast ; OPT-CI: addrspacecast
; OPT-CI: getelementptr ; OPT-CI: getelementptr
; OPT-CI: bitcast ; OPT-CI: bitcast
; OPT: br label ; OPT: br label
; GCN-LABEL: {{^}}test_sink_noop_addrspacecast_flat_to_global_i32: ; GCN-LABEL: {{^}}test_sink_noop_addrspacecast_flat_to_global_i32:
▲ Show 20 Lines • Show All 156 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/cgp-addressing-modes.ll

; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=tahiti < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-SI %s ; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=tahiti < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-SI %s
; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=bonaire < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-CI %s ; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=bonaire < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-CI %s
; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-VI %s ; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-VI %s
; RUN: llc -march=amdgcn -mcpu=tahiti -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -march=amdgcn -mcpu=tahiti -amdgpu-scalarize-global-loads=false -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=bonaire -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s ; RUN: llc -march=amdgcn -mcpu=bonaire -amdgpu-scalarize-global-loads=false -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -march=amdgcn -mcpu=tonga -amdgpu-scalarize-global-loads=false -mattr=-flat-for-global -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:64:64-p5:32:32-p24:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64" target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:64:64-p5:32:32-p24:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64"
; OPT-LABEL: @test_sink_global_small_offset_i32( ; OPT-LABEL: @test_sink_global_small_offset_i32(
; OPT-CI-NOT: getelementptr i32, i32 addrspace(1)* %in ; OPT-CI-NOT: getelementptr i32, i32 addrspace(1)* %in
; OPT-VI: getelementptr i32, i32 addrspace(1)* %in ; OPT-VI: getelementptr i32, i32 addrspace(1)* %in
; OPT: br i1 ; OPT: br i1
; OPT-CI: getelementptr i8, ; OPT-CI: getelementptr i8,
▲ Show 20 Lines • Show All 668 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/coalescer_remat.ll

; RUN: llc -march=amdgcn -verify-machineinstrs -mtriple=amdgcn-- -o - %s | FileCheck %s ; RUN: llc -march=amdgcn -verify-machineinstrs -mtriple=amdgcn-- -o - %s | FileCheck %s
declare float @llvm.fma.f32(float, float, float) declare float @llvm.fma.f32(float, float, float)
; This checks that rematerialization support of the coalescer does not ; This checks that rematerialization support of the coalescer does not
; unnecessarily widen the register class. Without those fixes > 20 VGprs ; unnecessarily widen the register class. Without those fixes > 20 VGprs
; are used here ; are used here
; Also check that some rematerialization of the 0 constant happened. ; Also check that some rematerialization of the 0 constant happened.
; CHECK-LABEL: foobar ; CHECK-LABEL: foobar
; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0 ; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0
; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0 ; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0
; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0 ; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0
; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0 ; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0
; It's probably OK if this is slightly higher: ; It's probably OK if this is slightly higher:
; CHECK: ; NumVgprs: 8 ; CHECK: ; NumVgprs: 4
define amdgpu_kernel void @foobar(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in, i32 %flag) { define amdgpu_kernel void @foobar(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in, i32 %flag) {
entry: entry:
%cmpflag = icmp eq i32 %flag, 1 %cmpflag = icmp eq i32 %flag, 1
br i1 %cmpflag, label %loop, label %exit br i1 %cmpflag, label %loop, label %exit
loop: loop:
%c = phi i32 [0, %entry], [%cnext, %loop] %c = phi i32 [0, %entry], [%cnext, %loop]
%v0 = phi float [0.0, %entry], [%fma.0, %loop] %v0 = phi float [0.0, %entry], [%fma.0, %loop]
Show All 34 Lines

llvm/trunk/test/CodeGen/AMDGPU/copy-illegal-type.ll

; RUN: llc -march=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -amdgpu-sdwa-peephole=0 < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -amdgpu-sdwa-peephole=0 < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=FUNC %s
declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
declare i32 @llvm.amdgcn.workitem.id.y() nounwind readnone declare i32 @llvm.amdgcn.workitem.id.y() nounwind readnone
; FUNC-LABEL: {{^}}test_copy_v4i8: ; FUNC-LABEL: {{^}}test_copy_v4i8:
; GCN: buffer_load_dword [[REG:v[0-9]+]] ; GCN: {{buffer|flat}}_load_dword [[REG:v[0-9]+]]
; GCN: buffer_store_dword [[REG]] ; GCN: buffer_store_dword [[REG]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind { define amdgpu_kernel void @test_copy_v4i8(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind {
%val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4 %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid.x
%val = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}test_copy_v4i8_x2: ; FUNC-LABEL: {{^}}test_copy_v4i8_x2:
; GCN: buffer_load_dword [[REG:v[0-9]+]] ; GCN: {{buffer|flat}}_load_dword [[REG:v[0-9]+]]
; GCN: buffer_store_dword [[REG]] ; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]] ; GCN: buffer_store_dword [[REG]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8_x2(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %in) nounwind { define amdgpu_kernel void @test_copy_v4i8_x2(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %in) nounwind {
%val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4 %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid.x
%val = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}test_copy_v4i8_x3: ; FUNC-LABEL: {{^}}test_copy_v4i8_x3:
; GCN: buffer_load_dword [[REG:v[0-9]+]] ; GCN: {{buffer|flat}}_load_dword [[REG:v[0-9]+]]
; GCN: buffer_store_dword [[REG]] ; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]] ; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]] ; GCN: buffer_store_dword [[REG]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8_x3(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %out2, <4 x i8> addrspace(1)* %in) nounwind { define amdgpu_kernel void @test_copy_v4i8_x3(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %out2, <4 x i8> addrspace(1)* %in) nounwind {
%val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4 %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid.x
%val = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out2, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out2, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}test_copy_v4i8_x4: ; FUNC-LABEL: {{^}}test_copy_v4i8_x4:
; GCN: buffer_load_dword [[REG:v[0-9]+]] ; GCN: {{buffer|flat}}_load_dword [[REG:v[0-9]+]]
; GCN: buffer_store_dword [[REG]] ; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]] ; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]] ; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]] ; GCN: buffer_store_dword [[REG]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8_x4(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %out2, <4 x i8> addrspace(1)* %out3, <4 x i8> addrspace(1)* %in) nounwind { define amdgpu_kernel void @test_copy_v4i8_x4(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %out2, <4 x i8> addrspace(1)* %out3, <4 x i8> addrspace(1)* %in) nounwind {
%val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4 %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid.x
%val = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out2, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out2, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out3, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out3, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}test_copy_v4i8_extra_use: ; FUNC-LABEL: {{^}}test_copy_v4i8_extra_use:
; GCN: buffer_load_dword ; GCN: {{buffer|flat}}_load_dword
; GCN-DAG: v_lshrrev_b32 ; GCN-DAG: v_lshrrev_b32
; GCN: v_and_b32 ; GCN: v_and_b32
; GCN: v_or_b32 ; GCN: v_or_b32
; GCN-DAG: buffer_store_dword ; GCN-DAG: buffer_store_dword
; GCN-DAG: buffer_store_dword ; GCN-DAG: buffer_store_dword
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8_extra_use(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %in) nounwind { define amdgpu_kernel void @test_copy_v4i8_extra_use(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %in) nounwind {
%val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4 %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid.x
%val = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
%add = add <4 x i8> %val, <i8 9, i8 9, i8 9, i8 9> %add = add <4 x i8> %val, <i8 9, i8 9, i8 9, i8 9>
store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
store <4 x i8> %add, <4 x i8> addrspace(1)* %out1, align 4 store <4 x i8> %add, <4 x i8> addrspace(1)* %out1, align 4
ret void ret void
} }
; FIXME: Need to handle non-uniform case for function below (load without gep). ; FIXME: Need to handle non-uniform case for function below (load without gep).
; FUNC-LABEL: {{^}}test_copy_v4i8_x2_extra_use: ; FUNC-LABEL: {{^}}test_copy_v4i8_x2_extra_use:
Show All 14 Linesdefine amdgpu_kernel void @test_copy_v4i8_x2_extra_use(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %out2, <4 x i8> addrspace(1)* %in) nounwind {
%add = add <4 x i8> %val, <i8 9, i8 9, i8 9, i8 9> %add = add <4 x i8> %val, <i8 9, i8 9, i8 9, i8 9>
store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
store <4 x i8> %add, <4 x i8> addrspace(1)* %out1, align 4 store <4 x i8> %add, <4 x i8> addrspace(1)* %out1, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out2, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out2, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}test_copy_v3i8_align4: ; FUNC-LABEL: {{^}}test_copy_v3i8_align4:
; GCN: buffer_load_dword ; GCN: {{buffer|flat}}_load_dword
; GCN-DAG: buffer_store_short v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}} ; GCN-DAG: buffer_store_short v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0 offset:2{{$}} ; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0 offset:2{{$}}
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v3i8_align4(<3 x i8> addrspace(1)* %out, <3 x i8> addrspace(1)* %in) nounwind { define amdgpu_kernel void @test_copy_v3i8_align4(<3 x i8> addrspace(1)* %out, <3 x i8> addrspace(1)* %in) nounwind {
%val = load <3 x i8>, <3 x i8> addrspace(1)* %in, align 4 %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <3 x i8>, <3 x i8> addrspace(1)* %in, i32 %tid.x
%val = load <3 x i8>, <3 x i8> addrspace(1)* %gep, align 4
store <3 x i8> %val, <3 x i8> addrspace(1)* %out, align 4 store <3 x i8> %val, <3 x i8> addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}test_copy_v3i8_align2: ; FUNC-LABEL: {{^}}test_copy_v3i8_align2:
; GCN-DAG: buffer_load_ushort v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}} ; GCN-DAG: {{buffer|flat}}_load_ushort v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
; GCN-DAG: buffer_load_ubyte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0 offset:2{{$}} ; GCN-DAG: {{buffer|flat}}_load_ubyte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0 offset:2{{$}}
; GCN-DAG: buffer_store_short v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}} ; GCN-DAG: buffer_store_short v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0 offset:2{{$}} ; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0 offset:2{{$}}
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v3i8_align2(<3 x i8> addrspace(1)* %out, <3 x i8> addrspace(1)* %in) nounwind { define amdgpu_kernel void @test_copy_v3i8_align2(<3 x i8> addrspace(1)* %out, <3 x i8> addrspace(1)* %in) nounwind {
%val = load <3 x i8>, <3 x i8> addrspace(1)* %in, align 2 %val = load <3 x i8>, <3 x i8> addrspace(1)* %in, align 2
store <3 x i8> %val, <3 x i8> addrspace(1)* %out, align 2 store <3 x i8> %val, <3 x i8> addrspace(1)* %out, align 2
ret void ret void
} }
; FUNC-LABEL: {{^}}test_copy_v3i8_align1: ; FUNC-LABEL: {{^}}test_copy_v3i8_align1:
; GCN: buffer_load_ubyte ; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_load_ubyte ; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_load_ubyte ; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_store_byte ; GCN: buffer_store_byte
; GCN: buffer_store_byte ; GCN: buffer_store_byte
; GCN: buffer_store_byte ; GCN: buffer_store_byte
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v3i8_align1(<3 x i8> addrspace(1)* %out, <3 x i8> addrspace(1)* %in) nounwind { define amdgpu_kernel void @test_copy_v3i8_align1(<3 x i8> addrspace(1)* %out, <3 x i8> addrspace(1)* %in) nounwind {
%val = load <3 x i8>, <3 x i8> addrspace(1)* %in, align 1 %val = load <3 x i8>, <3 x i8> addrspace(1)* %in, align 1
store <3 x i8> %val, <3 x i8> addrspace(1)* %out, align 1 store <3 x i8> %val, <3 x i8> addrspace(1)* %out, align 1
ret void ret void
} }
; FUNC-LABEL: {{^}}test_copy_v4i8_volatile_load: ; FUNC-LABEL: {{^}}test_copy_v4i8_volatile_load:
; GCN: buffer_load_ubyte ; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_load_ubyte ; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_load_ubyte ; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_load_ubyte ; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_store_dword ; GCN: buffer_store_dword
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8_volatile_load(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind { define amdgpu_kernel void @test_copy_v4i8_volatile_load(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind {
%val = load volatile <4 x i8>, <4 x i8> addrspace(1)* %in, align 4 %val = load volatile <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out, align 4 store <4 x i8> %val, <4 x i8> addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}test_copy_v4i8_volatile_store: ; FUNC-LABEL: {{^}}test_copy_v4i8_volatile_store:
; GCN: buffer_load_ubyte ; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_load_ubyte ; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_load_ubyte ; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_load_ubyte ; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_store_byte ; GCN: buffer_store_byte
; GCN: buffer_store_byte ; GCN: buffer_store_byte
; GCN: buffer_store_byte ; GCN: buffer_store_byte
; GCN: buffer_store_byte ; GCN: buffer_store_byte
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8_volatile_store(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind { define amdgpu_kernel void @test_copy_v4i8_volatile_store(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind {
%val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4 %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
store volatile <4 x i8> %val, <4 x i8> addrspace(1)* %out, align 4 store volatile <4 x i8> %val, <4 x i8> addrspace(1)* %out, align 4
ret void ret void
} }

llvm/trunk/test/CodeGen/AMDGPU/ctlz.ll

Show All 28 Lines
; EG: CNDE_INT ; EG: CNDE_INT
define amdgpu_kernel void @s_ctlz_i32(i32 addrspace(1)* noalias %out, i32 %val) nounwind { define amdgpu_kernel void @s_ctlz_i32(i32 addrspace(1)* noalias %out, i32 %val) nounwind {
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone
store i32 %ctlz, i32 addrspace(1)* %out, align 4 store i32 %ctlz, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_i32: ; FUNC-LABEL: {{^}}v_ctlz_i32:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN-DAG: v_ffbh_u32_e32 [[CTLZ:v[0-9]+]], [[VAL]] ; GCN-DAG: v_ffbh_u32_e32 [[CTLZ:v[0-9]+]], [[VAL]]
; GCN-DAG: v_cmp_ne_u32_e32 vcc, 0, [[CTLZ]] ; GCN-DAG: v_cmp_ne_u32_e32 vcc, 0, [[CTLZ]]
; GCN: v_cndmask_b32_e32 [[RESULT:v[0-9]+]], 32, [[CTLZ]], vcc ; GCN: v_cndmask_b32_e32 [[RESULT:v[0-9]+]], 32, [[CTLZ]], vcc
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
; EG: FFBH_UINT ; EG: FFBH_UINT
; EG: CNDE_INT ; EG: CNDE_INT
define amdgpu_kernel void @v_ctlz_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone
store i32 %ctlz, i32 addrspace(1)* %out, align 4 store i32 %ctlz, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_v2i32: ; FUNC-LABEL: {{^}}v_ctlz_v2i32:
; GCN: buffer_load_dwordx2 ; GCN: {{buffer|flat}}_load_dwordx2
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: buffer_store_dwordx2 ; GCN: buffer_store_dwordx2
; GCN: s_endpgm ; GCN: s_endpgm
; EG: FFBH_UINT ; EG: FFBH_UINT
; EG: CNDE_INT ; EG: CNDE_INT
; EG: FFBH_UINT ; EG: FFBH_UINT
; EG: CNDE_INT ; EG: CNDE_INT
define amdgpu_kernel void @v_ctlz_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) nounwind {
%val = load <2 x i32>, <2 x i32> addrspace(1)* %valptr, align 8 %tid = call i32 @llvm.r600.read.tidig.x()
%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
%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
} }
; FUNC-LABEL: {{^}}v_ctlz_v4i32: ; FUNC-LABEL: {{^}}v_ctlz_v4i32:
; GCN: buffer_load_dwordx4 ; GCN: {{buffer|flat}}_load_dwordx4
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: buffer_store_dwordx4 ; GCN: buffer_store_dwordx4
; GCN: s_endpgm ; GCN: s_endpgm
; EG-DAG: FFBH_UINT ; EG-DAG: FFBH_UINT
; EG-DAG: CNDE_INT ; EG-DAG: CNDE_INT
; EG-DAG: FFBH_UINT ; EG-DAG: FFBH_UINT
; EG-DAG: CNDE_INT ; EG-DAG: CNDE_INT
; EG-DAG: FFBH_UINT ; EG-DAG: FFBH_UINT
; EG-DAG: CNDE_INT ; EG-DAG: CNDE_INT
; EG-DAG: FFBH_UINT ; EG-DAG: FFBH_UINT
; EG-DAG: CNDE_INT ; EG-DAG: CNDE_INT
define amdgpu_kernel void @v_ctlz_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %valptr) nounwind {
%val = load <4 x i32>, <4 x i32> addrspace(1)* %valptr, align 16 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %valptr, i32 %tid
%val = load <4 x i32>, <4 x i32> addrspace(1)* %in.gep, align 16
%ctlz = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %val, i1 false) nounwind readnone %ctlz = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %val, i1 false) nounwind readnone
store <4 x i32> %ctlz, <4 x i32> addrspace(1)* %out, align 16 store <4 x i32> %ctlz, <4 x i32> addrspace(1)* %out, align 16
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_i8: ; FUNC-LABEL: {{^}}v_ctlz_i8:
; GCN: buffer_load_ubyte [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_ubyte [[VAL:v[0-9]+]],
; SI-DAG: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]] ; SI-DAG: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; VI-DAG: v_ffbh_u32_sdwa [[RESULT:v[0-9]+]], [[VAL]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 ; VI-DAG: v_ffbh_u32_sdwa [[RESULT:v[0-9]+]], [[VAL]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0
; GCN: buffer_store_byte [[RESULT]], ; GCN: buffer_store_byte [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctlz_i8(i8 addrspace(1)* noalias %out, i8 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_i8(i8 addrspace(1)* noalias %out, i8 addrspace(1)* noalias %valptr) nounwind {
%val = load i8, i8 addrspace(1)* %valptr %val = load i8, i8 addrspace(1)* %valptr
%ctlz = call i8 @llvm.ctlz.i8(i8 %val, i1 false) nounwind readnone %ctlz = call i8 @llvm.ctlz.i8(i8 %val, i1 false) nounwind readnone
store i8 %ctlz, i8 addrspace(1)* %out store i8 %ctlz, i8 addrspace(1)* %out
▲ Show 20 Lines • Show All 54 Lines • ▼ Show 20 Linesdefine amdgpu_kernel void @v_ctlz_i64_trunc(i32 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in) nounwind {
%val = load i64, i64 addrspace(1)* %in.gep %val = load i64, i64 addrspace(1)* %in.gep
%ctlz = call i64 @llvm.ctlz.i64(i64 %val, i1 false) %ctlz = call i64 @llvm.ctlz.i64(i64 %val, i1 false)
%trunc = trunc i64 %ctlz to i32 %trunc = trunc i64 %ctlz to i32
store i32 %trunc, i32 addrspace(1)* %out.gep store i32 %trunc, i32 addrspace(1)* %out.gep
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_i32_sel_eq_neg1: ; FUNC-LABEL: {{^}}v_ctlz_i32_sel_eq_neg1:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]] ; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctlz_i32_sel_eq_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_i32_sel_eq_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone
%cmp = icmp eq i32 %val, 0 %cmp = icmp eq i32 %val, 0
%sel = select i1 %cmp, i32 -1, i32 %ctlz %sel = select i1 %cmp, i32 -1, i32 %ctlz
store i32 %sel, i32 addrspace(1)* %out store i32 %sel, i32 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_i32_sel_ne_neg1: ; FUNC-LABEL: {{^}}v_ctlz_i32_sel_ne_neg1:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]] ; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctlz_i32_sel_ne_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_i32_sel_ne_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone
%cmp = icmp ne i32 %val, 0 %cmp = icmp ne i32 %val, 0
%sel = select i1 %cmp, i32 %ctlz, i32 -1 %sel = select i1 %cmp, i32 %ctlz, i32 -1
store i32 %sel, i32 addrspace(1)* %out store i32 %sel, i32 addrspace(1)* %out
ret void ret void
} }
; TODO: Should be able to eliminate select here as well. ; TODO: Should be able to eliminate select here as well.
; FUNC-LABEL: {{^}}v_ctlz_i32_sel_eq_bitwidth: ; FUNC-LABEL: {{^}}v_ctlz_i32_sel_eq_bitwidth:
; GCN: buffer_load_dword ; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_cmp ; GCN: v_cmp
; GCN: v_cndmask ; GCN: v_cndmask
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctlz_i32_sel_eq_bitwidth(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_i32_sel_eq_bitwidth(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone
%cmp = icmp eq i32 %ctlz, 32 %cmp = icmp eq i32 %ctlz, 32
%sel = select i1 %cmp, i32 -1, i32 %ctlz %sel = select i1 %cmp, i32 -1, i32 %ctlz
store i32 %sel, i32 addrspace(1)* %out store i32 %sel, i32 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_i32_sel_ne_bitwidth: ; FUNC-LABEL: {{^}}v_ctlz_i32_sel_ne_bitwidth:
; GCN: buffer_load_dword ; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_cmp ; GCN: v_cmp
; GCN: v_cndmask ; GCN: v_cndmask
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctlz_i32_sel_ne_bitwidth(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_i32_sel_ne_bitwidth(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone
%cmp = icmp ne i32 %ctlz, 32 %cmp = icmp ne i32 %ctlz, 32
%sel = select i1 %cmp, i32 %ctlz, i32 -1 %sel = select i1 %cmp, i32 %ctlz, i32 -1
store i32 %sel, i32 addrspace(1)* %out store i32 %sel, i32 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_i8_sel_eq_neg1: ; FUNC-LABEL: {{^}}v_ctlz_i8_sel_eq_neg1:
; GCN: {{buffer|flat}}_load_ubyte [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_ubyte [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[FFBH:v[0-9]+]], [[VAL]] ; GCN: v_ffbh_u32_e32 [[FFBH:v[0-9]+]], [[VAL]]
; GCN: {{buffer|flat}}_store_byte [[FFBH]], ; GCN: {{buffer|flat}}_store_byte [[FFBH]],
define amdgpu_kernel void @v_ctlz_i8_sel_eq_neg1(i8 addrspace(1)* noalias %out, i8 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_i8_sel_eq_neg1(i8 addrspace(1)* noalias %out, i8 addrspace(1)* noalias %valptr) nounwind {
%tid = call i32 @llvm.r600.read.tidig.x() %tid = call i32 @llvm.r600.read.tidig.x()
%valptr.gep = getelementptr i8, i8 addrspace(1)* %valptr, i32 %tid %valptr.gep = getelementptr i8, i8 addrspace(1)* %valptr, i32 %tid
%val = load i8, i8 addrspace(1)* %valptr.gep %val = load i8, i8 addrspace(1)* %valptr.gep
%ctlz = call i8 @llvm.ctlz.i8(i8 %val, i1 false) nounwind readnone %ctlz = call i8 @llvm.ctlz.i8(i8 %val, i1 false) nounwind readnone
%cmp = icmp eq i8 %val, 0 %cmp = icmp eq i8 %val, 0
%sel = select i1 %cmp, i8 -1, i8 %ctlz %sel = select i1 %cmp, i8 -1, i8 %ctlz
store i8 %sel, i8 addrspace(1)* %out store i8 %sel, i8 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_i16_sel_eq_neg1: ; FUNC-LABEL: {{^}}v_ctlz_i16_sel_eq_neg1:
; SI: buffer_load_ushort [[VAL:v[0-9]+]], ; SI: {{buffer|flat}}_load_ushort [[VAL:v[0-9]+]],
; SI: v_ffbh_u32_e32 [[FFBH:v[0-9]+]], [[VAL]] ; SI: v_ffbh_u32_e32 [[FFBH:v[0-9]+]], [[VAL]]
; SI: buffer_store_short [[FFBH]], ; SI: buffer_store_short [[FFBH]],
define amdgpu_kernel void @v_ctlz_i16_sel_eq_neg1(i16 addrspace(1)* noalias %out, i16 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_i16_sel_eq_neg1(i16 addrspace(1)* noalias %out, i16 addrspace(1)* noalias %valptr) nounwind {
%val = load i16, i16 addrspace(1)* %valptr %val = load i16, i16 addrspace(1)* %valptr
%ctlz = call i16 @llvm.ctlz.i16(i16 %val, i1 false) nounwind readnone %ctlz = call i16 @llvm.ctlz.i16(i16 %val, i1 false) nounwind readnone
%cmp = icmp eq i16 %val, 0 %cmp = icmp eq i16 %val, 0
%sel = select i1 %cmp, i16 -1, i16 %ctlz %sel = select i1 %cmp, i16 -1, i16 %ctlz
store i16 %sel, i16 addrspace(1)* %out store i16 %sel, i16 addrspace(1)* %out
Show All 19 Lines

llvm/trunk/test/CodeGen/AMDGPU/ctlz_zero_undef.ll

Show All 23 Lines
; EG: FFBH_UINT {{\*? *}}[[RESULT]] ; EG: FFBH_UINT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @s_ctlz_zero_undef_i32(i32 addrspace(1)* noalias %out, i32 %val) nounwind { define amdgpu_kernel void @s_ctlz_zero_undef_i32(i32 addrspace(1)* noalias %out, i32 %val) nounwind {
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
store i32 %ctlz, i32 addrspace(1)* %out, align 4 store i32 %ctlz, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]] ; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+\.[XYZW]]] ; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+\.[XYZW]]]
; EG: FFBH_UINT {{\*? *}}[[RESULT]] ; EG: FFBH_UINT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_ctlz_zero_undef_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
store i32 %ctlz, i32 addrspace(1)* %out, align 4 store i32 %ctlz, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_v2i32: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_v2i32:
; GCN: buffer_load_dwordx2 ; GCN: {{buffer|flat}}_load_dwordx2
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: buffer_store_dwordx2 ; GCN: buffer_store_dwordx2
; GCN: s_endpgm ; GCN: s_endpgm
; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+]]{{\.[XYZW]}} ; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+]]{{\.[XYZW]}}
; EG: FFBH_UINT {{\*? *}}[[RESULT]] ; EG: FFBH_UINT {{\*? *}}[[RESULT]]
; EG: FFBH_UINT {{\*? *}}[[RESULT]] ; EG: FFBH_UINT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_ctlz_zero_undef_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) nounwind {
%val = load <2 x i32>, <2 x i32> addrspace(1)* %valptr, align 8 %tid = call i32 @llvm.r600.read.tidig.x()
%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
%ctlz = call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> %val, i1 true) nounwind readnone %ctlz = call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> %val, i1 true) 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
} }
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_v4i32: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_v4i32:
; GCN: buffer_load_dwordx4 ; GCN: {{buffer|flat}}_load_dwordx4
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: buffer_store_dwordx4 ; GCN: buffer_store_dwordx4
; GCN: s_endpgm ; GCN: s_endpgm
; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+]]{{\.[XYZW]}} ; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+]]{{\.[XYZW]}}
; EG: FFBH_UINT {{\*? *}}[[RESULT]] ; EG: FFBH_UINT {{\*? *}}[[RESULT]]
; EG: FFBH_UINT {{\*? *}}[[RESULT]] ; EG: FFBH_UINT {{\*? *}}[[RESULT]]
; EG: FFBH_UINT {{\*? *}}[[RESULT]] ; EG: FFBH_UINT {{\*? *}}[[RESULT]]
; EG: FFBH_UINT {{\*? *}}[[RESULT]] ; EG: FFBH_UINT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_ctlz_zero_undef_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %valptr) nounwind {
%val = load <4 x i32>, <4 x i32> addrspace(1)* %valptr, align 16 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %valptr, i32 %tid
%val = load <4 x i32>, <4 x i32> addrspace(1)* %in.gep, align 16
%ctlz = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %val, i1 true) nounwind readnone %ctlz = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %val, i1 true) nounwind readnone
store <4 x i32> %ctlz, <4 x i32> addrspace(1)* %out, align 16 store <4 x i32> %ctlz, <4 x i32> addrspace(1)* %out, align 16
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i8: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i8:
; GCN: buffer_load_ubyte [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_ubyte [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]] ; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_byte [[RESULT]], ; GCN: buffer_store_byte [[RESULT]],
define amdgpu_kernel void @v_ctlz_zero_undef_i8(i8 addrspace(1)* noalias %out, i8 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_i8(i8 addrspace(1)* noalias %out, i8 addrspace(1)* noalias %valptr) nounwind {
%val = load i8, i8 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i8, i8 addrspace(1)* %valptr, i32 %tid
%val = load i8, i8 addrspace(1)* %in.gep
%ctlz = call i8 @llvm.ctlz.i8(i8 %val, i1 true) nounwind readnone %ctlz = call i8 @llvm.ctlz.i8(i8 %val, i1 true) nounwind readnone
store i8 %ctlz, i8 addrspace(1)* %out store i8 %ctlz, i8 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}s_ctlz_zero_undef_i64: ; FUNC-LABEL: {{^}}s_ctlz_zero_undef_i64:
; GCN: s_load_dwordx2 s{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, {{0xb|0x2c}} ; GCN: s_load_dwordx2 s{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, {{0xb|0x2c}}
; GCN-DAG: v_cmp_eq_u32_e64 vcc, s[[HI]], 0{{$}} ; GCN-DAG: v_cmp_eq_u32_e64 vcc, s[[HI]], 0{{$}}
▲ Show 20 Lines • Show All 45 Lines • ▼ Show 20 Linesdefine amdgpu_kernel void @v_ctlz_zero_undef_i64_trunc(i32 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in) nounwind {
%val = load i64, i64 addrspace(1)* %in.gep %val = load i64, i64 addrspace(1)* %in.gep
%ctlz = call i64 @llvm.ctlz.i64(i64 %val, i1 true) %ctlz = call i64 @llvm.ctlz.i64(i64 %val, i1 true)
%trunc = trunc i64 %ctlz to i32 %trunc = trunc i64 %ctlz to i32
store i32 %trunc, i32 addrspace(1)* %out.gep store i32 %trunc, i32 addrspace(1)* %out.gep
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_neg1: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_neg1:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]] ; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp eq i32 %val, 0 %cmp = icmp eq i32 %val, 0
%sel = select i1 %cmp, i32 -1, i32 %ctlz %sel = select i1 %cmp, i32 -1, i32 %ctlz
store i32 %sel, i32 addrspace(1)* %out store i32 %sel, i32 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_ne_neg1: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_ne_neg1:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]] ; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_ne_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_ne_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp ne i32 %val, 0 %cmp = icmp ne i32 %val, 0
%sel = select i1 %cmp, i32 %ctlz, i32 -1 %sel = select i1 %cmp, i32 %ctlz, i32 -1
store i32 %sel, i32 addrspace(1)* %out store i32 %sel, i32 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i8_sel_eq_neg1: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i8_sel_eq_neg1:
; GCN: {{buffer|flat}}_load_ubyte [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_ubyte [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[FFBH:v[0-9]+]], [[VAL]] ; GCN: v_ffbh_u32_e32 [[FFBH:v[0-9]+]], [[VAL]]
; GCN: {{buffer|flat}}_store_byte [[FFBH]], ; GCN: {{buffer|flat}}_store_byte [[FFBH]],
define amdgpu_kernel void @v_ctlz_zero_undef_i8_sel_eq_neg1(i8 addrspace(1)* noalias %out, i8 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_i8_sel_eq_neg1(i8 addrspace(1)* noalias %out, i8 addrspace(1)* noalias %valptr) nounwind {
%tid = call i32 @llvm.r600.read.tidig.x() %tid = call i32 @llvm.r600.read.tidig.x()
%valptr.gep = getelementptr i8, i8 addrspace(1)* %valptr, i32 %tid %valptr.gep = getelementptr i8, i8 addrspace(1)* %valptr, i32 %tid
%val = load i8, i8 addrspace(1)* %valptr.gep %val = load i8, i8 addrspace(1)* %valptr.gep
%ctlz = call i8 @llvm.ctlz.i8(i8 %val, i1 true) nounwind readnone %ctlz = call i8 @llvm.ctlz.i8(i8 %val, i1 true) nounwind readnone
%cmp = icmp eq i8 %val, 0 %cmp = icmp eq i8 %val, 0
%sel = select i1 %cmp, i8 -1, i8 %ctlz %sel = select i1 %cmp, i8 -1, i8 %ctlz
store i8 %sel, i8 addrspace(1)* %out store i8 %sel, i8 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_neg1_two_use: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_neg1_two_use:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN-DAG: v_ffbh_u32_e32 [[RESULT0:v[0-9]+]], [[VAL]] ; GCN-DAG: v_ffbh_u32_e32 [[RESULT0:v[0-9]+]], [[VAL]]
; GCN-DAG: v_cmp_eq_u32_e32 vcc, 0, [[VAL]] ; GCN-DAG: v_cmp_eq_u32_e32 vcc, 0, [[VAL]]
; GCN-DAG: v_cndmask_b32_e64 [[RESULT1:v[0-9]+]], 0, 1, vcc ; GCN-DAG: v_cndmask_b32_e64 [[RESULT1:v[0-9]+]], 0, 1, vcc
; GCN-DAG: buffer_store_dword [[RESULT0]] ; GCN-DAG: buffer_store_dword [[RESULT0]]
; GCN-DAG: buffer_store_byte [[RESULT1]] ; GCN-DAG: buffer_store_byte [[RESULT1]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_neg1_two_use(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_neg1_two_use(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp eq i32 %val, 0 %cmp = icmp eq i32 %val, 0
%sel = select i1 %cmp, i32 -1, i32 %ctlz %sel = select i1 %cmp, i32 -1, i32 %ctlz
store volatile i32 %sel, i32 addrspace(1)* %out store volatile i32 %sel, i32 addrspace(1)* %out
store volatile i1 %cmp, i1 addrspace(1)* undef store volatile i1 %cmp, i1 addrspace(1)* undef
ret void ret void
} }
; Selected on wrong constant ; Selected on wrong constant
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_0: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_0:
; GCN: buffer_load_dword ; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_cmp ; GCN: v_cmp
; GCN: v_cndmask ; GCN: v_cndmask
; GCN: buffer_store_dword ; GCN: buffer_store_dword
define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp eq i32 %val, 0 %cmp = icmp eq i32 %val, 0
%sel = select i1 %cmp, i32 0, i32 %ctlz %sel = select i1 %cmp, i32 0, i32 %ctlz
store i32 %sel, i32 addrspace(1)* %out store i32 %sel, i32 addrspace(1)* %out
ret void ret void
} }
; Selected on wrong constant ; Selected on wrong constant
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_ne_0: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_ne_0:
; GCN: buffer_load_dword ; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_cmp ; GCN: v_cmp
; GCN: v_cndmask ; GCN: v_cndmask
; GCN: buffer_store_dword ; GCN: buffer_store_dword
define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_ne_0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_ne_0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp ne i32 %val, 0 %cmp = icmp ne i32 %val, 0
%sel = select i1 %cmp, i32 %ctlz, i32 0 %sel = select i1 %cmp, i32 %ctlz, i32 0
store i32 %sel, i32 addrspace(1)* %out store i32 %sel, i32 addrspace(1)* %out
ret void ret void
} }
; Compare on wrong constant ; Compare on wrong constant
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_cmp_non0: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_cmp_non0:
; GCN: buffer_load_dword ; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_cmp ; GCN: v_cmp
; GCN: v_cndmask ; GCN: v_cndmask
; GCN: buffer_store_dword ; GCN: buffer_store_dword
define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_cmp_non0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_cmp_non0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp eq i32 %val, 1 %cmp = icmp eq i32 %val, 1
%sel = select i1 %cmp, i32 0, i32 %ctlz %sel = select i1 %cmp, i32 0, i32 %ctlz
store i32 %sel, i32 addrspace(1)* %out store i32 %sel, i32 addrspace(1)* %out
ret void ret void
} }
; Selected on wrong constant ; Selected on wrong constant
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_ne_cmp_non0: ; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_ne_cmp_non0:
; GCN: buffer_load_dword ; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32 ; GCN: v_ffbh_u32_e32
; GCN: v_cmp ; GCN: v_cmp
; GCN: v_cndmask ; GCN: v_cndmask
; GCN: buffer_store_dword ; GCN: buffer_store_dword
define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_ne_cmp_non0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_ne_cmp_non0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone %ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp ne i32 %val, 1 %cmp = icmp ne i32 %val, 1
%sel = select i1 %cmp, i32 %ctlz, i32 0 %sel = select i1 %cmp, i32 %ctlz, i32 0
store i32 %sel, i32 addrspace(1)* %out store i32 %sel, i32 addrspace(1)* %out
ret void ret void
} }

llvm/trunk/test/CodeGen/AMDGPU/ctpop.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC -check-prefix=SI %s ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC -check-prefix=VI %s ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC -check-prefix=VI %s
; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i32 @llvm.ctpop.i32(i32) nounwind readnone declare i32 @llvm.ctpop.i32(i32) nounwind readnone
declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32>) nounwind readnone declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32>) nounwind readnone
declare <4 x i32> @llvm.ctpop.v4i32(<4 x i32>) nounwind readnone declare <4 x i32> @llvm.ctpop.v4i32(<4 x i32>) nounwind readnone
declare <8 x i32> @llvm.ctpop.v8i32(<8 x i32>) nounwind readnone declare <8 x i32> @llvm.ctpop.v8i32(<8 x i32>) nounwind readnone
declare <16 x i32> @llvm.ctpop.v16i32(<16 x i32>) nounwind readnone declare <16 x i32> @llvm.ctpop.v16i32(<16 x i32>) nounwind readnone
declare i32 @llvm.r600.read.tidig.x() nounwind readnone
; FUNC-LABEL: {{^}}s_ctpop_i32: ; FUNC-LABEL: {{^}}s_ctpop_i32:
; GCN: s_load_dword [[SVAL:s[0-9]+]], ; GCN: s_load_dword [[SVAL:s[0-9]+]],
; GCN: s_bcnt1_i32_b32 [[SRESULT:s[0-9]+]], [[SVAL]] ; GCN: s_bcnt1_i32_b32 [[SRESULT:s[0-9]+]], [[SVAL]]
; GCN: v_mov_b32_e32 [[VRESULT:v[0-9]+]], [[SRESULT]] ; GCN: v_mov_b32_e32 [[VRESULT:v[0-9]+]], [[SRESULT]]
; GCN: buffer_store_dword [[VRESULT]], ; GCN: buffer_store_dword [[VRESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @s_ctpop_i32(i32 addrspace(1)* noalias %out, i32 %val) nounwind { define amdgpu_kernel void @s_ctpop_i32(i32 addrspace(1)* noalias %out, i32 %val) nounwind {
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone %ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
store i32 %ctpop, i32 addrspace(1)* %out, align 4 store i32 %ctpop, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; XXX - Why 0 in register? ; XXX - Why 0 in register?
; FUNC-LABEL: {{^}}v_ctpop_i32: ; FUNC-LABEL: {{^}}v_ctpop_i32:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], 0 ; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], 0
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
%val = load i32, i32 addrspace(1)* %in, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone %ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
store i32 %ctpop, i32 addrspace(1)* %out, align 4 store i32 %ctpop, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_add_chain_i32: ; FUNC-LABEL: {{^}}v_ctpop_add_chain_i32:
; GCN: buffer_load_dword [[VAL1:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL0:v[0-9]+]],
; GCN: buffer_load_dword [[VAL0:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL1:v[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT:v[0-9]+]], [[VAL1]], 0 ; GCN: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT:v[0-9]+]], [[VAL1]], 0
; SI: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], [[VAL0]], [[MIDRESULT]] ; SI: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], [[VAL0]], [[MIDRESULT]]
; VI: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], [[VAL0]], [[MIDRESULT]] ; VI: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], [[VAL0]], [[MIDRESULT]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_add_chain_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in0, i32 addrspace(1)* noalias %in1) nounwind { define amdgpu_kernel void @v_ctpop_add_chain_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in0, i32 addrspace(1)* noalias %in1) nounwind {
%val0 = load i32, i32 addrspace(1)* %in0, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%val1 = load i32, i32 addrspace(1)* %in1, align 4 %in0.gep = getelementptr i32, i32 addrspace(1)* %in0, i32 %tid
%in1.gep = getelementptr i32, i32 addrspace(1)* %in1, i32 %tid
%val0 = load i32, i32 addrspace(1)* %in0.gep, align 4
%val1 = load i32, i32 addrspace(1)* %in1.gep, align 4
%ctpop0 = call i32 @llvm.ctpop.i32(i32 %val0) nounwind readnone %ctpop0 = call i32 @llvm.ctpop.i32(i32 %val0) nounwind readnone
%ctpop1 = call i32 @llvm.ctpop.i32(i32 %val1) nounwind readnone %ctpop1 = call i32 @llvm.ctpop.i32(i32 %val1) nounwind readnone
%add = add i32 %ctpop0, %ctpop1 %add = add i32 %ctpop0, %ctpop1
store i32 %add, i32 addrspace(1)* %out, align 4 store i32 %add, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_add_sgpr_i32: ; FUNC-LABEL: {{^}}v_ctpop_add_sgpr_i32:
; GCN: buffer_load_dword [[VAL0:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL0:v[0-9]+]],
; GCN: s_waitcnt ; GCN: s_waitcnt
; GCN-NEXT: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL0]], s{{[0-9]+}} ; GCN-NEXT: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL0]], s{{[0-9]+}}
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_add_sgpr_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in0, i32 addrspace(1)* noalias %in1, i32 %sval) nounwind { define amdgpu_kernel void @v_ctpop_add_sgpr_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in, i32 %sval) nounwind {
%val0 = load i32, i32 addrspace(1)* %in0, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%ctpop0 = call i32 @llvm.ctpop.i32(i32 %val0) nounwind readnone %in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%add = add i32 %ctpop0, %sval %val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%add = add i32 %ctpop, %sval
store i32 %add, i32 addrspace(1)* %out, align 4 store i32 %add, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_v2i32: ; FUNC-LABEL: {{^}}v_ctpop_v2i32:
; GCN: v_bcnt_u32_b32{{(_e64)*}} ; GCN: v_bcnt_u32_b32{{(_e64)*}}
; GCN: v_bcnt_u32_b32{{(_e64)*}} ; GCN: v_bcnt_u32_b32{{(_e64)*}}
; GCN: s_endpgm ; GCN: s_endpgm
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %in) nounwind {
%val = load <2 x i32>, <2 x i32> addrspace(1)* %in, align 8 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %in, i32 %tid
%val = load <2 x i32>, <2 x i32> addrspace(1)* %in.gep, align 8
%ctpop = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %val) nounwind readnone %ctpop = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %val) nounwind readnone
store <2 x i32> %ctpop, <2 x i32> addrspace(1)* %out, align 8 store <2 x i32> %ctpop, <2 x i32> addrspace(1)* %out, align 8
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_v4i32: ; FUNC-LABEL: {{^}}v_ctpop_v4i32:
; GCN: v_bcnt_u32_b32{{(_e64)*}} ; GCN: v_bcnt_u32_b32{{(_e64)*}}
; GCN: v_bcnt_u32_b32{{(_e64)*}} ; GCN: v_bcnt_u32_b32{{(_e64)*}}
; GCN: v_bcnt_u32_b32{{(_e64)*}} ; GCN: v_bcnt_u32_b32{{(_e64)*}}
; GCN: v_bcnt_u32_b32{{(_e64)*}} ; GCN: v_bcnt_u32_b32{{(_e64)*}}
; GCN: s_endpgm ; GCN: s_endpgm
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %in) nounwind {
%val = load <4 x i32>, <4 x i32> addrspace(1)* %in, align 16 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %in, i32 %tid
%val = load <4 x i32>, <4 x i32> addrspace(1)* %in.gep, align 16
%ctpop = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %val) nounwind readnone %ctpop = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %val) nounwind readnone
store <4 x i32> %ctpop, <4 x i32> addrspace(1)* %out, align 16 store <4 x i32> %ctpop, <4 x i32> addrspace(1)* %out, align 16
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_v8i32: ; FUNC-LABEL: {{^}}v_ctpop_v8i32:
; GCN: v_bcnt_u32_b32{{(_e64)*}} ; GCN: v_bcnt_u32_b32{{(_e64)*}}
; GCN: v_bcnt_u32_b32{{(_e64)*}} ; GCN: v_bcnt_u32_b32{{(_e64)*}}
Show All 9 Lines
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_v8i32(<8 x i32> addrspace(1)* noalias %out, <8 x i32> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_v8i32(<8 x i32> addrspace(1)* noalias %out, <8 x i32> addrspace(1)* noalias %in) nounwind {
%val = load <8 x i32>, <8 x i32> addrspace(1)* %in, align 32 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr <8 x i32>, <8 x i32> addrspace(1)* %in, i32 %tid
%val = load <8 x i32>, <8 x i32> addrspace(1)* %in.gep, align 32
%ctpop = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %val) nounwind readnone %ctpop = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %val) nounwind readnone
store <8 x i32> %ctpop, <8 x i32> addrspace(1)* %out, align 32 store <8 x i32> %ctpop, <8 x i32> addrspace(1)* %out, align 32
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_v16i32: ; FUNC-LABEL: {{^}}v_ctpop_v16i32:
; GCN: v_bcnt_u32_b32{{(_e64)*}} ; GCN: v_bcnt_u32_b32{{(_e64)*}}
; GCN: v_bcnt_u32_b32{{(_e64)*}} ; GCN: v_bcnt_u32_b32{{(_e64)*}}
Show All 25 Lines
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_v16i32(<16 x i32> addrspace(1)* noalias %out, <16 x i32> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_v16i32(<16 x i32> addrspace(1)* noalias %out, <16 x i32> addrspace(1)* noalias %in) nounwind {
%val = load <16 x i32>, <16 x i32> addrspace(1)* %in, align 32 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr <16 x i32>, <16 x i32> addrspace(1)* %in, i32 %tid
%val = load <16 x i32>, <16 x i32> addrspace(1)* %in.gep, align 32
%ctpop = call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %val) nounwind readnone %ctpop = call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %val) nounwind readnone
store <16 x i32> %ctpop, <16 x i32> addrspace(1)* %out, align 32 store <16 x i32> %ctpop, <16 x i32> addrspace(1)* %out, align 32
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_i32_add_inline_constant: ; FUNC-LABEL: {{^}}v_ctpop_i32_add_inline_constant:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], 4 ; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], 4
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32_add_inline_constant(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_i32_add_inline_constant(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
%val = load i32, i32 addrspace(1)* %in, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone %ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%add = add i32 %ctpop, 4 %add = add i32 %ctpop, 4
store i32 %add, i32 addrspace(1)* %out, align 4 store i32 %add, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_i32_add_inline_constant_inv: ; FUNC-LABEL: {{^}}v_ctpop_i32_add_inline_constant_inv:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], 4 ; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], 4
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32_add_inline_constant_inv(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_i32_add_inline_constant_inv(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
%val = load i32, i32 addrspace(1)* %in, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone %ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%add = add i32 4, %ctpop %add = add i32 4, %ctpop
store i32 %add, i32 addrspace(1)* %out, align 4 store i32 %add, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_i32_add_literal: ; FUNC-LABEL: {{^}}v_ctpop_i32_add_literal:
; GCN-DAG: buffer_load_dword [[VAL:v[0-9]+]], ; GCN-DAG: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN-DAG: v_mov_b32_e32 [[LIT:v[0-9]+]], 0x1869f ; GCN-DAG: v_mov_b32_e32 [[LIT:v[0-9]+]], 0x1869f
; SI: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], [[VAL]], [[LIT]] ; SI: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], [[VAL]], [[LIT]]
; VI: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], [[VAL]], [[LIT]] ; VI: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], [[VAL]], [[LIT]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_i32_add_literal(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_i32_add_literal(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
%val = load i32, i32 addrspace(1)* %in, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone %ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%add = add i32 %ctpop, 99999 %add = add i32 %ctpop, 99999
store i32 %add, i32 addrspace(1)* %out, align 4 store i32 %add, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_i32_add_var: ; FUNC-LABEL: {{^}}v_ctpop_i32_add_var:
; GCN-DAG: buffer_load_dword [[VAL:v[0-9]+]], ; GCN-DAG: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN-DAG: s_load_dword [[VAR:s[0-9]+]], ; GCN-DAG: s_load_dword [[VAR:s[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], [[VAR]] ; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], [[VAR]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32_add_var(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in, i32 %const) nounwind { define amdgpu_kernel void @v_ctpop_i32_add_var(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in, i32 %const) nounwind {
%val = load i32, i32 addrspace(1)* %in, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone %ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%add = add i32 %ctpop, %const %add = add i32 %ctpop, %const
store i32 %add, i32 addrspace(1)* %out, align 4 store i32 %add, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_i32_add_var_inv: ; FUNC-LABEL: {{^}}v_ctpop_i32_add_var_inv:
; GCN-DAG: buffer_load_dword [[VAL:v[0-9]+]], ; GCN-DAG: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN-DAG: s_load_dword [[VAR:s[0-9]+]], ; GCN-DAG: s_load_dword [[VAR:s[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], [[VAR]] ; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], [[VAR]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32_add_var_inv(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in, i32 %const) nounwind { define amdgpu_kernel void @v_ctpop_i32_add_var_inv(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in, i32 %const) nounwind {
%val = load i32, i32 addrspace(1)* %in, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone %ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%add = add i32 %const, %ctpop %add = add i32 %const, %ctpop
store i32 %add, i32 addrspace(1)* %out, align 4 store i32 %add, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_i32_add_vvar_inv: ; FUNC-LABEL: {{^}}v_ctpop_i32_add_vvar_inv:
; GCN-DAG: buffer_load_dword [[VAL:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], {{0$}} ; SI: buffer_load_dword [[VAR:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0 addr64
; GCN-DAG: buffer_load_dword [[VAR:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], 0 offset:16 ; SI: buffer_load_dword [[VAL:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0 addr64
; SI: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], [[VAL]], [[VAR]] ; SI: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], [[VAR]], [[VAL]]
; VI: flat_load_dword [[VAL:v[0-9]+]], v[{{[0-9]+:[0-9]+}}]
; VI: flat_load_dword [[VAR:v[0-9]+]], v[{{[0-9]+:[0-9]+}}]
; VI: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], [[VAL]], [[VAR]] ; VI: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], [[VAL]], [[VAR]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
; EG: BCNT_INT ; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32_add_vvar_inv(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in, i32 addrspace(1)* noalias %constptr) nounwind { define amdgpu_kernel void @v_ctpop_i32_add_vvar_inv(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in, i32 addrspace(1)* noalias %constptr) nounwind {
%val = load i32, i32 addrspace(1)* %in, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone %ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%gep = getelementptr i32, i32 addrspace(1)* %constptr, i32 4 %gep = getelementptr i32, i32 addrspace(1)* %constptr, i32 %tid
%const = load i32, i32 addrspace(1)* %gep, align 4 %const = load i32, i32 addrspace(1)* %gep, align 4
%add = add i32 %const, %ctpop %add = add i32 %const, %ctpop
store i32 %add, i32 addrspace(1)* %out, align 4 store i32 %add, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FIXME: We currently disallow SALU instructions in all branches, ; FIXME: We currently disallow SALU instructions in all branches,
; but there are some cases when the should be allowed. ; but there are some cases when the should be allowed.
Show All 28 Lines

llvm/trunk/test/CodeGen/AMDGPU/ctpop64.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=GCN -check-prefix=FUNC %s
declare i32 @llvm.r600.read.tidig.x() nounwind readnone
declare i64 @llvm.ctpop.i64(i64) nounwind readnone declare i64 @llvm.ctpop.i64(i64) nounwind readnone
declare <2 x i64> @llvm.ctpop.v2i64(<2 x i64>) nounwind readnone declare <2 x i64> @llvm.ctpop.v2i64(<2 x i64>) nounwind readnone
declare <4 x i64> @llvm.ctpop.v4i64(<4 x i64>) nounwind readnone declare <4 x i64> @llvm.ctpop.v4i64(<4 x i64>) nounwind readnone
declare <8 x i64> @llvm.ctpop.v8i64(<8 x i64>) nounwind readnone declare <8 x i64> @llvm.ctpop.v8i64(<8 x i64>) nounwind readnone
declare <16 x i64> @llvm.ctpop.v16i64(<16 x i64>) nounwind readnone declare <16 x i64> @llvm.ctpop.v16i64(<16 x i64>) nounwind readnone
declare i65 @llvm.ctpop.i65(i65) nounwind readnone declare i65 @llvm.ctpop.i65(i65) nounwind readnone
declare i128 @llvm.ctpop.i128(i128) nounwind readnone declare i128 @llvm.ctpop.i128(i128) nounwind readnone
; FUNC-LABEL: {{^}}s_ctpop_i64: ; FUNC-LABEL: {{^}}s_ctpop_i64:
; SI: s_load_dwordx2 [[SVAL:s\[[0-9]+:[0-9]+\]]], s{{\[[0-9]+:[0-9]+\]}}, 0xb ; SI: s_load_dwordx2 [[SVAL:s\[[0-9]+:[0-9]+\]]], s{{\[[0-9]+:[0-9]+\]}}, 0xb
; VI: s_load_dwordx2 [[SVAL:s\[[0-9]+:[0-9]+\]]], s{{\[[0-9]+:[0-9]+\]}}, 0x2c ; VI: s_load_dwordx2 [[SVAL:s\[[0-9]+:[0-9]+\]]], s{{\[[0-9]+:[0-9]+\]}}, 0x2c
; GCN: s_bcnt1_i32_b64 [[SRESULT:s[0-9]+]], [[SVAL]] ; GCN: s_bcnt1_i32_b64 [[SRESULT:s[0-9]+]], [[SVAL]]
; GCN: v_mov_b32_e32 [[VRESULT:v[0-9]+]], [[SRESULT]] ; GCN: v_mov_b32_e32 [[VRESULT:v[0-9]+]], [[SRESULT]]
; GCN: buffer_store_dword [[VRESULT]], ; GCN: buffer_store_dword [[VRESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @s_ctpop_i64(i32 addrspace(1)* noalias %out, i64 %val) nounwind { define amdgpu_kernel void @s_ctpop_i64(i32 addrspace(1)* noalias %out, i64 %val) nounwind {
%ctpop = call i64 @llvm.ctpop.i64(i64 %val) nounwind readnone %ctpop = call i64 @llvm.ctpop.i64(i64 %val) nounwind readnone
%truncctpop = trunc i64 %ctpop to i32 %truncctpop = trunc i64 %ctpop to i32
store i32 %truncctpop, i32 addrspace(1)* %out, align 4 store i32 %truncctpop, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_i64: ; FUNC-LABEL: {{^}}v_ctpop_i64:
; GCN: buffer_load_dwordx2 v{{\[}}[[LOVAL:[0-9]+]]:[[HIVAL:[0-9]+]]{{\]}}, ; GCN: {{buffer|flat}}_load_dwordx2 v{{\[}}[[LOVAL:[0-9]+]]:[[HIVAL:[0-9]+]]{{\]}},
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT:v[0-9]+]], v[[LOVAL]], 0 ; GCN: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT:v[0-9]+]], v[[LOVAL]], 0
; SI-NEXT: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]] ; SI-NEXT: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]]
; VI-NEXT: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]] ; VI-NEXT: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_i64(i32 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_i64(i32 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in) nounwind {
%val = load i64, i64 addrspace(1)* %in, align 8 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i64, i64 addrspace(1)* %in, i32 %tid
%val = load i64, i64 addrspace(1)* %in.gep, align 8
%ctpop = call i64 @llvm.ctpop.i64(i64 %val) nounwind readnone %ctpop = call i64 @llvm.ctpop.i64(i64 %val) nounwind readnone
%truncctpop = trunc i64 %ctpop to i32 %truncctpop = trunc i64 %ctpop to i32
store i32 %truncctpop, i32 addrspace(1)* %out, align 4 store i32 %truncctpop, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_i64_user: ; FUNC-LABEL: {{^}}v_ctpop_i64_user:
; GCN: buffer_load_dwordx2 v{{\[}}[[LOVAL:[0-9]+]]:[[HIVAL:[0-9]+]]{{\]}}, ; GCN: {{buffer|flat}}_load_dwordx2 v{{\[}}[[LOVAL:[0-9]+]]:[[HIVAL:[0-9]+]]{{\]}},
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT:v[0-9]+]], v[[LOVAL]], 0 ; GCN: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT:v[0-9]+]], v[[LOVAL]], 0
; SI-NEXT: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]] ; SI-NEXT: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]]
; VI-NEXT: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]] ; VI-NEXT: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]]
; GCN-DAG: v_or_b32_e32 v[[RESULT_LO:[0-9]+]], s{{[0-9]+}}, [[RESULT]] ; GCN-DAG: v_or_b32_e32 v[[RESULT_LO:[0-9]+]], s{{[0-9]+}}, [[RESULT]]
; GCN-DAG: v_mov_b32_e32 v[[RESULT_HI:[0-9]+]], s{{[0-9]+}} ; GCN-DAG: v_mov_b32_e32 v[[RESULT_HI:[0-9]+]], s{{[0-9]+}}
; GCN: buffer_store_dwordx2 v{{\[}}[[RESULT_LO]]:[[RESULT_HI]]{{\]}} ; GCN: buffer_store_dwordx2 v{{\[}}[[RESULT_LO]]:[[RESULT_HI]]{{\]}}
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_i64_user(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in, i64 %s.val) nounwind { define amdgpu_kernel void @v_ctpop_i64_user(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in, i64 %s.val) nounwind {
%val = load i64, i64 addrspace(1)* %in, align 8 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i64, i64 addrspace(1)* %in, i32 %tid
%val = load i64, i64 addrspace(1)* %in.gep, align 8
%ctpop = call i64 @llvm.ctpop.i64(i64 %val) nounwind readnone %ctpop = call i64 @llvm.ctpop.i64(i64 %val) nounwind readnone
%or = or i64 %ctpop, %s.val %or = or i64 %ctpop, %s.val
store i64 %or, i64 addrspace(1)* %out store i64 %or, i64 addrspace(1)* %out
ret void ret void
} }
; FUNC-LABEL: {{^}}s_ctpop_v2i64: ; FUNC-LABEL: {{^}}s_ctpop_v2i64:
; GCN: s_bcnt1_i32_b64 ; GCN: s_bcnt1_i32_b64
Show All 21 Lines
; FUNC-LABEL: {{^}}v_ctpop_v2i64: ; FUNC-LABEL: {{^}}v_ctpop_v2i64:
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_v2i64(<2 x i32> addrspace(1)* noalias %out, <2 x i64> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_v2i64(<2 x i32> addrspace(1)* noalias %out, <2 x i64> addrspace(1)* noalias %in) nounwind {
%val = load <2 x i64>, <2 x i64> addrspace(1)* %in, align 16 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr <2 x i64>, <2 x i64> addrspace(1)* %in, i32 %tid
%val = load <2 x i64>, <2 x i64> addrspace(1)* %in.gep, align 16
%ctpop = call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %val) nounwind readnone %ctpop = call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %val) nounwind readnone
%truncctpop = trunc <2 x i64> %ctpop to <2 x i32> %truncctpop = trunc <2 x i64> %ctpop to <2 x i32>
store <2 x i32> %truncctpop, <2 x i32> addrspace(1)* %out, align 8 store <2 x i32> %truncctpop, <2 x i32> addrspace(1)* %out, align 8
ret void ret void
} }
; FUNC-LABEL: {{^}}v_ctpop_v4i64: ; FUNC-LABEL: {{^}}v_ctpop_v4i64:
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: v_bcnt_u32_b32 ; GCN: v_bcnt_u32_b32
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_v4i64(<4 x i32> addrspace(1)* noalias %out, <4 x i64> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_v4i64(<4 x i32> addrspace(1)* noalias %out, <4 x i64> addrspace(1)* noalias %in) nounwind {
%val = load <4 x i64>, <4 x i64> addrspace(1)* %in, align 32 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr <4 x i64>, <4 x i64> addrspace(1)* %in, i32 %tid
%val = load <4 x i64>, <4 x i64> addrspace(1)* %in.gep, align 32
%ctpop = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %val) nounwind readnone %ctpop = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %val) nounwind readnone
%truncctpop = trunc <4 x i64> %ctpop to <4 x i32> %truncctpop = trunc <4 x i64> %ctpop to <4 x i32>
store <4 x i32> %truncctpop, <4 x i32> addrspace(1)* %out, align 16 store <4 x i32> %truncctpop, <4 x i32> addrspace(1)* %out, align 16
ret void ret void
} }
; FUNC-LABEL: {{^}}ctpop_i64_in_br: ; FUNC-LABEL: {{^}}ctpop_i64_in_br:
; SI-DAG: s_load_dwordx2 s{{\[}}[[LOVAL:[0-9]+]]:[[HIVAL:[0-9]+]]{{\]}}, s[{{[0-9]+:[0-9]+}}], 0xd ; SI-DAG: s_load_dwordx2 s{{\[}}[[LOVAL:[0-9]+]]:[[HIVAL:[0-9]+]]{{\]}}, s[{{[0-9]+:[0-9]+}}], 0xd
▲ Show 20 Lines • Show All 47 Lines • ▼ Show 20 Linesdefine amdgpu_kernel void @s_ctpop_i65(i32 addrspace(1)* noalias %out, i65 %val) nounwind {
%truncctpop = trunc i65 %ctpop to i32 %truncctpop = trunc i65 %ctpop to i32
store i32 %truncctpop, i32 addrspace(1)* %out, align 4 store i32 %truncctpop, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FIXME: Should not have extra add ; FIXME: Should not have extra add
; FUNC-LABEL: {{^}}v_ctpop_i128: ; FUNC-LABEL: {{^}}v_ctpop_i128:
; GCN: buffer_load_dwordx4 v{{\[}}[[VAL0:[0-9]+]]:[[VAL3:[0-9]+]]{{\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}} ; SI: buffer_load_dwordx4 v{{\[}}[[VAL0:[0-9]+]]:[[VAL3:[0-9]+]]{{\]}}, v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64
; VI: flat_load_dwordx4 v{{\[}}[[VAL0:[0-9]+]]:[[VAL3:[0-9]+]]{{\]}}, v{{\[[0-9]+:[0-9]+\]}}
; GCN-DAG: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT0:v[0-9]+]], v{{[0-9]+}}, 0 ; GCN-DAG: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT0:v[0-9]+]], v{{[0-9]+}}, 0
; GCN-DAG: v_bcnt_u32_b32{{(_e32)*(_e64)*}} [[MIDRESULT1:v[0-9]+]], v[[VAL3]], [[MIDRESULT0]] ; GCN-DAG: v_bcnt_u32_b32{{(_e32)*(_e64)*}} [[MIDRESULT1:v[0-9]+]], v[[VAL3]], [[MIDRESULT0]]
; GCN-DAG: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT2:v[0-9]+]], v[[VAL0]], 0 ; GCN-DAG: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT2:v[0-9]+]], v[[VAL0]], 0
; GCN-DAG: v_bcnt_u32_b32{{(_e32)*(_e64)*}} [[MIDRESULT3:v[0-9]+]], v{{[0-9]+}}, [[MIDRESULT2]] ; GCN-DAG: v_bcnt_u32_b32{{(_e32)*(_e64)*}} [[MIDRESULT3:v[0-9]+]], v{{[0-9]+}}, [[MIDRESULT2]]
; GCN: v_add_i32_e32 [[RESULT:v[0-9]+]], vcc, [[MIDRESULT1]], [[MIDRESULT2]] ; GCN: v_add_i32_e32 [[RESULT:v[0-9]+]], vcc, [[MIDRESULT1]], [[MIDRESULT2]]
; GCN: buffer_store_dword [[RESULT]], ; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_i128(i32 addrspace(1)* noalias %out, i128 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v_ctpop_i128(i32 addrspace(1)* noalias %out, i128 addrspace(1)* noalias %in) nounwind {
%val = load i128, i128 addrspace(1)* %in, align 8 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i128, i128 addrspace(1)* %in, i32 %tid
%val = load i128, i128 addrspace(1)* %in.gep, align 8
%ctpop = call i128 @llvm.ctpop.i128(i128 %val) nounwind readnone %ctpop = call i128 @llvm.ctpop.i128(i128 %val) nounwind readnone
%truncctpop = trunc i128 %ctpop to i32 %truncctpop = trunc i128 %ctpop to i32
store i32 %truncctpop, i32 addrspace(1)* %out, align 4 store i32 %truncctpop, i32 addrspace(1)* %out, align 4
ret void ret void
} }

llvm/trunk/test/CodeGen/AMDGPU/cttz_zero_undef.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i32 @llvm.cttz.i32(i32, i1) nounwind readnone declare i32 @llvm.cttz.i32(i32, i1) nounwind readnone
declare <2 x i32> @llvm.cttz.v2i32(<2 x i32>, i1) nounwind readnone declare <2 x i32> @llvm.cttz.v2i32(<2 x i32>, i1) nounwind readnone
declare <4 x i32> @llvm.cttz.v4i32(<4 x i32>, i1) nounwind readnone declare <4 x i32> @llvm.cttz.v4i32(<4 x i32>, i1) nounwind readnone
declare i32 @llvm.r600.read.tidig.x() nounwind readnone
; FUNC-LABEL: {{^}}s_cttz_zero_undef_i32: ; FUNC-LABEL: {{^}}s_cttz_zero_undef_i32:
; SI: s_load_dword [[VAL:s[0-9]+]], ; SI: s_load_dword [[VAL:s[0-9]+]],
; SI: s_ff1_i32_b32 [[SRESULT:s[0-9]+]], [[VAL]] ; SI: s_ff1_i32_b32 [[SRESULT:s[0-9]+]], [[VAL]]
; SI: v_mov_b32_e32 [[VRESULT:v[0-9]+]], [[SRESULT]] ; SI: v_mov_b32_e32 [[VRESULT:v[0-9]+]], [[SRESULT]]
; SI: buffer_store_dword [[VRESULT]], ; SI: buffer_store_dword [[VRESULT]],
; SI: s_endpgm ; SI: s_endpgm
; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+\.[XYZW]]] ; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+\.[XYZW]]]
; EG: FFBL_INT {{\*? *}}[[RESULT]] ; EG: FFBL_INT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @s_cttz_zero_undef_i32(i32 addrspace(1)* noalias %out, i32 %val) nounwind { define amdgpu_kernel void @s_cttz_zero_undef_i32(i32 addrspace(1)* noalias %out, i32 %val) nounwind {
%cttz = call i32 @llvm.cttz.i32(i32 %val, i1 true) nounwind readnone %cttz = call i32 @llvm.cttz.i32(i32 %val, i1 true) nounwind readnone
store i32 %cttz, i32 addrspace(1)* %out, align 4 store i32 %cttz, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_cttz_zero_undef_i32: ; FUNC-LABEL: {{^}}v_cttz_zero_undef_i32:
; SI: buffer_load_dword [[VAL:v[0-9]+]], ; SI: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; SI: v_ffbl_b32_e32 [[RESULT:v[0-9]+]], [[VAL]] ; SI: v_ffbl_b32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; SI: buffer_store_dword [[RESULT]], ; SI: buffer_store_dword [[RESULT]],
; SI: s_endpgm ; SI: s_endpgm
; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+\.[XYZW]]] ; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+\.[XYZW]]]
; EG: FFBL_INT {{\*? *}}[[RESULT]] ; EG: FFBL_INT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_cttz_zero_undef_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_cttz_zero_undef_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
%val = load i32, i32 addrspace(1)* %valptr, align 4 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
%val = load i32, i32 addrspace(1)* %in.gep, align 4
%cttz = call i32 @llvm.cttz.i32(i32 %val, i1 true) nounwind readnone %cttz = call i32 @llvm.cttz.i32(i32 %val, i1 true) nounwind readnone
store i32 %cttz, i32 addrspace(1)* %out, align 4 store i32 %cttz, i32 addrspace(1)* %out, align 4
ret void ret void
} }
; FUNC-LABEL: {{^}}v_cttz_zero_undef_v2i32: ; FUNC-LABEL: {{^}}v_cttz_zero_undef_v2i32:
; SI: buffer_load_dwordx2 ; SI: {{buffer|flat}}_load_dwordx2
; SI: v_ffbl_b32_e32 ; SI: v_ffbl_b32_e32
; SI: v_ffbl_b32_e32 ; SI: v_ffbl_b32_e32
; SI: buffer_store_dwordx2 ; SI: buffer_store_dwordx2
; SI: s_endpgm ; SI: s_endpgm
; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+]]{{\.[XYZW]}} ; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+]]{{\.[XYZW]}}
; EG: FFBL_INT {{\*? *}}[[RESULT]] ; EG: FFBL_INT {{\*? *}}[[RESULT]]
; EG: FFBL_INT {{\*? *}}[[RESULT]] ; EG: FFBL_INT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_cttz_zero_undef_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_cttz_zero_undef_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) nounwind {
%val = load <2 x i32>, <2 x i32> addrspace(1)* %valptr, align 8 %tid = call i32 @llvm.r600.read.tidig.x()
%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
%cttz = call <2 x i32> @llvm.cttz.v2i32(<2 x i32> %val, i1 true) nounwind readnone %cttz = call <2 x i32> @llvm.cttz.v2i32(<2 x i32> %val, i1 true) nounwind readnone
store <2 x i32> %cttz, <2 x i32> addrspace(1)* %out, align 8 store <2 x i32> %cttz, <2 x i32> addrspace(1)* %out, align 8
ret void ret void
} }
; FUNC-LABEL: {{^}}v_cttz_zero_undef_v4i32: ; FUNC-LABEL: {{^}}v_cttz_zero_undef_v4i32:
; SI: buffer_load_dwordx4 ; SI: {{buffer|flat}}_load_dwordx4
; SI: v_ffbl_b32_e32 ; SI: v_ffbl_b32_e32
; SI: v_ffbl_b32_e32 ; SI: v_ffbl_b32_e32
; SI: v_ffbl_b32_e32 ; SI: v_ffbl_b32_e32
; SI: v_ffbl_b32_e32 ; SI: v_ffbl_b32_e32
; SI: buffer_store_dwordx4 ; SI: buffer_store_dwordx4
; SI: s_endpgm ; SI: s_endpgm
; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+]]{{\.[XYZW]}} ; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+]]{{\.[XYZW]}}
; EG: FFBL_INT {{\*? *}}[[RESULT]] ; EG: FFBL_INT {{\*? *}}[[RESULT]]
; EG: FFBL_INT {{\*? *}}[[RESULT]] ; EG: FFBL_INT {{\*? *}}[[RESULT]]
; EG: FFBL_INT {{\*? *}}[[RESULT]] ; EG: FFBL_INT {{\*? *}}[[RESULT]]
; EG: FFBL_INT {{\*? *}}[[RESULT]] ; EG: FFBL_INT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_cttz_zero_undef_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %valptr) nounwind { define amdgpu_kernel void @v_cttz_zero_undef_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %valptr) nounwind {
%val = load <4 x i32>, <4 x i32> addrspace(1)* %valptr, align 16 %tid = call i32 @llvm.r600.read.tidig.x()
%in.gep = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %valptr, i32 %tid
%val = load <4 x i32>, <4 x i32> addrspace(1)* %in.gep, align 16
%cttz = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> %val, i1 true) nounwind readnone %cttz = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> %val, i1 true) nounwind readnone
store <4 x i32> %cttz, <4 x i32> addrspace(1)* %out, align 16 store <4 x i32> %cttz, <4 x i32> addrspace(1)* %out, align 16
ret void ret void
} }

llvm/trunk/test/CodeGen/AMDGPU/cvt_f32_ubyte.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
declare i32 @llvm.amdgcn.workitem.id.y() nounwind readnone declare i32 @llvm.amdgcn.workitem.id.y() nounwind readnone
; GCN-LABEL: {{^}}load_i8_to_f32: ; GCN-LABEL: {{^}}load_i8_to_f32:
; GCN: buffer_load_ubyte [[LOADREG:v[0-9]+]], ; GCN: {{buffer|flat}}_load_ubyte [[LOADREG:v[0-9]+]],
; GCN-NOT: bfe ; GCN-NOT: bfe
; GCN-NOT: lshr ; GCN-NOT: lshr
; GCN: v_cvt_f32_ubyte0_e32 [[CONV:v[0-9]+]], [[LOADREG]] ; GCN: v_cvt_f32_ubyte0_e32 [[CONV:v[0-9]+]], [[LOADREG]]
; GCN: buffer_store_dword [[CONV]], ; GCN: buffer_store_dword [[CONV]],
define amdgpu_kernel void @load_i8_to_f32(float addrspace(1)* noalias %out, i8 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @load_i8_to_f32(float addrspace(1)* noalias %out, i8 addrspace(1)* noalias %in) nounwind {
%load = load i8, i8 addrspace(1)* %in, align 1 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i8, i8 addrspace(1)* %in, i32 %tid
%load = load i8, i8 addrspace(1)* %gep, align 1
%cvt = uitofp i8 %load to float %cvt = uitofp i8 %load to float
store float %cvt, float addrspace(1)* %out, align 4 store float %cvt, float addrspace(1)* %out, align 4
ret void ret void
} }
; GCN-LABEL: {{^}}load_v2i8_to_v2f32: ; GCN-LABEL: {{^}}load_v2i8_to_v2f32:
; GCN: buffer_load_ushort [[LD:v[0-9]+]] ; GCN: {{buffer|flat}}_load_ushort [[LD:v[0-9]+]]
; GCN-DAG: v_cvt_f32_ubyte1_e32 v[[HIRESULT:[0-9]+]], [[LD]] ; GCN-DAG: v_cvt_f32_ubyte1_e32 v[[HIRESULT:[0-9]+]], [[LD]]
; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]], [[LD]] ; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]], [[LD]]
; GCN: buffer_store_dwordx2 v{{\[}}[[LORESULT]]:[[HIRESULT]]{{\]}}, ; GCN: buffer_store_dwordx2 v{{\[}}[[LORESULT]]:[[HIRESULT]]{{\]}},
define amdgpu_kernel void @load_v2i8_to_v2f32(<2 x float> addrspace(1)* noalias %out, <2 x i8> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @load_v2i8_to_v2f32(<2 x float> addrspace(1)* noalias %out, <2 x i8> addrspace(1)* noalias %in) nounwind {
%load = load <2 x i8>, <2 x i8> addrspace(1)* %in, align 2 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <2 x i8>, <2 x i8> addrspace(1)* %in, i32 %tid
%load = load <2 x i8>, <2 x i8> addrspace(1)* %gep, align 2
%cvt = uitofp <2 x i8> %load to <2 x float> %cvt = uitofp <2 x i8> %load to <2 x float>
store <2 x float> %cvt, <2 x float> addrspace(1)* %out, align 16 store <2 x float> %cvt, <2 x float> addrspace(1)* %out, align 16
ret void ret void
} }
; GCN-LABEL: {{^}}load_v3i8_to_v3f32: ; GCN-LABEL: {{^}}load_v3i8_to_v3f32:
; GCN: buffer_load_dword [[VAL:v[0-9]+]] ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; GCN-NOT: v_cvt_f32_ubyte3_e32 ; GCN-NOT: v_cvt_f32_ubyte3_e32
; GCN-DAG: v_cvt_f32_ubyte2_e32 v{{[0-9]+}}, [[VAL]] ; GCN-DAG: v_cvt_f32_ubyte2_e32 v{{[0-9]+}}, [[VAL]]
; GCN-DAG: v_cvt_f32_ubyte1_e32 v[[HIRESULT:[0-9]+]], [[VAL]] ; GCN-DAG: v_cvt_f32_ubyte1_e32 v[[HIRESULT:[0-9]+]], [[VAL]]
; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]], [[VAL]] ; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]], [[VAL]]
; GCN: buffer_store_dwordx2 v{{\[}}[[LORESULT]]:[[HIRESULT]]{{\]}}, ; GCN: buffer_store_dwordx2 v{{\[}}[[LORESULT]]:[[HIRESULT]]{{\]}},
define amdgpu_kernel void @load_v3i8_to_v3f32(<3 x float> addrspace(1)* noalias %out, <3 x i8> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @load_v3i8_to_v3f32(<3 x float> addrspace(1)* noalias %out, <3 x i8> addrspace(1)* noalias %in) nounwind {
%load = load <3 x i8>, <3 x i8> addrspace(1)* %in, align 4 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <3 x i8>, <3 x i8> addrspace(1)* %in, i32 %tid
%load = load <3 x i8>, <3 x i8> addrspace(1)* %gep, align 4
%cvt = uitofp <3 x i8> %load to <3 x float> %cvt = uitofp <3 x i8> %load to <3 x float>
store <3 x float> %cvt, <3 x float> addrspace(1)* %out, align 16 store <3 x float> %cvt, <3 x float> addrspace(1)* %out, align 16
ret void ret void
} }
; GCN-LABEL: {{^}}load_v4i8_to_v4f32: ; GCN-LABEL: {{^}}load_v4i8_to_v4f32:
; GCN: buffer_load_dword [[LOADREG:v[0-9]+]] ; GCN: {{buffer|flat}}_load_dword [[LOADREG:v[0-9]+]]
; GCN-NOT: bfe ; GCN-NOT: bfe
; GCN-NOT: lshr ; GCN-NOT: lshr
; GCN-DAG: v_cvt_f32_ubyte3_e32 v[[HIRESULT:[0-9]+]], [[LOADREG]] ; GCN-DAG: v_cvt_f32_ubyte3_e32 v[[HIRESULT:[0-9]+]], [[LOADREG]]
; GCN-DAG: v_cvt_f32_ubyte2_e32 v{{[0-9]+}}, [[LOADREG]] ; GCN-DAG: v_cvt_f32_ubyte2_e32 v{{[0-9]+}}, [[LOADREG]]
; GCN-DAG: v_cvt_f32_ubyte1_e32 v{{[0-9]+}}, [[LOADREG]] ; GCN-DAG: v_cvt_f32_ubyte1_e32 v{{[0-9]+}}, [[LOADREG]]
; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]], [[LOADREG]] ; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]], [[LOADREG]]
; GCN: buffer_store_dwordx4 v{{\[}}[[LORESULT]]:[[HIRESULT]]{{\]}}, ; GCN: buffer_store_dwordx4 v{{\[}}[[LORESULT]]:[[HIRESULT]]{{\]}},
define amdgpu_kernel void @load_v4i8_to_v4f32(<4 x float> addrspace(1)* noalias %out, <4 x i8> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @load_v4i8_to_v4f32(<4 x float> addrspace(1)* noalias %out, <4 x i8> addrspace(1)* noalias %in) nounwind {
%load = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid
%load = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
%cvt = uitofp <4 x i8> %load to <4 x float> %cvt = uitofp <4 x i8> %load to <4 x float>
store <4 x float> %cvt, <4 x float> addrspace(1)* %out, align 16 store <4 x float> %cvt, <4 x float> addrspace(1)* %out, align 16
ret void ret void
} }
; This should not be adding instructions to shift into the correct ; This should not be adding instructions to shift into the correct
; position in the word for the component. ; position in the word for the component.
; FIXME: Packing bytes ; FIXME: Packing bytes
; GCN-LABEL: {{^}}load_v4i8_to_v4f32_unaligned: ; GCN-LABEL: {{^}}load_v4i8_to_v4f32_unaligned:
; GCN: buffer_load_ubyte [[LOADREG3:v[0-9]+]] ; GCN: {{buffer|flat}}_load_ubyte [[LOADREG3:v[0-9]+]]
; GCN: buffer_load_ubyte [[LOADREG2:v[0-9]+]] ; GCN: {{buffer|flat}}_load_ubyte [[LOADREG2:v[0-9]+]]
; GCN: buffer_load_ubyte [[LOADREG1:v[0-9]+]] ; GCN: {{buffer|flat}}_load_ubyte [[LOADREG1:v[0-9]+]]
; GCN: buffer_load_ubyte [[LOADREG0:v[0-9]+]] ; GCN: {{buffer|flat}}_load_ubyte [[LOADREG0:v[0-9]+]]
; GCN-DAG: v_lshlrev_b32 ; GCN-DAG: v_lshlrev_b32
; GCN-DAG: v_or_b32 ; GCN-DAG: v_or_b32
; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]], ; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]],
; GCN-DAG: v_cvt_f32_ubyte0_e32 v{{[0-9]+}}, ; GCN-DAG: v_cvt_f32_ubyte0_e32 v{{[0-9]+}},
; GCN-DAG: v_cvt_f32_ubyte0_e32 v{{[0-9]+}}, ; GCN-DAG: v_cvt_f32_ubyte0_e32 v{{[0-9]+}},
; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[HIRESULT:[0-9]+]] ; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[HIRESULT:[0-9]+]]
; GCN: buffer_store_dwordx4 ; GCN: buffer_store_dwordx4
define amdgpu_kernel void @load_v4i8_to_v4f32_unaligned(<4 x float> addrspace(1)* noalias %out, <4 x i8> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @load_v4i8_to_v4f32_unaligned(<4 x float> addrspace(1)* noalias %out, <4 x i8> addrspace(1)* noalias %in) nounwind {
%load = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 1 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid
%load = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 1
%cvt = uitofp <4 x i8> %load to <4 x float> %cvt = uitofp <4 x i8> %load to <4 x float>
store <4 x float> %cvt, <4 x float> addrspace(1)* %out, align 16 store <4 x float> %cvt, <4 x float> addrspace(1)* %out, align 16
ret void ret void
} }
; FIXME: Need to handle non-uniform case for function below (load without gep). ; FIXME: Need to handle non-uniform case for function below (load without gep).
; Instructions still emitted to repack bytes for add use. ; Instructions still emitted to repack bytes for add use.
Show All 30 Linesdefine amdgpu_kernel void @load_v4i8_to_v4f32_2_uses(<4 x float> addrspace(1)* noalias %out, <4 x i8> addrspace(1)* noalias %out2, <4 x i8> addrspace(1)* noalias %in) nounwind {
store <4 x i8> %add, <4 x i8> addrspace(1)* %out2, align 4 store <4 x i8> %add, <4 x i8> addrspace(1)* %out2, align 4
ret void ret void
} }
; Make sure this doesn't crash. ; Make sure this doesn't crash.
; GCN-LABEL: {{^}}load_v7i8_to_v7f32: ; GCN-LABEL: {{^}}load_v7i8_to_v7f32:
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @load_v7i8_to_v7f32(<7 x float> addrspace(1)* noalias %out, <7 x i8> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @load_v7i8_to_v7f32(<7 x float> addrspace(1)* noalias %out, <7 x i8> addrspace(1)* noalias %in) nounwind {
%load = load <7 x i8>, <7 x i8> addrspace(1)* %in, align 1 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%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
%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
} }
; GCN-LABEL: {{^}}load_v8i8_to_v8f32: ; GCN-LABEL: {{^}}load_v8i8_to_v8f32:
; GCN: buffer_load_dwordx2 v{{\[}}[[LOLOAD:[0-9]+]]:[[HILOAD:[0-9]+]]{{\]}}, ; GCN: {{buffer|flat}}_load_dwordx2 v{{\[}}[[LOLOAD:[0-9]+]]:[[HILOAD:[0-9]+]]{{\]}},
; GCN-NOT: bfe ; GCN-NOT: bfe
; GCN-NOT: lshr ; GCN-NOT: lshr
; GCN-DAG: v_cvt_f32_ubyte3_e32 v{{[0-9]+}}, v[[LOLOAD]] ; GCN-DAG: v_cvt_f32_ubyte3_e32 v{{[0-9]+}}, v[[LOLOAD]]
; GCN-DAG: v_cvt_f32_ubyte2_e32 v{{[0-9]+}}, v[[LOLOAD]] ; GCN-DAG: v_cvt_f32_ubyte2_e32 v{{[0-9]+}}, v[[LOLOAD]]
; GCN-DAG: v_cvt_f32_ubyte1_e32 v{{[0-9]+}}, v[[LOLOAD]] ; GCN-DAG: v_cvt_f32_ubyte1_e32 v{{[0-9]+}}, v[[LOLOAD]]
; GCN-DAG: v_cvt_f32_ubyte0_e32 v{{[0-9]+}}, v[[LOLOAD]] ; GCN-DAG: v_cvt_f32_ubyte0_e32 v{{[0-9]+}}, v[[LOLOAD]]
; GCN-DAG: v_cvt_f32_ubyte3_e32 v{{[0-9]+}}, v[[HILOAD]] ; GCN-DAG: v_cvt_f32_ubyte3_e32 v{{[0-9]+}}, v[[HILOAD]]
; GCN-DAG: v_cvt_f32_ubyte2_e32 v{{[0-9]+}}, v[[HILOAD]] ; GCN-DAG: v_cvt_f32_ubyte2_e32 v{{[0-9]+}}, v[[HILOAD]]
; GCN-DAG: v_cvt_f32_ubyte1_e32 v{{[0-9]+}}, v[[HILOAD]] ; GCN-DAG: v_cvt_f32_ubyte1_e32 v{{[0-9]+}}, v[[HILOAD]]
; GCN-DAG: v_cvt_f32_ubyte0_e32 v{{[0-9]+}}, v[[HILOAD]] ; GCN-DAG: v_cvt_f32_ubyte0_e32 v{{[0-9]+}}, v[[HILOAD]]
; GCN-NOT: bfe ; GCN-NOT: bfe
; GCN-NOT: lshr ; GCN-NOT: lshr
; GCN: buffer_store_dwordx4 ; GCN: buffer_store_dwordx4
; GCN: buffer_store_dwordx4 ; GCN: buffer_store_dwordx4
define amdgpu_kernel void @load_v8i8_to_v8f32(<8 x float> addrspace(1)* noalias %out, <8 x i8> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @load_v8i8_to_v8f32(<8 x float> addrspace(1)* noalias %out, <8 x i8> addrspace(1)* noalias %in) nounwind {
%load = load <8 x i8>, <8 x i8> addrspace(1)* %in, align 8 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <8 x i8>, <8 x i8> addrspace(1)* %in, i32 %tid
%load = load <8 x i8>, <8 x i8> addrspace(1)* %gep, align 8
%cvt = uitofp <8 x i8> %load to <8 x float> %cvt = uitofp <8 x i8> %load to <8 x float>
store <8 x float> %cvt, <8 x float> addrspace(1)* %out, align 16 store <8 x float> %cvt, <8 x float> addrspace(1)* %out, align 16
ret void ret void
} }
; GCN-LABEL: {{^}}i8_zext_inreg_i32_to_f32: ; GCN-LABEL: {{^}}i8_zext_inreg_i32_to_f32:
; GCN: buffer_load_dword [[LOADREG:v[0-9]+]], ; GCN: {{buffer|flat}}_load_dword [[LOADREG:v[0-9]+]],
; GCN: v_add_i32_e32 [[ADD:v[0-9]+]], vcc, 2, [[LOADREG]] ; GCN: v_add_i32_e32 [[ADD:v[0-9]+]], vcc, 2, [[LOADREG]]
; GCN-NEXT: v_cvt_f32_ubyte0_e32 [[CONV:v[0-9]+]], [[ADD]] ; GCN-NEXT: v_cvt_f32_ubyte0_e32 [[CONV:v[0-9]+]], [[ADD]]
; GCN: buffer_store_dword [[CONV]], ; GCN: buffer_store_dword [[CONV]],
define amdgpu_kernel void @i8_zext_inreg_i32_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @i8_zext_inreg_i32_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
%load = load i32, i32 addrspace(1)* %in, align 4 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%load = load i32, i32 addrspace(1)* %gep, align 4
%add = add i32 %load, 2 %add = add i32 %load, 2
%inreg = and i32 %add, 255 %inreg = and i32 %add, 255
%cvt = uitofp i32 %inreg to float %cvt = uitofp i32 %inreg to float
store float %cvt, float addrspace(1)* %out, align 4 store float %cvt, float addrspace(1)* %out, align 4
ret void ret void
} }
; GCN-LABEL: {{^}}i8_zext_inreg_hi1_to_f32: ; GCN-LABEL: {{^}}i8_zext_inreg_hi1_to_f32:
define amdgpu_kernel void @i8_zext_inreg_hi1_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @i8_zext_inreg_hi1_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
%load = load i32, i32 addrspace(1)* %in, align 4 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%load = load i32, i32 addrspace(1)* %gep, align 4
%inreg = and i32 %load, 65280 %inreg = and i32 %load, 65280
%shr = lshr i32 %inreg, 8 %shr = lshr i32 %inreg, 8
%cvt = uitofp i32 %shr to float %cvt = uitofp i32 %shr to float
store float %cvt, float addrspace(1)* %out, align 4 store float %cvt, float addrspace(1)* %out, align 4
ret void ret void
} }
; We don't get these ones because of the zext, but instcombine removes ; We don't get these ones because of the zext, but instcombine removes
; them so it shouldn't really matter. ; them so it shouldn't really matter.
; GCN-LABEL: {{^}}i8_zext_i32_to_f32: ; GCN-LABEL: {{^}}i8_zext_i32_to_f32:
define amdgpu_kernel void @i8_zext_i32_to_f32(float addrspace(1)* noalias %out, i8 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @i8_zext_i32_to_f32(float addrspace(1)* noalias %out, i8 addrspace(1)* noalias %in) nounwind {
%load = load i8, i8 addrspace(1)* %in, align 1 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i8, i8 addrspace(1)* %in, i32 %tid
%load = load i8, i8 addrspace(1)* %gep, align 1
%ext = zext i8 %load to i32 %ext = zext i8 %load to i32
%cvt = uitofp i32 %ext to float %cvt = uitofp i32 %ext to float
store float %cvt, float addrspace(1)* %out, align 4 store float %cvt, float addrspace(1)* %out, align 4
ret void ret void
} }
; GCN-LABEL: {{^}}v4i8_zext_v4i32_to_v4f32: ; GCN-LABEL: {{^}}v4i8_zext_v4i32_to_v4f32:
define amdgpu_kernel void @v4i8_zext_v4i32_to_v4f32(<4 x float> addrspace(1)* noalias %out, <4 x i8> addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @v4i8_zext_v4i32_to_v4f32(<4 x float> addrspace(1)* noalias %out, <4 x i8> addrspace(1)* noalias %in) nounwind {
%load = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 1 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid
%load = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 1
%ext = zext <4 x i8> %load to <4 x i32> %ext = zext <4 x i8> %load to <4 x i32>
%cvt = uitofp <4 x i32> %ext to <4 x float> %cvt = uitofp <4 x i32> %ext to <4 x float>
store <4 x float> %cvt, <4 x float> addrspace(1)* %out, align 16 store <4 x float> %cvt, <4 x float> addrspace(1)* %out, align 16
ret void ret void
} }
; GCN-LABEL: {{^}}extract_byte0_to_f32: ; GCN-LABEL: {{^}}extract_byte0_to_f32:
; GCN: buffer_load_dword [[VAL:v[0-9]+]] ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; GCN-NOT: [[VAL]] ; GCN-NOT: [[VAL]]
; GCN: v_cvt_f32_ubyte0_e32 [[CONV:v[0-9]+]], [[VAL]] ; GCN: v_cvt_f32_ubyte0_e32 [[CONV:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[CONV]] ; GCN: buffer_store_dword [[CONV]]
define amdgpu_kernel void @extract_byte0_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @extract_byte0_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
%val = load i32, i32 addrspace(1)* %in %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%val = load i32, i32 addrspace(1)* %gep
%and = and i32 %val, 255 %and = and i32 %val, 255
%cvt = uitofp i32 %and to float %cvt = uitofp i32 %and to float
store float %cvt, float addrspace(1)* %out store float %cvt, float addrspace(1)* %out
ret void ret void
} }
; GCN-LABEL: {{^}}extract_byte1_to_f32: ; GCN-LABEL: {{^}}extract_byte1_to_f32:
; GCN: buffer_load_dword [[VAL:v[0-9]+]] ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; GCN-NOT: [[VAL]] ; GCN-NOT: [[VAL]]
; GCN: v_cvt_f32_ubyte1_e32 [[CONV:v[0-9]+]], [[VAL]] ; GCN: v_cvt_f32_ubyte1_e32 [[CONV:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[CONV]] ; GCN: buffer_store_dword [[CONV]]
define amdgpu_kernel void @extract_byte1_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @extract_byte1_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
%val = load i32, i32 addrspace(1)* %in %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%val = load i32, i32 addrspace(1)* %gep
%srl = lshr i32 %val, 8 %srl = lshr i32 %val, 8
%and = and i32 %srl, 255 %and = and i32 %srl, 255
%cvt = uitofp i32 %and to float %cvt = uitofp i32 %and to float
store float %cvt, float addrspace(1)* %out store float %cvt, float addrspace(1)* %out
ret void ret void
} }
; GCN-LABEL: {{^}}extract_byte2_to_f32: ; GCN-LABEL: {{^}}extract_byte2_to_f32:
; GCN: buffer_load_dword [[VAL:v[0-9]+]] ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; GCN-NOT: [[VAL]] ; GCN-NOT: [[VAL]]
; GCN: v_cvt_f32_ubyte2_e32 [[CONV:v[0-9]+]], [[VAL]] ; GCN: v_cvt_f32_ubyte2_e32 [[CONV:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[CONV]] ; GCN: buffer_store_dword [[CONV]]
define amdgpu_kernel void @extract_byte2_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @extract_byte2_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
%val = load i32, i32 addrspace(1)* %in %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%val = load i32, i32 addrspace(1)* %gep
%srl = lshr i32 %val, 16 %srl = lshr i32 %val, 16
%and = and i32 %srl, 255 %and = and i32 %srl, 255
%cvt = uitofp i32 %and to float %cvt = uitofp i32 %and to float
store float %cvt, float addrspace(1)* %out store float %cvt, float addrspace(1)* %out
ret void ret void
} }
; GCN-LABEL: {{^}}extract_byte3_to_f32: ; GCN-LABEL: {{^}}extract_byte3_to_f32:
; GCN: buffer_load_dword [[VAL:v[0-9]+]] ; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; GCN-NOT: [[VAL]] ; GCN-NOT: [[VAL]]
; GCN: v_cvt_f32_ubyte3_e32 [[CONV:v[0-9]+]], [[VAL]] ; GCN: v_cvt_f32_ubyte3_e32 [[CONV:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[CONV]] ; GCN: buffer_store_dword [[CONV]]
define amdgpu_kernel void @extract_byte3_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind { define amdgpu_kernel void @extract_byte3_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
%val = load i32, i32 addrspace(1)* %in %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%val = load i32, i32 addrspace(1)* %gep
%srl = lshr i32 %val, 24 %srl = lshr i32 %val, 24
%and = and i32 %srl, 255 %and = and i32 %srl, 255
%cvt = uitofp i32 %and to float %cvt = uitofp i32 %and to float
store float %cvt, float addrspace(1)* %out store float %cvt, float addrspace(1)* %out
ret void ret void
} }

llvm/trunk/test/CodeGen/AMDGPU/early-if-convert-cost.ll

; RUN: llc -stress-early-ifcvt -amdgpu-early-ifcvt=1 -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -stress-early-ifcvt -amdgpu-early-ifcvt=1 -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; FIXME: Most of these cases that don't trigger because of broken cost ; FIXME: Most of these cases that don't trigger because of broken cost
; heuristics. Should not need -stress-early-ifcvt ; heuristics. Should not need -stress-early-ifcvt
; GCN-LABEL: {{^}}test_vccnz_ifcvt_triangle64: ; GCN-LABEL: {{^}}test_vccnz_ifcvt_triangle64:
; GCN: buffer_load_dwordx2 v{{\[}}[[VAL_LO:[0-9]+]]:[[VAL_HI:[0-9]+]]{{\]}} ; GCN: buffer_load_dwordx2 v{{\[}}[[VAL_LO:[0-9]+]]:[[VAL_HI:[0-9]+]]{{\]}}
; GCN: v_cmp_neq_f64_e32 vcc, 1.0, v{{\[}}[[VAL_LO]]:[[VAL_HI]]{{\]}} ; GCN: v_cmp_neq_f64_e32 vcc, 1.0, v{{\[}}[[VAL_LO]]:[[VAL_HI]]{{\]}}
; GCN: v_add_f64 v{{\[}}[[ADD_LO:[0-9]+]]:[[ADD_HI:[0-9]+]]{{\]}}, v{{\[}}[[VAL_LO]]:[[VAL_HI]]{{\]}}, v{{\[}}[[VAL_LO]]:[[VAL_HI]]{{\]}} ; GCN: v_add_f64 v{{\[}}[[ADD_LO:[0-9]+]]:[[ADD_HI:[0-9]+]]{{\]}}, v{{\[}}[[VAL_LO]]:[[VAL_HI]]{{\]}}, v{{\[}}[[VAL_LO]]:[[VAL_HI]]{{\]}}
▲ Show 20 Lines • Show All 101 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/early-if-convert.ll

; RUN: llc -march=amdgcn -mcpu=verde -amdgpu-early-ifcvt=1 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -amdgpu-early-ifcvt=1 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; XUN: llc -march=amdgcn -mcpu=tonga -amdgpu-early-ifcvt=1 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; XUN: llc -march=amdgcn -mcpu=tonga -amdgpu-early-ifcvt=1 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; FIXME: This leaves behind a now unnecessary and with exec ; FIXME: This leaves behind a now unnecessary and with exec
; GCN-LABEL: {{^}}test_vccnz_ifcvt_triangle: ; GCN-LABEL: {{^}}test_vccnz_ifcvt_triangle:
; GCN: buffer_load_dword [[VAL:v[0-9]+]] ; GCN: buffer_load_dword [[VAL:v[0-9]+]]
; GCN: v_cmp_neq_f32_e32 vcc, 1.0, [[VAL]] ; GCN: v_cmp_neq_f32_e32 vcc, 1.0, [[VAL]]
; GCN: v_add_f32_e32 [[ADD:v[0-9]+]], [[VAL]], [[VAL]] ; GCN: v_add_f32_e32 [[ADD:v[0-9]+]], [[VAL]], [[VAL]]
▲ Show 20 Lines • Show All 445 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/enable-no-signed-zeros-fp-math.ll

; RUN: llc -march=amdgcn -enable-no-signed-zeros-fp-math=0 < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-SAFE %s ; RUN: llc -march=amdgcn -enable-no-signed-zeros-fp-math=0 < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-SAFE %s
; RUN: llc -march=amdgcn -enable-no-signed-zeros-fp-math=1 < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-UNSAFE %s ; RUN: llc -march=amdgcn -enable-no-signed-zeros-fp-math=1 < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-UNSAFE %s
; RUN: llc -march=amdgcn -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-UNSAFE %s ; RUN: llc -march=amdgcn -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-UNSAFE %s
declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
; Test that the -enable-no-signed-zeros-fp-math flag works ; Test that the -enable-no-signed-zeros-fp-math flag works
; GCN-LABEL: {{^}}fneg_fsub_f32: ; GCN-LABEL: {{^}}fneg_fsub_f32:
; GCN: v_subrev_f32_e32 [[SUB:v[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}} ; GCN: v_subrev_f32_e32 [[SUB:v[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}
; GCN-SAFE: v_xor_b32_e32 v{{[0-9]+}}, 0x80000000, [[SUB]] ; GCN-SAFE: v_xor_b32_e32 v{{[0-9]+}}, 0x80000000, [[SUB]]
; GCN-UNSAFE-NOT: xor ; GCN-UNSAFE-NOT: xor
define amdgpu_kernel void @fneg_fsub_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 { define amdgpu_kernel void @fneg_fsub_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
%b_ptr = getelementptr float, float addrspace(1)* %in, i32 1 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%a = load float, float addrspace(1)* %in, align 4 %add = add i32 %tid, 1
%gep = getelementptr float, float addrspace(1)* %in, i32 %tid
%b_ptr = getelementptr float, float addrspace(1)* %in, i32 %add
%a = load float, float addrspace(1)* %gep, align 4
%b = load float, float addrspace(1)* %b_ptr, align 4 %b = load float, float addrspace(1)* %b_ptr, align 4
%result = fsub float %a, %b %result = fsub float %a, %b
%neg.result = fsub float -0.0, %result %neg.result = fsub float -0.0, %result
store float %neg.result, float addrspace(1)* %out, align 4 store float %neg.result, float addrspace(1)* %out, align 4
ret void ret void
} }
attributes #0 = { nounwind } attributes #0 = { nounwind }

llvm/trunk/test/CodeGen/AMDGPU/extractelt-to-trunc.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
; Make sure the add and load are reduced to 32-bits even with the ; Make sure the add and load are reduced to 32-bits even with the
; bitcast to vector. ; bitcast to vector.
; GCN-LABEL: {{^}}bitcast_int_to_vector_extract_0: ; GCN-LABEL: {{^}}bitcast_int_to_vector_extract_0:
; GCN-DAG: s_load_dword [[B:s[0-9]+]] ; GCN-DAG: s_load_dword [[B:s[0-9]+]]
; GCN-DAG: buffer_load_dword [[A:v[0-9]+]] ; GCN-DAG: buffer_load_dword [[A:v[0-9]+]]
; GCN: v_add_i32_e32 [[ADD:v[0-9]+]], vcc, [[B]], [[A]] ; GCN: v_add_i32_e32 [[ADD:v[0-9]+]], vcc, [[B]], [[A]]
; GCN: buffer_store_dword [[ADD]] ; GCN: buffer_store_dword [[ADD]]
define amdgpu_kernel void @bitcast_int_to_vector_extract_0(i32 addrspace(1)* %out, i64 addrspace(1)* %in, i64 %b) { define amdgpu_kernel void @bitcast_int_to_vector_extract_0(i32 addrspace(1)* %out, i64 addrspace(1)* %in, i64 %b) {
%a = load i64, i64 addrspace(1)* %in %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i64, i64 addrspace(1)* %in, i32 %tid
%a = load i64, i64 addrspace(1)* %gep
%add = add i64 %a, %b %add = add i64 %a, %b
%val.bc = bitcast i64 %add to <2 x i32> %val.bc = bitcast i64 %add to <2 x i32>
%extract = extractelement <2 x i32> %val.bc, i32 0 %extract = extractelement <2 x i32> %val.bc, i32 0
store i32 %extract, i32 addrspace(1)* %out store i32 %extract, i32 addrspace(1)* %out
ret void ret void
} }
; GCN-LABEL: {{^}}bitcast_fp_to_vector_extract_0: ; GCN-LABEL: {{^}}bitcast_fp_to_vector_extract_0:
; GCN: buffer_load_dwordx2 ; GCN: buffer_load_dwordx2
; GCN: v_add_f64 ; GCN: v_add_f64
; GCN: buffer_store_dword v ; GCN: buffer_store_dword v
define amdgpu_kernel void @bitcast_fp_to_vector_extract_0(i32 addrspace(1)* %out, double addrspace(1)* %in, double %b) { define amdgpu_kernel void @bitcast_fp_to_vector_extract_0(i32 addrspace(1)* %out, double addrspace(1)* %in, double %b) {
%a = load double, double addrspace(1)* %in %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr double, double addrspace(1)* %in, i32 %tid
%a = load double, double addrspace(1)* %gep
%add = fadd double %a, %b %add = fadd double %a, %b
%val.bc = bitcast double %add to <2 x i32> %val.bc = bitcast double %add to <2 x i32>
%extract = extractelement <2 x i32> %val.bc, i32 0 %extract = extractelement <2 x i32> %val.bc, i32 0
store i32 %extract, i32 addrspace(1)* %out store i32 %extract, i32 addrspace(1)* %out
ret void ret void
} }
; GCN-LABEL: {{^}}bitcast_int_to_fpvector_extract_0: ; GCN-LABEL: {{^}}bitcast_int_to_fpvector_extract_0:
; GCN: buffer_load_dwordx2 ; GCN: buffer_load_dwordx2
; GCN: v_add_i32 ; GCN: v_add_i32
; GCN: buffer_store_dword ; GCN: buffer_store_dword
define amdgpu_kernel void @bitcast_int_to_fpvector_extract_0(float addrspace(1)* %out, i64 addrspace(1)* %in, i64 %b) { define amdgpu_kernel void @bitcast_int_to_fpvector_extract_0(float addrspace(1)* %out, i64 addrspace(1)* %in, i64 %b) {
%a = load i64, i64 addrspace(1)* %in %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i64, i64 addrspace(1)* %in, i32 %tid
%a = load i64, i64 addrspace(1)* %gep
%add = add i64 %a, %b %add = add i64 %a, %b
%val.bc = bitcast i64 %add to <2 x float> %val.bc = bitcast i64 %add to <2 x float>
%extract = extractelement <2 x float> %val.bc, i32 0 %extract = extractelement <2 x float> %val.bc, i32 0
store float %extract, float addrspace(1)* %out store float %extract, float addrspace(1)* %out
ret void ret void
} }
; GCN-LABEL: {{^}}no_extract_volatile_load_extract0: ; GCN-LABEL: {{^}}no_extract_volatile_load_extract0:
Show All 32 Lines

llvm/trunk/test/CodeGen/AMDGPU/fabs.f16.ll

Show First 20 Lines • Show All 128 Lines • ▼ Show 20 Lines
; VI: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, ; VI: v_lshrrev_b32_e32 v{{[0-9]+}}, 16,
; VI: v_mul_f16_sdwa v{{[0-9]+}}, |v{{[0-9]+}}|, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; VI: v_mul_f16_sdwa v{{[0-9]+}}, |v{{[0-9]+}}|, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI: v_mul_f16_e64 v{{[0-9]+}}, |v{{[0-9]+}}|, v{{[0-9]+}} ; VI: v_mul_f16_e64 v{{[0-9]+}}, |v{{[0-9]+}}|, v{{[0-9]+}}
; GFX9: v_and_b32_e32 [[FABS:v[0-9]+]], 0x7fff7fff, [[VAL]] ; GFX9: v_and_b32_e32 [[FABS:v[0-9]+]], 0x7fff7fff, [[VAL]]
; GFX9: v_pk_mul_f16 v{{[0-9]+}}, [[FABS]], v{{[0-9]+$}} ; GFX9: v_pk_mul_f16 v{{[0-9]+}}, [[FABS]], v{{[0-9]+$}}
define amdgpu_kernel void @v_fabs_fold_v2f16(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 { define amdgpu_kernel void @v_fabs_fold_v2f16(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
%val = load <2 x half>, <2 x half> addrspace(1)* %in %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <2 x half>, <2 x half> addrspace(1)* %in, i32 %tid
%val = load <2 x half>, <2 x half> addrspace(1)* %gep
%fabs = call <2 x half> @llvm.fabs.v2f16(<2 x half> %val) %fabs = call <2 x half> @llvm.fabs.v2f16(<2 x half> %val)
%fmul = fmul <2 x half> %fabs, %val %fmul = fmul <2 x half> %fabs, %val
store <2 x half> %fmul, <2 x half> addrspace(1)* %out store <2 x half> %fmul, <2 x half> addrspace(1)* %out
ret void ret void
} }
declare half @llvm.fabs.f16(half) #1 declare half @llvm.fabs.f16(half) #1
declare <2 x half> @llvm.fabs.v2f16(<2 x half>) #1 declare <2 x half> @llvm.fabs.v2f16(<2 x half>) #1
declare <4 x half> @llvm.fabs.v4f16(<4 x half>) #1 declare <4 x half> @llvm.fabs.v4f16(<4 x half>) #1
declare i32 @llvm.amdgcn.workitem.id.x() #1 declare i32 @llvm.amdgcn.workitem.id.x() #1
attributes #0 = { nounwind } attributes #0 = { nounwind }
attributes #1 = { nounwind readnone } attributes #1 = { nounwind readnone }

llvm/trunk/test/CodeGen/AMDGPU/fadd.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}fadd_f16 ; GCN-LABEL: {{^}}fadd_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: {{buffer|flat}}_load_ushort v[[A_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[B_F16:[0-9]+]] ; GCN: {{buffer|flat}}_load_ushort v[[B_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]] ; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]]
; SI: v_add_f32_e32 v[[R_F32:[0-9]+]], v[[B_F32]], v[[A_F32]] ; SI: v_add_f32_e32 v[[R_F32:[0-9]+]], v[[B_F32]], v[[A_F32]]
; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]] ; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]]
; VI: v_add_f16_e32 v[[R_F16:[0-9]+]], v[[B_F16]], v[[A_F16]] ; VI: v_add_f16_e32 v[[R_F16:[0-9]+]], v[[B_F16]], v[[A_F16]]
; GCN: buffer_store_short v[[R_F16]] ; GCN: buffer_store_short v[[R_F16]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @fadd_f16( define amdgpu_kernel void @fadd_f16(
half addrspace(1)* %r, half addrspace(1)* %r,
half addrspace(1)* %a, half addrspace(1)* %a,
half addrspace(1)* %b) { half addrspace(1)* %b) {
entry: entry:
%a.val = load half, half addrspace(1)* %a %a.val = load half, half addrspace(1)* %a
%b.val = load half, half addrspace(1)* %b %b.val = load half, half addrspace(1)* %b
%r.val = fadd half %a.val, %b.val %r.val = fadd half %a.val, %b.val
store half %r.val, half addrspace(1)* %r store half %r.val, half addrspace(1)* %r
ret void ret void
} }
; GCN-LABEL: {{^}}fadd_f16_imm_a ; GCN-LABEL: {{^}}fadd_f16_imm_a
; GCN: buffer_load_ushort v[[B_F16:[0-9]+]] ; GCN: {{buffer|flat}}_load_ushort v[[B_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]] ; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]]
; SI: v_add_f32_e32 v[[R_F32:[0-9]+]], 1.0, v[[B_F32]] ; SI: v_add_f32_e32 v[[R_F32:[0-9]+]], 1.0, v[[B_F32]]
; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]] ; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]]
; VI: v_add_f16_e32 v[[R_F16:[0-9]+]], 1.0, v[[B_F16]] ; VI: v_add_f16_e32 v[[R_F16:[0-9]+]], 1.0, v[[B_F16]]
; GCN: buffer_store_short v[[R_F16]] ; GCN: buffer_store_short v[[R_F16]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @fadd_f16_imm_a( define amdgpu_kernel void @fadd_f16_imm_a(
half addrspace(1)* %r, half addrspace(1)* %r,
half addrspace(1)* %b) { half addrspace(1)* %b) {
entry: entry:
%b.val = load half, half addrspace(1)* %b %b.val = load half, half addrspace(1)* %b
%r.val = fadd half 1.0, %b.val %r.val = fadd half 1.0, %b.val
store half %r.val, half addrspace(1)* %r store half %r.val, half addrspace(1)* %r
ret void ret void
} }
; GCN-LABEL: {{^}}fadd_f16_imm_b ; GCN-LABEL: {{^}}fadd_f16_imm_b
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: {{buffer|flat}}_load_ushort v[[A_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_add_f32_e32 v[[R_F32:[0-9]+]], 2.0, v[[A_F32]] ; SI: v_add_f32_e32 v[[R_F32:[0-9]+]], 2.0, v[[A_F32]]
; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]] ; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]]
; VI: v_add_f16_e32 v[[R_F16:[0-9]+]], 2.0, v[[A_F16]] ; VI: v_add_f16_e32 v[[R_F16:[0-9]+]], 2.0, v[[A_F16]]
; GCN: buffer_store_short v[[R_F16]] ; GCN: buffer_store_short v[[R_F16]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @fadd_f16_imm_b( define amdgpu_kernel void @fadd_f16_imm_b(
half addrspace(1)* %r, half addrspace(1)* %r,
half addrspace(1)* %a) { half addrspace(1)* %a) {
entry: entry:
%a.val = load half, half addrspace(1)* %a %a.val = load half, half addrspace(1)* %a
%r.val = fadd half %a.val, 2.0 %r.val = fadd half %a.val, 2.0
store half %r.val, half addrspace(1)* %r store half %r.val, half addrspace(1)* %r
ret void ret void
} }
; GCN-LABEL: {{^}}fadd_v2f16: ; GCN-LABEL: {{^}}fadd_v2f16:
; GCN: buffer_load_dword v[[A_V2_F16:[0-9]+]] ; GCN: {{buffer|flat}}_load_dword v[[A_V2_F16:[0-9]+]]
; GCN: buffer_load_dword v[[B_V2_F16:[0-9]+]] ; GCN: {{buffer|flat}}_load_dword v[[B_V2_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_V2_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_V2_F16]]
; SI: v_lshrrev_b32_e32 v[[A_F16_1:[0-9]+]], 16, v[[A_V2_F16]] ; SI: v_lshrrev_b32_e32 v[[A_F16_1:[0-9]+]], 16, v[[A_V2_F16]]
; SI: v_cvt_f32_f16_e32 v[[B_F32_0:[0-9]+]], v[[B_V2_F16]] ; SI: v_cvt_f32_f16_e32 v[[B_F32_0:[0-9]+]], v[[B_V2_F16]]
; SI: v_lshrrev_b32_e32 v[[B_F16_1:[0-9]+]], 16, v[[B_V2_F16]] ; SI: v_lshrrev_b32_e32 v[[B_F16_1:[0-9]+]], 16, v[[B_V2_F16]]
; SI-DAG: v_cvt_f32_f16_e32 v[[A_F32_1:[0-9]+]], v[[A_F16_1]] ; SI-DAG: v_cvt_f32_f16_e32 v[[A_F32_1:[0-9]+]], v[[A_F16_1]]
; SI-DAG: v_cvt_f32_f16_e32 v[[B_F32_1:[0-9]+]], v[[B_F16_1]] ; SI-DAG: v_cvt_f32_f16_e32 v[[B_F32_1:[0-9]+]], v[[B_F16_1]]
; SI: v_add_f32_e32 v[[R_F32_0:[0-9]+]], v[[B_F32_0]], v[[A_F32_0]] ; SI-DAG: v_add_f32_e32 v[[R_F32_0:[0-9]+]], v[[B_F32_0]], v[[A_F32_0]]
; SI: v_add_f32_e32 v[[R_F32_1:[0-9]+]], v[[B_F32_1]], v[[A_F32_1]] ; SI-DAG: v_add_f32_e32 v[[R_F32_1:[0-9]+]], v[[B_F32_1]], v[[A_F32_1]]
; SI-DAG: v_cvt_f16_f32_e32 v[[R_F16_0:[0-9]+]], v[[R_F32_0]] ; SI-DAG: v_cvt_f16_f32_e32 v[[R_F16_0:[0-9]+]], v[[R_F32_0]]
; SI-DAG: v_cvt_f16_f32_e32 v[[R_F16_1:[0-9]+]], v[[R_F32_1]] ; SI-DAG: v_cvt_f16_f32_e32 v[[R_F16_1:[0-9]+]], v[[R_F32_1]]
; SI: v_lshlrev_b32_e32 v[[R_F16_HI:[0-9]+]], 16, v[[R_F16_1]] ; SI: v_lshlrev_b32_e32 v[[R_F16_HI:[0-9]+]], 16, v[[R_F16_1]]
; SI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_0]] ; SI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_0]]
; VI-DAG: v_add_f16_e32 v[[R_F16_LO:[0-9]+]], v[[B_V2_F16]], v[[A_V2_F16]] ; VI-DAG: v_add_f16_e32 v[[R_F16_LO:[0-9]+]], v[[B_V2_F16]], v[[A_V2_F16]]
; VI-DAG: v_add_f16_sdwa v[[R_F16_HI:[0-9]+]], v[[A_V2_F16]], v[[B_V2_F16]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 ; VI-DAG: v_add_f16_sdwa v[[R_F16_HI:[0-9]+]], v[[A_V2_F16]], v[[B_V2_F16]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
; VI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_LO]] ; VI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_LO]]
; GCN: buffer_store_dword v[[R_V2_F16]] ; GCN: buffer_store_dword v[[R_V2_F16]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @fadd_v2f16( define amdgpu_kernel void @fadd_v2f16(
<2 x half> addrspace(1)* %r, <2 x half> addrspace(1)* %r,
<2 x half> addrspace(1)* %a, <2 x half> addrspace(1)* %a,
<2 x half> addrspace(1)* %b) { <2 x half> addrspace(1)* %b) {
entry: entry:
%a.val = load <2 x half>, <2 x half> addrspace(1)* %a %tid = call i32 @llvm.amdgcn.workitem.id.x()
%b.val = load <2 x half>, <2 x half> addrspace(1)* %b %gep.a = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %a, i32 %tid
%gep.b = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %b, i32 %tid
%a.val = load <2 x half>, <2 x half> addrspace(1)* %gep.a
%b.val = load <2 x half>, <2 x half> addrspace(1)* %gep.b
%r.val = fadd <2 x half> %a.val, %b.val %r.val = fadd <2 x half> %a.val, %b.val
store <2 x half> %r.val, <2 x half> addrspace(1)* %r store <2 x half> %r.val, <2 x half> addrspace(1)* %r
ret void ret void
} }
; GCN-LABEL: {{^}}fadd_v2f16_imm_a: ; GCN-LABEL: {{^}}fadd_v2f16_imm_a:
; GCN-DAG: buffer_load_dword v[[B_V2_F16:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_dword v[[B_V2_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[B_F32_0:[0-9]+]], v[[B_V2_F16]] ; SI: v_cvt_f32_f16_e32 v[[B_F32_0:[0-9]+]], v[[B_V2_F16]]
; SI: v_lshrrev_b32_e32 v[[B_F16_1:[0-9]+]], 16, v[[B_V2_F16]] ; SI: v_lshrrev_b32_e32 v[[B_F16_1:[0-9]+]], 16, v[[B_V2_F16]]
; SI: v_cvt_f32_f16_e32 v[[B_F32_1:[0-9]+]], v[[B_F16_1]] ; SI: v_cvt_f32_f16_e32 v[[B_F32_1:[0-9]+]], v[[B_F16_1]]
; SI: v_add_f32_e32 v[[R_F32_0:[0-9]+]], 1.0, v[[B_F32_0]] ; SI: v_add_f32_e32 v[[R_F32_0:[0-9]+]], 1.0, v[[B_F32_0]]
; SI: v_cvt_f16_f32_e32 v[[R_F16_0:[0-9]+]], v[[R_F32_0]] ; SI: v_cvt_f16_f32_e32 v[[R_F16_0:[0-9]+]], v[[R_F32_0]]
; SI: v_add_f32_e32 v[[R_F32_1:[0-9]+]], 2.0, v[[B_F32_1]] ; SI: v_add_f32_e32 v[[R_F32_1:[0-9]+]], 2.0, v[[B_F32_1]]
; SI: v_cvt_f16_f32_e32 v[[R_F16_1:[0-9]+]], v[[R_F32_1]] ; SI: v_cvt_f16_f32_e32 v[[R_F16_1:[0-9]+]], v[[R_F32_1]]
; SI-DAG: v_lshlrev_b32_e32 v[[R_F16_HI:[0-9]+]], 16, v[[R_F16_1]] ; SI-DAG: v_lshlrev_b32_e32 v[[R_F16_HI:[0-9]+]], 16, v[[R_F16_1]]
; SI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_0]] ; SI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_0]]
; VI-DAG: v_mov_b32_e32 v[[CONST2:[0-9]+]], 0x4000 ; VI-DAG: v_mov_b32_e32 v[[CONST2:[0-9]+]], 0x4000
; VI-DAG: v_add_f16_sdwa v[[R_F16_HI:[0-9]+]], v[[B_V2_F16]], v[[CONST2]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; VI-DAG: v_add_f16_sdwa v[[R_F16_HI:[0-9]+]], v[[B_V2_F16]], v[[CONST2]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-DAG: v_add_f16_e32 v[[R_F16_0:[0-9]+]], 1.0, v[[B_V2_F16]] ; VI-DAG: v_add_f16_e32 v[[R_F16_0:[0-9]+]], 1.0, v[[B_V2_F16]]
; VI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_0]] ; VI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_0]]
; GCN: buffer_store_dword v[[R_V2_F16]] ; GCN: buffer_store_dword v[[R_V2_F16]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @fadd_v2f16_imm_a( define amdgpu_kernel void @fadd_v2f16_imm_a(
<2 x half> addrspace(1)* %r, <2 x half> addrspace(1)* %r,
<2 x half> addrspace(1)* %b) { <2 x half> addrspace(1)* %b) {
entry: entry:
%b.val = load <2 x half>, <2 x half> addrspace(1)* %b %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep.b = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %b, i32 %tid
%b.val = load <2 x half>, <2 x half> addrspace(1)* %gep.b
%r.val = fadd <2 x half> <half 1.0, half 2.0>, %b.val %r.val = fadd <2 x half> <half 1.0, half 2.0>, %b.val
store <2 x half> %r.val, <2 x half> addrspace(1)* %r store <2 x half> %r.val, <2 x half> addrspace(1)* %r
ret void ret void
} }
; GCN-LABEL: {{^}}fadd_v2f16_imm_b: ; GCN-LABEL: {{^}}fadd_v2f16_imm_b:
; GCN-DAG: buffer_load_dword v[[A_V2_F16:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_dword v[[A_V2_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_V2_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_V2_F16]]
; SI-DAG: v_lshrrev_b32_e32 v[[A_F16_1:[0-9]+]], 16, v[[A_V2_F16]] ; SI-DAG: v_lshrrev_b32_e32 v[[A_F16_1:[0-9]+]], 16, v[[A_V2_F16]]
; SI: v_cvt_f32_f16_e32 v[[A_F32_1:[0-9]+]], v[[A_F16_1]] ; SI: v_cvt_f32_f16_e32 v[[A_F32_1:[0-9]+]], v[[A_F16_1]]
; SI: v_add_f32_e32 v[[R_F32_0:[0-9]+]], 2.0, v[[A_F32_0]] ; SI: v_add_f32_e32 v[[R_F32_0:[0-9]+]], 2.0, v[[A_F32_0]]
; SI: v_cvt_f16_f32_e32 v[[R_F16_0:[0-9]+]], v[[R_F32_0]] ; SI: v_cvt_f16_f32_e32 v[[R_F16_0:[0-9]+]], v[[R_F32_0]]
; SI: v_add_f32_e32 v[[R_F32_1:[0-9]+]], 1.0, v[[A_F32_1]] ; SI: v_add_f32_e32 v[[R_F32_1:[0-9]+]], 1.0, v[[A_F32_1]]
; SI: v_cvt_f16_f32_e32 v[[R_F16_1:[0-9]+]], v[[R_F32_1]] ; SI: v_cvt_f16_f32_e32 v[[R_F16_1:[0-9]+]], v[[R_F32_1]]
; SI-DAG: v_lshlrev_b32_e32 v[[R_F16_HI:[0-9]+]], 16, v[[R_F16_1]] ; SI-DAG: v_lshlrev_b32_e32 v[[R_F16_HI:[0-9]+]], 16, v[[R_F16_1]]
; SI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_0]] ; SI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_HI]], v[[R_F16_0]]
; VI-DAG: v_mov_b32_e32 v[[CONST1:[0-9]+]], 0x3c00 ; VI-DAG: v_mov_b32_e32 v[[CONST1:[0-9]+]], 0x3c00
; VI-DAG: v_add_f16_sdwa v[[R_F16_0:[0-9]+]], v[[A_V2_F16]], v[[CONST1]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; VI-DAG: v_add_f16_sdwa v[[R_F16_0:[0-9]+]], v[[A_V2_F16]], v[[CONST1]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-DAG: v_add_f16_e32 v[[R_F16_1:[0-9]+]], 2.0, v[[A_V2_F16]] ; VI-DAG: v_add_f16_e32 v[[R_F16_1:[0-9]+]], 2.0, v[[A_V2_F16]]
; VI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_0]], v[[R_F16_1]] ; VI: v_or_b32_e32 v[[R_V2_F16:[0-9]+]], v[[R_F16_0]], v[[R_F16_1]]
; GCN: buffer_store_dword v[[R_V2_F16]] ; GCN: buffer_store_dword v[[R_V2_F16]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @fadd_v2f16_imm_b( define amdgpu_kernel void @fadd_v2f16_imm_b(
<2 x half> addrspace(1)* %r, <2 x half> addrspace(1)* %r,
<2 x half> addrspace(1)* %a) { <2 x half> addrspace(1)* %a) {
entry: entry:
%a.val = load <2 x half>, <2 x half> addrspace(1)* %a %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep.a = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %a, i32 %tid
%a.val = load <2 x half>, <2 x half> addrspace(1)* %gep.a
%r.val = fadd <2 x half> %a.val, <half 2.0, half 1.0> %r.val = fadd <2 x half> %a.val, <half 2.0, half 1.0>
store <2 x half> %r.val, <2 x half> addrspace(1)* %r store <2 x half> %r.val, <2 x half> addrspace(1)* %r
ret void ret void
} }
declare i32 @llvm.amdgcn.workitem.id.x() #1
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }

llvm/trunk/test/CodeGen/AMDGPU/fadd64.ll

; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck %s ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck %s ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck %s
; CHECK-LABEL: {{^}}v_fadd_f64: ; CHECK-LABEL: {{^}}v_fadd_f64:
; CHECK: v_add_f64 {{v[[0-9]+:[0-9]+]}}, {{v[[0-9]+:[0-9]+]}}, {{v[[0-9]+:[0-9]+]}} ; CHECK: v_add_f64 {{v[[0-9]+:[0-9]+]}}, {{v[[0-9]+:[0-9]+]}}, {{v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @v_fadd_f64(double addrspace(1)* %out, double addrspace(1)* %in1, define amdgpu_kernel void @v_fadd_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) { double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1 %tid = call i32 @llvm.amdgcn.workitem.id.x()
%r1 = load double, double addrspace(1)* %in2 %gep1 = getelementptr inbounds double, double addrspace(1)* %in1, i32 %tid
%gep2 = getelementptr inbounds double, double addrspace(1)* %in2, i32 %tid
%r0 = load double, double addrspace(1)* %gep1
%r1 = load double, double addrspace(1)* %gep2
%r2 = fadd double %r0, %r1 %r2 = fadd double %r0, %r1
store double %r2, double addrspace(1)* %out store double %r2, double addrspace(1)* %out
ret void ret void
} }
; CHECK-LABEL: {{^}}s_fadd_f64: ; CHECK-LABEL: {{^}}s_fadd_f64:
; CHECK: v_add_f64 {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}} ; CHECK: v_add_f64 {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}}
define amdgpu_kernel void @s_fadd_f64(double addrspace(1)* %out, double %r0, double %r1) { define amdgpu_kernel void @s_fadd_f64(double addrspace(1)* %out, double %r0, double %r1) {
Show All 19 Lines
; CHECK: v_add_f64 {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}} ; CHECK: v_add_f64 {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}}
; CHECK: v_add_f64 {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}} ; CHECK: v_add_f64 {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}}
; CHECK: _store_dwordx4 ; CHECK: _store_dwordx4
define amdgpu_kernel void @s_fadd_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %r0, <2 x double> %r1) { define amdgpu_kernel void @s_fadd_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %r0, <2 x double> %r1) {
%r2 = fadd <2 x double> %r0, %r1 %r2 = fadd <2 x double> %r0, %r1
store <2 x double> %r2, <2 x double> addrspace(1)* %out store <2 x double> %r2, <2 x double> addrspace(1)* %out
ret void ret void
} }
declare i32 @llvm.amdgcn.workitem.id.x() #1
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }

llvm/trunk/test/CodeGen/AMDGPU/fcanonicalize.f16.ll

; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s ; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
declare half @llvm.fabs.f16(half) #0 declare half @llvm.fabs.f16(half) #0
declare half @llvm.canonicalize.f16(half) #0 declare half @llvm.canonicalize.f16(half) #0
declare <2 x half> @llvm.fabs.v2f16(<2 x half>) #0 declare <2 x half> @llvm.fabs.v2f16(<2 x half>) #0
declare <2 x half> @llvm.canonicalize.v2f16(<2 x half>) #0 declare <2 x half> @llvm.canonicalize.v2f16(<2 x half>) #0
declare i32 @llvm.amdgcn.workitem.id.x() #0
; GCN-LABEL: {{^}}v_test_canonicalize_var_f16: ; GCN-LABEL: {{^}}v_test_canonicalize_var_f16:
; GCN: v_mul_f16_e32 [[REG:v[0-9]+]], 1.0, {{v[0-9]+}} ; GCN: v_mul_f16_e32 [[REG:v[0-9]+]], 1.0, {{v[0-9]+}}
; GCN: buffer_store_short [[REG]] ; GCN: buffer_store_short [[REG]]
define amdgpu_kernel void @v_test_canonicalize_var_f16(half addrspace(1)* %out) #1 { define amdgpu_kernel void @v_test_canonicalize_var_f16(half addrspace(1)* %out) #1 {
%val = load half, half addrspace(1)* %out %val = load half, half addrspace(1)* %out
%canonicalized = call half @llvm.canonicalize.f16(half %val) %canonicalized = call half @llvm.canonicalize.f16(half %val)
store half %canonicalized, half addrspace(1)* %out store half %canonicalized, half addrspace(1)* %out
▲ Show 20 Lines • Show All 192 Lines • ▼ Show 20 Lines
; VI: v_mov_b32_e32 v[[CONST1:[0-9]+]], 0x3c00 ; VI: v_mov_b32_e32 v[[CONST1:[0-9]+]], 0x3c00
; VI-DAG: v_mul_f16_sdwa [[REG0:v[0-9]+]], v[[CONST1]], {{v[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; VI-DAG: v_mul_f16_sdwa [[REG0:v[0-9]+]], v[[CONST1]], {{v[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; VI-DAG: v_mul_f16_e32 [[REG1:v[0-9]+]], 1.0, {{v[0-9]+}} ; VI-DAG: v_mul_f16_e32 [[REG1:v[0-9]+]], 1.0, {{v[0-9]+}}
; VI-NOT: v_and_b32 ; VI-NOT: v_and_b32
; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, {{v[0-9]+$}} ; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, {{v[0-9]+$}}
; GFX9: buffer_store_dword [[REG]] ; GFX9: buffer_store_dword [[REG]]
define amdgpu_kernel void @v_test_canonicalize_var_v2f16(<2 x half> addrspace(1)* %out) #1 { define amdgpu_kernel void @v_test_canonicalize_var_v2f16(<2 x half> addrspace(1)* %out) #1 {
%val = load <2 x half>, <2 x half> addrspace(1)* %out %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <2 x half>, <2 x half> addrspace(1)* %out, i32 %tid
%val = load <2 x half>, <2 x half> addrspace(1)* %gep
%canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %val) %canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %val)
store <2 x half> %canonicalized, <2 x half> addrspace(1)* %out store <2 x half> %canonicalized, <2 x half> addrspace(1)* %out
ret void ret void
} }
; FIXME: Fold modifier ; FIXME: Fold modifier
; GCN-LABEL: {{^}}v_test_canonicalize_fabs_var_v2f16: ; GCN-LABEL: {{^}}v_test_canonicalize_fabs_var_v2f16:
; VI-DAG: v_bfe_u32 ; VI-DAG: v_bfe_u32
; VI-DAG: v_and_b32_e32 v{{[0-9]+}}, 0x7fff7fff, v{{[0-9]+}} ; VI-DAG: v_and_b32_e32 v{{[0-9]+}}, 0x7fff7fff, v{{[0-9]+}}
; VI-DAG: v_mov_b32_e32 v[[CONST1:[0-9]+]], 0x3c00 ; VI-DAG: v_mov_b32_e32 v[[CONST1:[0-9]+]], 0x3c00
; VI: v_mul_f16_sdwa [[REG0:v[0-9]+]], v[[CONST1]], v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD ; VI: v_mul_f16_sdwa [[REG0:v[0-9]+]], v[[CONST1]], v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; VI: v_mul_f16_e32 [[REG1:v[0-9]+]], 1.0, v{{[0-9]+}} ; VI: v_mul_f16_e32 [[REG1:v[0-9]+]], 1.0, v{{[0-9]+}}
; VI-NOT: 0xffff ; VI-NOT: 0xffff
; VI: v_or_b32 ; VI: v_or_b32
; GFX9: v_and_b32_e32 [[ABS:v[0-9]+]], 0x7fff7fff, v{{[0-9]+}} ; GFX9: v_and_b32_e32 [[ABS:v[0-9]+]], 0x7fff7fff, v{{[0-9]+}}
; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, [[ABS]]{{$}} ; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, [[ABS]]{{$}}
; GCN: buffer_store_dword ; GCN: buffer_store_dword
define amdgpu_kernel void @v_test_canonicalize_fabs_var_v2f16(<2 x half> addrspace(1)* %out) #1 { define amdgpu_kernel void @v_test_canonicalize_fabs_var_v2f16(<2 x half> addrspace(1)* %out) #1 {
%val = load <2 x half>, <2 x half> addrspace(1)* %out %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <2 x half>, <2 x half> addrspace(1)* %out, i32 %tid
%val = load <2 x half>, <2 x half> addrspace(1)* %gep
%val.fabs = call <2 x half> @llvm.fabs.v2f16(<2 x half> %val) %val.fabs = call <2 x half> @llvm.fabs.v2f16(<2 x half> %val)
%canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %val.fabs) %canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %val.fabs)
store <2 x half> %canonicalized, <2 x half> addrspace(1)* %out store <2 x half> %canonicalized, <2 x half> addrspace(1)* %out
ret void ret void
} }
; GCN-LABEL: {{^}}v_test_canonicalize_fneg_fabs_var_v2f16: ; GCN-LABEL: {{^}}v_test_canonicalize_fneg_fabs_var_v2f16:
; VI-DAG: v_mov_b32_e32 v[[CONST1:[0-9]+]], 0x3c00 ; VI-DAG: v_mov_b32_e32 v[[CONST1:[0-9]+]], 0x3c00
; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, 0x80008000, v{{[0-9]+}} ; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, 0x80008000, v{{[0-9]+}}
; VI-DAG: v_mul_f16_sdwa [[REG0:v[0-9]+]], v[[CONST1]], v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; VI-DAG: v_mul_f16_sdwa [[REG0:v[0-9]+]], v[[CONST1]], v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; VI-DAG: v_mul_f16_e32 [[REG1:v[0-9]+]], 1.0, v{{[0-9]+}} ; VI-DAG: v_mul_f16_e32 [[REG1:v[0-9]+]], 1.0, v{{[0-9]+}}
; VI: v_or_b32 ; VI: v_or_b32
; GFX9: v_and_b32_e32 [[ABS:v[0-9]+]], 0x7fff7fff, v{{[0-9]+}} ; GFX9: v_and_b32_e32 [[ABS:v[0-9]+]], 0x7fff7fff, v{{[0-9]+}}
; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, [[ABS]] neg_lo:[0,1] neg_hi:[0,1]{{$}} ; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, [[ABS]] neg_lo:[0,1] neg_hi:[0,1]{{$}}
; GCN: buffer_store_dword ; GCN: buffer_store_dword
define amdgpu_kernel void @v_test_canonicalize_fneg_fabs_var_v2f16(<2 x half> addrspace(1)* %out) #1 { define amdgpu_kernel void @v_test_canonicalize_fneg_fabs_var_v2f16(<2 x half> addrspace(1)* %out) #1 {
%val = load <2 x half>, <2 x half> addrspace(1)* %out %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <2 x half>, <2 x half> addrspace(1)* %out, i32 %tid
%val = load <2 x half>, <2 x half> addrspace(1)* %gep
%val.fabs = call <2 x half> @llvm.fabs.v2f16(<2 x half> %val) %val.fabs = call <2 x half> @llvm.fabs.v2f16(<2 x half> %val)
%val.fabs.fneg = fsub <2 x half> <half -0.0, half -0.0>, %val.fabs %val.fabs.fneg = fsub <2 x half> <half -0.0, half -0.0>, %val.fabs
%canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %val.fabs.fneg) %canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %val.fabs.fneg)
store <2 x half> %canonicalized, <2 x half> addrspace(1)* %out store <2 x half> %canonicalized, <2 x half> addrspace(1)* %out
ret void ret void
} }
; FIXME: Fold modifier ; FIXME: Fold modifier
; GCN-LABEL: {{^}}v_test_canonicalize_fneg_var_v2f16: ; GCN-LABEL: {{^}}v_test_canonicalize_fneg_var_v2f16:
; VI-DAG: v_mov_b32_e32 v[[CONST1:[0-9]+]], 0x3c00 ; VI-DAG: v_mov_b32_e32 v[[CONST1:[0-9]+]], 0x3c00
; VI-DAG: v_xor_b32_e32 [[FNEG:v[0-9]+]], 0x80008000, v{{[0-9]+}} ; VI-DAG: v_xor_b32_e32 [[FNEG:v[0-9]+]], 0x80008000, v{{[0-9]+}}
; VI-DAG: v_mul_f16_sdwa [[REG1:v[0-9]+]], v[[CONST1]], [[FNEG]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; VI-DAG: v_mul_f16_sdwa [[REG1:v[0-9]+]], v[[CONST1]], [[FNEG]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; VI-DAG: v_mul_f16_e32 [[REG0:v[0-9]+]], 1.0, [[FNEG]] ; VI-DAG: v_mul_f16_e32 [[REG0:v[0-9]+]], 1.0, [[FNEG]]
; VI-NOT: 0xffff ; VI-NOT: 0xffff
; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, {{v[0-9]+}} neg_lo:[0,1] neg_hi:[0,1]{{$}} ; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, {{v[0-9]+}} neg_lo:[0,1] neg_hi:[0,1]{{$}}
; GFX9: buffer_store_dword [[REG]] ; GFX9: buffer_store_dword [[REG]]
define amdgpu_kernel void @v_test_canonicalize_fneg_var_v2f16(<2 x half> addrspace(1)* %out) #1 { define amdgpu_kernel void @v_test_canonicalize_fneg_var_v2f16(<2 x half> addrspace(1)* %out) #1 {
%val = load <2 x half>, <2 x half> addrspace(1)* %out %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr <2 x half>, <2 x half> addrspace(1)* %out, i32 %tid
%val = load <2 x half>, <2 x half> addrspace(1)* %gep
%fneg.val = fsub <2 x half> <half -0.0, half -0.0>, %val %fneg.val = fsub <2 x half> <half -0.0, half -0.0>, %val
%canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %fneg.val) %canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %fneg.val)
store <2 x half> %canonicalized, <2 x half> addrspace(1)* %out store <2 x half> %canonicalized, <2 x half> addrspace(1)* %out
ret void ret void
} }
; GCN-LABEL: {{^}}s_test_canonicalize_var_v2f16: ; GCN-LABEL: {{^}}s_test_canonicalize_var_v2f16:
; VI: v_mov_b32_e32 [[ONE:v[0-9]+]], 0x3c00 ; VI: v_mov_b32_e32 [[ONE:v[0-9]+]], 0x3c00
▲ Show 20 Lines • Show All 161 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fcanonicalize.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
declare float @llvm.fabs.f32(float) #0 declare float @llvm.fabs.f32(float) #0
declare float @llvm.canonicalize.f32(float) #0 declare float @llvm.canonicalize.f32(float) #0
declare double @llvm.fabs.f64(double) #0 declare double @llvm.fabs.f64(double) #0
declare double @llvm.canonicalize.f64(double) #0 declare double @llvm.canonicalize.f64(double) #0
; GCN-LABEL: {{^}}v_test_canonicalize_var_f32: ; GCN-LABEL: {{^}}v_test_canonicalize_var_f32:
; GCN: v_mul_f32_e32 [[REG:v[0-9]+]], 1.0, {{v[0-9]+}} ; GCN: v_mul_f32_e32 [[REG:v[0-9]+]], 1.0, {{v[0-9]+}}
▲ Show 20 Lines • Show All 412 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fcmp.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}fcmp_f16_lt ; GCN-LABEL: {{^}}fcmp_f16_lt
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[B_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[B_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]] ; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]]
; SI: v_cmp_lt_f32_e32 vcc, v[[A_F32]], v[[B_F32]] ; SI: v_cmp_lt_f32_e32 vcc, v[[A_F32]], v[[B_F32]]
; VI: v_cmp_lt_f16_e32 vcc, v[[A_F16]], v[[B_F16]] ; VI: v_cmp_lt_f16_e32 vcc, v[[A_F16]], v[[B_F16]]
▲ Show 20 Lines • Show All 767 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fcmp64.ll

; RUN: llc < %s -march=amdgcn -mcpu=tahiti -verify-machineinstrs | FileCheck %s ; RUN: llc < %s -march=amdgcn -mcpu=tahiti -verify-machineinstrs | FileCheck %s
; RUN: llc < %s -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs | FileCheck %s ; RUN: llc < %s -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs | FileCheck %s
; CHECK-LABEL: {{^}}flt_f64: ; CHECK-LABEL: {{^}}flt_f64:
; CHECK: v_cmp_nge_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}} ; CHECK: v_cmp_nge_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @flt_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1, define amdgpu_kernel void @flt_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) { double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1 %r0 = load double, double addrspace(1)* %in1
%r1 = load double, double addrspace(1)* %in2 %r1 = load double, double addrspace(1)* %in2
%r2 = fcmp ult double %r0, %r1 %r2 = fcmp ult double %r0, %r1
%r3 = zext i1 %r2 to i32 %r3 = zext i1 %r2 to i32
store i32 %r3, i32 addrspace(1)* %out store i32 %r3, i32 addrspace(1)* %out
ret void ret void
} }
; CHECK-LABEL: {{^}}fle_f64: ; CHECK-LABEL: {{^}}fle_f64:
; CHECK: v_cmp_ngt_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}} ; CHECK: v_cmp_ngt_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @fle_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1, define amdgpu_kernel void @fle_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) { double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1 %r0 = load double, double addrspace(1)* %in1
%r1 = load double, double addrspace(1)* %in2 %r1 = load double, double addrspace(1)* %in2
%r2 = fcmp ule double %r0, %r1 %r2 = fcmp ule double %r0, %r1
%r3 = zext i1 %r2 to i32 %r3 = zext i1 %r2 to i32
store i32 %r3, i32 addrspace(1)* %out store i32 %r3, i32 addrspace(1)* %out
ret void ret void
} }
; CHECK-LABEL: {{^}}fgt_f64: ; CHECK-LABEL: {{^}}fgt_f64:
; CHECK: v_cmp_nle_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}} ; CHECK: v_cmp_nle_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @fgt_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1, define amdgpu_kernel void @fgt_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) { double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1 %r0 = load double, double addrspace(1)* %in1
%r1 = load double, double addrspace(1)* %in2 %r1 = load double, double addrspace(1)* %in2
%r2 = fcmp ugt double %r0, %r1 %r2 = fcmp ugt double %r0, %r1
%r3 = zext i1 %r2 to i32 %r3 = zext i1 %r2 to i32
store i32 %r3, i32 addrspace(1)* %out store i32 %r3, i32 addrspace(1)* %out
ret void ret void
} }
; CHECK-LABEL: {{^}}fge_f64: ; CHECK-LABEL: {{^}}fge_f64:
; CHECK: v_cmp_nlt_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}} ; CHECK: v_cmp_nlt_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @fge_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1, define amdgpu_kernel void @fge_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) { double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1 %r0 = load double, double addrspace(1)* %in1
%r1 = load double, double addrspace(1)* %in2 %r1 = load double, double addrspace(1)* %in2
%r2 = fcmp uge double %r0, %r1 %r2 = fcmp uge double %r0, %r1
%r3 = zext i1 %r2 to i32 %r3 = zext i1 %r2 to i32
store i32 %r3, i32 addrspace(1)* %out store i32 %r3, i32 addrspace(1)* %out
ret void ret void
} }
; CHECK-LABEL: {{^}}fne_f64: ; CHECK-LABEL: {{^}}fne_f64:
; CHECK: v_cmp_neq_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}} ; CHECK: v_cmp_neq_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @fne_f64(double addrspace(1)* %out, double addrspace(1)* %in1, define amdgpu_kernel void @fne_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) { double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1 %r0 = load double, double addrspace(1)* %in1
%r1 = load double, double addrspace(1)* %in2 %r1 = load double, double addrspace(1)* %in2
%r2 = fcmp une double %r0, %r1 %r2 = fcmp une double %r0, %r1
%r3 = select i1 %r2, double %r0, double %r1 %r3 = select i1 %r2, double %r0, double %r1
store double %r3, double addrspace(1)* %out store double %r3, double addrspace(1)* %out
ret void ret void
} }
; CHECK-LABEL: {{^}}feq_f64: ; CHECK-LABEL: {{^}}feq_f64:
; CHECK: v_cmp_nlg_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}} ; CHECK: v_cmp_nlg_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @feq_f64(double addrspace(1)* %out, double addrspace(1)* %in1, define amdgpu_kernel void @feq_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) { double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1 %r0 = load double, double addrspace(1)* %in1
%r1 = load double, double addrspace(1)* %in2 %r1 = load double, double addrspace(1)* %in2
%r2 = fcmp ueq double %r0, %r1 %r2 = fcmp ueq double %r0, %r1
%r3 = select i1 %r2, double %r0, double %r1 %r3 = select i1 %r2, double %r0, double %r1
store double %r3, double addrspace(1)* %out store double %r3, double addrspace(1)* %out
ret void ret void
} }

llvm/trunk/test/CodeGen/AMDGPU/fconst64.ll

; RUN: llc < %s -march=amdgcn -mcpu=tahiti -verify-machineinstrs | FileCheck %s ; RUN: llc < %s -march=amdgcn -mcpu=tahiti -verify-machineinstrs | FileCheck %s
; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck %s ; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck %s
; CHECK: {{^}}fconst_f64: ; CHECK: {{^}}fconst_f64:
; CHECK-DAG: s_mov_b32 {{s[0-9]+}}, 0x40140000 ; CHECK-DAG: s_mov_b32 {{s[0-9]+}}, 0x40140000
; CHECK-DAG: s_mov_b32 {{s[0-9]+}}, 0 ; CHECK-DAG: s_mov_b32 {{s[0-9]+}}, 0
define amdgpu_kernel void @fconst_f64(double addrspace(1)* %out, double addrspace(1)* %in) { define amdgpu_kernel void @fconst_f64(double addrspace(1)* %out, double addrspace(1)* %in) {
%r1 = load double, double addrspace(1)* %in %tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr inbounds double, double addrspace(1)* %in, i32 %tid
%r1 = load double, double addrspace(1)* %gep
%r2 = fadd double %r1, 5.000000e+00 %r2 = fadd double %r1, 5.000000e+00
store double %r2, double addrspace(1)* %out store double %r2, double addrspace(1)* %out
ret void ret void
} }
declare i32 @llvm.amdgcn.workitem.id.x() #1
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }

llvm/trunk/test/CodeGen/AMDGPU/fcopysign.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX8 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX8 %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s
declare half @llvm.copysign.f16(half, half) declare half @llvm.copysign.f16(half, half)
declare float @llvm.copysign.f32(float, float) declare float @llvm.copysign.f32(float, float)
declare double @llvm.copysign.f64(double, double) declare double @llvm.copysign.f64(double, double)
declare <2 x half> @llvm.copysign.v2f16(<2 x half>, <2 x half>) declare <2 x half> @llvm.copysign.v2f16(<2 x half>, <2 x half>)
declare <3 x half> @llvm.copysign.v3f16(<3 x half>, <3 x half>) declare <3 x half> @llvm.copysign.v3f16(<3 x half>, <3 x half>)
declare <4 x half> @llvm.copysign.v4f16(<4 x half>, <4 x half>) declare <4 x half> @llvm.copysign.v4f16(<4 x half>, <4 x half>)
declare i32 @llvm.amdgcn.workitem.id.x()
; GCN-LABEL: {{^}}test_copysign_f16: ; GCN-LABEL: {{^}}test_copysign_f16:
; SI: buffer_load_ushort v[[SIGN:[0-9]+]] ; SI: {{buffer|flat}}_load_ushort v[[SIGN:[0-9]+]]
; SI: buffer_load_ushort v[[MAG:[0-9]+]] ; SI: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
; SI: s_brev_b32 s[[CONST:[0-9]+]], -2 ; SI: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f32_f16_e32 v[[MAG_F32:[0-9]+]], v[[MAG]] ; SI-DAG: v_cvt_f32_f16_e32 v[[MAG_F32:[0-9]+]], v[[MAG]]
; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]] ; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]]
; SI: v_bfi_b32 v[[OUT_F32:[0-9]+]], s[[CONST]], v[[MAG_F32]], v[[SIGN_F32]] ; SI: v_bfi_b32 v[[OUT_F32:[0-9]+]], s[[CONST]], v[[MAG_F32]], v[[SIGN_F32]]
; SI: v_cvt_f16_f32_e32 v[[OUT:[0-9]+]], v[[OUT_F32]] ; SI: v_cvt_f16_f32_e32 v[[OUT:[0-9]+]], v[[OUT_F32]]
; GFX89: buffer_load_ushort v[[SIGN:[0-9]+]] ; GFX89: {{buffer|flat}}_load_ushort v[[SIGN:[0-9]+]]
; GFX89: buffer_load_ushort v[[MAG:[0-9]+]] ; GFX89: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
; GFX89: s_movk_i32 s[[CONST:[0-9]+]], 0x7fff ; GFX89: s_movk_i32 s[[CONST:[0-9]+]], 0x7fff
; GFX89: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN]] ; GFX89: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN]]
; GCN: buffer_store_short v[[OUT]] ; GCN: buffer_store_short v[[OUT]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copysign_f16( define amdgpu_kernel void @test_copysign_f16(
half addrspace(1)* %arg_out, half addrspace(1)* %arg_out,
half addrspace(1)* %arg_mag, half addrspace(1)* %arg_mag,
half addrspace(1)* %arg_sign) { half addrspace(1)* %arg_sign) {
entry: entry:
%mag = load half, half addrspace(1)* %arg_mag %mag = load half, half addrspace(1)* %arg_mag
%sign = load half, half addrspace(1)* %arg_sign %sign = load half, half addrspace(1)* %arg_sign
%out = call half @llvm.copysign.f16(half %mag, half %sign) %out = call half @llvm.copysign.f16(half %mag, half %sign)
store half %out, half addrspace(1)* %arg_out store half %out, half addrspace(1)* %arg_out
ret void ret void
} }
; GCN-LABEL: {{^}}test_copysign_out_f32_mag_f16_sign_f32: ; GCN-LABEL: {{^}}test_copysign_out_f32_mag_f16_sign_f32:
; GCN-DAG: buffer_load_ushort v[[MAG:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
; GCN-DAG: buffer_load_dword v[[SIGN:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_dword v[[SIGN:[0-9]+]]
; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2 ; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; GCN-DAG: v_cvt_f32_f16_e32 v[[MAG_EXT:[0-9]+]], v[[MAG]] ; GCN-DAG: v_cvt_f32_f16_e32 v[[MAG_EXT:[0-9]+]], v[[MAG]]
; GCN: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG_EXT]], v[[SIGN]] ; GCN: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG_EXT]], v[[SIGN]]
; GCN: buffer_store_dword v[[OUT]] ; GCN: buffer_store_dword v[[OUT]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copysign_out_f32_mag_f16_sign_f32( define amdgpu_kernel void @test_copysign_out_f32_mag_f16_sign_f32(
float addrspace(1)* %arg_out, float addrspace(1)* %arg_out,
half addrspace(1)* %arg_mag, half addrspace(1)* %arg_mag,
float addrspace(1)* %arg_sign) { float addrspace(1)* %arg_sign) {
entry: entry:
%mag = load half, half addrspace(1)* %arg_mag %tid = call i32 @llvm.amdgcn.workitem.id.x()
%arg_mag_gep = getelementptr half, half addrspace(1)* %arg_mag, i32 %tid
%mag = load half, half addrspace(1)* %arg_mag_gep
%mag.ext = fpext half %mag to float %mag.ext = fpext half %mag to float
%sign = load float, float addrspace(1)* %arg_sign %arg_sign_gep = getelementptr float, float addrspace(1)* %arg_sign, i32 %tid
%sign = load float, float addrspace(1)* %arg_sign_gep
%out = call float @llvm.copysign.f32(float %mag.ext, float %sign) %out = call float @llvm.copysign.f32(float %mag.ext, float %sign)
store float %out, float addrspace(1)* %arg_out store float %out, float addrspace(1)* %arg_out
ret void ret void
} }
; GCN-LABEL: {{^}}test_copysign_out_f64_mag_f16_sign_f64: ; GCN-LABEL: {{^}}test_copysign_out_f64_mag_f16_sign_f64:
; GCN-DAG: buffer_load_ushort v[[MAG:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
; GCN-DAG: buffer_load_dwordx2 v{{\[}}[[SIGN_LO:[0-9]+]]:[[SIGN_HI:[0-9]+]]{{\]}} ; GCN-DAG: {{buffer|flat}}_load_dwordx2 v{{\[}}[[SIGN_LO:[0-9]+]]:[[SIGN_HI:[0-9]+]]{{\]}}
; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2 ; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; GCN-DAG: v_cvt_f32_f16_e32 v[[MAG_EXT:[0-9]+]], v[[MAG]] ; GCN-DAG: v_cvt_f32_f16_e32 v[[MAG_EXT:[0-9]+]], v[[MAG]]
; GCN-DAG: v_cvt_f64_f32_e32 v{{\[}}[[MAG_EXT_LO:[0-9]+]]:[[MAG_EXT_HI:[0-9]+]]{{\]}}, v[[MAG_EXT]] ; GCN-DAG: v_cvt_f64_f32_e32 v{{\[}}[[MAG_EXT_LO:[0-9]+]]:[[MAG_EXT_HI:[0-9]+]]{{\]}}, v[[MAG_EXT]]
; GCN: v_bfi_b32 v[[OUT_HI:[0-9]+]], s[[CONST]], v[[MAG_EXT_HI]], v[[SIGN_HI]] ; GCN: v_bfi_b32 v[[OUT_HI:[0-9]+]], s[[CONST]], v[[MAG_EXT_HI]], v[[SIGN_HI]]
; GCN: buffer_store_dwordx2 v{{\[}}[[MAG_EXT_LO]]:[[OUT_HI]]{{\]}} ; GCN: buffer_store_dwordx2 v{{\[}}[[MAG_EXT_LO]]:[[OUT_HI]]{{\]}}
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copysign_out_f64_mag_f16_sign_f64( define amdgpu_kernel void @test_copysign_out_f64_mag_f16_sign_f64(
double addrspace(1)* %arg_out, double addrspace(1)* %arg_out,
half addrspace(1)* %arg_mag, half addrspace(1)* %arg_mag,
double addrspace(1)* %arg_sign) { double addrspace(1)* %arg_sign) {
entry: entry:
%mag = load half, half addrspace(1)* %arg_mag %tid = call i32 @llvm.amdgcn.workitem.id.x()
%arg_mag_gep = getelementptr half, half addrspace(1)* %arg_mag, i32 %tid
%mag = load half, half addrspace(1)* %arg_mag_gep
%mag.ext = fpext half %mag to double %mag.ext = fpext half %mag to double
%sign = load double, double addrspace(1)* %arg_sign %arg_sign_gep = getelementptr double, double addrspace(1)* %arg_sign, i32 %tid
%sign = load double, double addrspace(1)* %arg_sign_gep
%out = call double @llvm.copysign.f64(double %mag.ext, double %sign) %out = call double @llvm.copysign.f64(double %mag.ext, double %sign)
store double %out, double addrspace(1)* %arg_out store double %out, double addrspace(1)* %arg_out
ret void ret void
} }
; GCN-LABEL: {{^}}test_copysign_out_f32_mag_f32_sign_f16: ; GCN-LABEL: {{^}}test_copysign_out_f32_mag_f32_sign_f16:
; GCN-DAG: buffer_load_dword v[[MAG:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_dword v[[MAG:[0-9]+]]
; GCN-DAG: buffer_load_ushort v[[SIGN:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_ushort v[[SIGN:[0-9]+]]
; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2 ; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]] ; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]]
; SI: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN_F32]] ; SI: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN_F32]]
; GFX89-DAG: v_lshlrev_b32_e32 v[[SIGN_SHIFT:[0-9]+]], 16, v[[SIGN]] ; GFX89-DAG: v_lshlrev_b32_e32 v[[SIGN_SHIFT:[0-9]+]], 16, v[[SIGN]]
; GFX89: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN_SHIFT]] ; GFX89: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN_SHIFT]]
; GCN: buffer_store_dword v[[OUT]] ; GCN: buffer_store_dword v[[OUT]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copysign_out_f32_mag_f32_sign_f16( define amdgpu_kernel void @test_copysign_out_f32_mag_f32_sign_f16(
float addrspace(1)* %arg_out, float addrspace(1)* %arg_out,
float addrspace(1)* %arg_mag, float addrspace(1)* %arg_mag,
half addrspace(1)* %arg_sign) { half addrspace(1)* %arg_sign) {
entry: entry:
%mag = load float, float addrspace(1)* %arg_mag %tid = call i32 @llvm.amdgcn.workitem.id.x()
%sign = load half, half addrspace(1)* %arg_sign %arg_mag_gep = getelementptr float, float addrspace(1)* %arg_mag, i32 %tid
%mag = load float, float addrspace(1)* %arg_mag_gep
%arg_sign_gep = getelementptr half, half addrspace(1)* %arg_sign, i32 %tid
%sign = load half, half addrspace(1)* %arg_sign_gep
%sign.ext = fpext half %sign to float %sign.ext = fpext half %sign to float
%out = call float @llvm.copysign.f32(float %mag, float %sign.ext) %out = call float @llvm.copysign.f32(float %mag, float %sign.ext)
store float %out, float addrspace(1)* %arg_out store float %out, float addrspace(1)* %arg_out
ret void ret void
} }
; GCN-LABEL: {{^}}test_copysign_out_f64_mag_f64_sign_f16: ; GCN-LABEL: {{^}}test_copysign_out_f64_mag_f64_sign_f16:
; GCN-DAG: buffer_load_dwordx2 v{{\[}}[[MAG_LO:[0-9]+]]:[[MAG_HI:[0-9]+]]{{\]}} ; GCN-DAG: {{buffer|flat}}_load_dwordx2 v{{\[}}[[MAG_LO:[0-9]+]]:[[MAG_HI:[0-9]+]]{{\]}}
; GCN-DAG: buffer_load_ushort v[[SIGN:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_ushort v[[SIGN:[0-9]+]]
; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2 ; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]] ; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]]
; SI: v_bfi_b32 v[[OUT_HI:[0-9]+]], s[[CONST]], v[[MAG_HI]], v[[SIGN_F32]] ; SI: v_bfi_b32 v[[OUT_HI:[0-9]+]], s[[CONST]], v[[MAG_HI]], v[[SIGN_F32]]
; GFX89-DAG: v_lshlrev_b32_e32 v[[SIGN_SHIFT:[0-9]+]], 16, v[[SIGN]] ; GFX89-DAG: v_lshlrev_b32_e32 v[[SIGN_SHIFT:[0-9]+]], 16, v[[SIGN]]
; GFX89: v_bfi_b32 v[[OUT_HI:[0-9]+]], s[[CONST]], v[[MAG_HI]], v[[SIGN_SHIFT]] ; GFX89: v_bfi_b32 v[[OUT_HI:[0-9]+]], s[[CONST]], v[[MAG_HI]], v[[SIGN_SHIFT]]
; GCN: buffer_store_dwordx2 v{{\[}}[[MAG_LO]]:[[OUT_HI]]{{\]}} ; GCN: buffer_store_dwordx2 v{{\[}}[[MAG_LO]]:[[OUT_HI]]{{\]}}
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copysign_out_f64_mag_f64_sign_f16( define amdgpu_kernel void @test_copysign_out_f64_mag_f64_sign_f16(
double addrspace(1)* %arg_out, double addrspace(1)* %arg_out,
double addrspace(1)* %arg_mag, double addrspace(1)* %arg_mag,
half addrspace(1)* %arg_sign) { half addrspace(1)* %arg_sign) {
entry: entry:
%mag = load double, double addrspace(1)* %arg_mag %tid = call i32 @llvm.amdgcn.workitem.id.x()
%sign = load half, half addrspace(1)* %arg_sign %arg_mag_gep = getelementptr double, double addrspace(1)* %arg_mag, i32 %tid
%mag = load double, double addrspace(1)* %arg_mag_gep
%arg_sign_gep = getelementptr half, half addrspace(1)* %arg_sign, i32 %tid
%sign = load half, half addrspace(1)* %arg_sign_gep
%sign.ext = fpext half %sign to double %sign.ext = fpext half %sign to double
%out = call double @llvm.copysign.f64(double %mag, double %sign.ext) %out = call double @llvm.copysign.f64(double %mag, double %sign.ext)
store double %out, double addrspace(1)* %arg_out store double %out, double addrspace(1)* %arg_out
ret void ret void
} }
; GCN-LABEL: {{^}}test_copysign_out_f16_mag_f16_sign_f32: ; GCN-LABEL: {{^}}test_copysign_out_f16_mag_f16_sign_f32:
; GCN-DAG: buffer_load_ushort v[[MAG:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
; GCN-DAG: buffer_load_dword v[[SIGN:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_dword v[[SIGN:[0-9]+]]
; SI-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2 ; SI-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f32_f16_e32 v[[MAG_F32:[0-9]+]], v[[MAG]] ; SI-DAG: v_cvt_f32_f16_e32 v[[MAG_F32:[0-9]+]], v[[MAG]]
; SI: v_bfi_b32 v[[OUT_F32:[0-9]+]], s[[CONST]], v[[MAG_F32]], v[[SIGN]] ; SI: v_bfi_b32 v[[OUT_F32:[0-9]+]], s[[CONST]], v[[MAG_F32]], v[[SIGN]]
; SI: v_cvt_f16_f32_e32 v[[OUT:[0-9]+]], v[[OUT_F32]] ; SI: v_cvt_f16_f32_e32 v[[OUT:[0-9]+]], v[[OUT_F32]]
; GFX89-DAG: s_movk_i32 s[[CONST:[0-9]+]], 0x7fff ; GFX89-DAG: s_movk_i32 s[[CONST:[0-9]+]], 0x7fff
; GFX89-DAG: v_lshrrev_b32_e32 v[[SIGN_SHIFT:[0-9]+]], 16, v[[SIGN]] ; GFX89-DAG: v_lshrrev_b32_e32 v[[SIGN_SHIFT:[0-9]+]], 16, v[[SIGN]]
; GFX89: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN_SHIFT]] ; GFX89: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN_SHIFT]]
; GCN: buffer_store_short v[[OUT]] ; GCN: buffer_store_short v[[OUT]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copysign_out_f16_mag_f16_sign_f32( define amdgpu_kernel void @test_copysign_out_f16_mag_f16_sign_f32(
half addrspace(1)* %arg_out, half addrspace(1)* %arg_out,
half addrspace(1)* %arg_mag, half addrspace(1)* %arg_mag,
float addrspace(1)* %arg_sign) { float addrspace(1)* %arg_sign) {
entry: entry:
%mag = load half, half addrspace(1)* %arg_mag %tid = call i32 @llvm.amdgcn.workitem.id.x()
%sign = load float, float addrspace(1)* %arg_sign %arg_mag_gep = getelementptr half, half addrspace(1)* %arg_mag, i32 %tid
%mag = load half, half addrspace(1)* %arg_mag_gep
%arg_sign_gep = getelementptr float, float addrspace(1)* %arg_sign, i32 %tid
%sign = load float, float addrspace(1)* %arg_sign_gep
%sign.trunc = fptrunc float %sign to half %sign.trunc = fptrunc float %sign to half
%out = call half @llvm.copysign.f16(half %mag, half %sign.trunc) %out = call half @llvm.copysign.f16(half %mag, half %sign.trunc)
store half %out, half addrspace(1)* %arg_out store half %out, half addrspace(1)* %arg_out
ret void ret void
} }
; GCN-LABEL: {{^}}test_copysign_out_f16_mag_f16_sign_f64: ; GCN-LABEL: {{^}}test_copysign_out_f16_mag_f16_sign_f64:
; GCN-DAG: buffer_load_ushort v[[MAG:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
; GCN-DAG: buffer_load_dwordx2 v{{\[}}[[SIGN_LO:[0-9]+]]:[[SIGN_HI:[0-9]+]]{{\]}} ; GCN-DAG: {{buffer|flat}}_load_dwordx2 v{{\[}}[[SIGN_LO:[0-9]+]]:[[SIGN_HI:[0-9]+]]{{\]}}
; SI-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2 ; SI-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f32_f16_e32 v[[MAG_F32:[0-9]+]], v[[MAG]] ; SI-DAG: v_cvt_f32_f16_e32 v[[MAG_F32:[0-9]+]], v[[MAG]]
; SI: v_bfi_b32 v[[OUT_F32:[0-9]+]], s[[CONST]], v[[MAG_F32]], v[[SIGN_HI]] ; SI: v_bfi_b32 v[[OUT_F32:[0-9]+]], s[[CONST]], v[[MAG_F32]], v[[SIGN_HI]]
; SI: v_cvt_f16_f32_e32 v[[OUT:[0-9]+]], v[[OUT_F32]] ; SI: v_cvt_f16_f32_e32 v[[OUT:[0-9]+]], v[[OUT_F32]]
; GFX89-DAG: s_movk_i32 s[[CONST:[0-9]+]], 0x7fff ; GFX89-DAG: s_movk_i32 s[[CONST:[0-9]+]], 0x7fff
; GFX89-DAG: v_lshrrev_b32_e32 v[[SIGN_SHIFT:[0-9]+]], 16, v[[SIGN_HI]] ; GFX89-DAG: v_lshrrev_b32_e32 v[[SIGN_SHIFT:[0-9]+]], 16, v[[SIGN_HI]]
; GFX89: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN_SHIFT]] ; GFX89: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN_SHIFT]]
; GCN: buffer_store_short v[[OUT]] ; GCN: buffer_store_short v[[OUT]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copysign_out_f16_mag_f16_sign_f64( define amdgpu_kernel void @test_copysign_out_f16_mag_f16_sign_f64(
half addrspace(1)* %arg_out, half addrspace(1)* %arg_out,
half addrspace(1)* %arg_mag, half addrspace(1)* %arg_mag,
double addrspace(1)* %arg_sign) { double addrspace(1)* %arg_sign) {
entry: entry:
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%arg_mag_gep = getelementptr half, half addrspace(1)* %arg_mag, i32 %tid
%mag = load half, half addrspace(1)* %arg_mag %mag = load half, half addrspace(1)* %arg_mag
%sign = load double, double addrspace(1)* %arg_sign %arg_sign_gep = getelementptr double, double addrspace(1)* %arg_sign, i32 %tid
%sign = load double, double addrspace(1)* %arg_sign_gep
%sign.trunc = fptrunc double %sign to half %sign.trunc = fptrunc double %sign to half
%out = call half @llvm.copysign.f16(half %mag, half %sign.trunc) %out = call half @llvm.copysign.f16(half %mag, half %sign.trunc)
store half %out, half addrspace(1)* %arg_out store half %out, half addrspace(1)* %arg_out
ret void ret void
} }
; GCN-LABEL: {{^}}test_copysign_out_f16_mag_f32_sign_f16: ; GCN-LABEL: {{^}}test_copysign_out_f16_mag_f32_sign_f16:
; GCN-DAG: buffer_load_dword v[[MAG:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_dword v[[MAG:[0-9]+]]
; GCN-DAG: buffer_load_ushort v[[SIGN:[0-9]+]] ; GCN-DAG: {{buffer|flat}}_load_ushort v[[SIGN:[0-9]+]]
; SI-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2 ; SI-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f16_f32_e32 v[[MAG_TRUNC:[0-9]+]], v[[MAG]] ; SI-DAG: v_cvt_f16_f32_e32 v[[MAG_TRUNC:[0-9]+]], v[[MAG]]
; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]] ; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]]
; SI-DAG: v_cvt_f32_f16_e32 v[[MAG_F32:[0-9]+]], v[[MAG_TRUNC]] ; SI-DAG: v_cvt_f32_f16_e32 v[[MAG_F32:[0-9]+]], v[[MAG_TRUNC]]
; SI: v_bfi_b32 v[[OUT_F32:[0-9]+]], s[[CONST]], v[[MAG_F32]], v[[SIGN_F32]] ; SI: v_bfi_b32 v[[OUT_F32:[0-9]+]], s[[CONST]], v[[MAG_F32]], v[[SIGN_F32]]
; SI: v_cvt_f16_f32_e32 v[[OUT:[0-9]+]], v[[OUT_F32]] ; SI: v_cvt_f16_f32_e32 v[[OUT:[0-9]+]], v[[OUT_F32]]
; GFX89-DAG: s_movk_i32 s[[CONST:[0-9]+]], 0x7fff ; GFX89-DAG: s_movk_i32 s[[CONST:[0-9]+]], 0x7fff
; GFX89-DAG: v_cvt_f16_f32_e32 v[[MAG_TRUNC:[0-9]+]], v[[MAG]] ; GFX89-DAG: v_cvt_f16_f32_e32 v[[MAG_TRUNC:[0-9]+]], v[[MAG]]
; GFX89: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG_TRUNC]], v[[SIGN]] ; GFX89: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG_TRUNC]], v[[SIGN]]
; GCN: buffer_store_short v[[OUT]] ; GCN: buffer_store_short v[[OUT]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @test_copysign_out_f16_mag_f32_sign_f16( define amdgpu_kernel void @test_copysign_out_f16_mag_f32_sign_f16(
half addrspace(1)* %arg_out, half addrspace(1)* %arg_out,
float addrspace(1)* %arg_mag, float addrspace(1)* %arg_mag,
half addrspace(1)* %arg_sign) { half addrspace(1)* %arg_sign) {
entry: entry:
%mag = load float, float addrspace(1)* %arg_mag %tid = call i32 @llvm.amdgcn.workitem.id.x()
%arg_mag_gep = getelementptr float, float addrspace(1)* %arg_mag, i32 %tid
%mag = load float, float addrspace(1)* %arg_mag_gep
%mag.trunc = fptrunc float %mag to half %mag.trunc = fptrunc float %mag to half
%sign = load half, half addrspace(1)* %arg_sign %arg_sign_gep = getelementptr half, half addrspace(1)* %arg_sign, i32 %tid
%sign = load half, half addrspace(1)* %arg_sign_gep
%out = call half @llvm.copysign.f16(half %mag.trunc, half %sign) %out = call half @llvm.copysign.f16(half %mag.trunc, half %sign)
store half %out, half addrspace(1)* %arg_out store half %out, half addrspace(1)* %arg_out
ret void ret void
} }
; GCN-LABEL: {{^}}test_copysign_out_f16_mag_f64_sign_f16: ; GCN-LABEL: {{^}}test_copysign_out_f16_mag_f64_sign_f16:
; GCN: v_bfi_b32 ; GCN: v_bfi_b32
; GCN: s_endpgm ; GCN: s_endpgm
▲ Show 20 Lines • Show All 58 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fma-combine.ll

; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast -enable-no-infs-fp-math -enable-unsafe-fp-math -mattr=+fp32-denormals < %s | FileCheck -check-prefix=SI-FMA -check-prefix=SI-UNSAFE -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast -enable-no-infs-fp-math -enable-unsafe-fp-math -mattr=+fp32-denormals < %s | FileCheck -check-prefix=SI-FMA -check-prefix=SI-UNSAFE -check-prefix=SI -check-prefix=FUNC %s
; Note: The SI-FMA conversions of type x * (y + 1) --> x * y + x would be ; Note: The SI-FMA conversions of type x * (y + 1) --> x * y + x would be
; beneficial even without fp32 denormals, but they do require no-infs-fp-math ; beneficial even without fp32 denormals, but they do require no-infs-fp-math
; for correctness. ; for correctness.
declare i32 @llvm.amdgcn.workitem.id.x() #0 declare i32 @llvm.amdgcn.workitem.id.x() #0
declare double @llvm.fabs.f64(double) #0 declare double @llvm.fabs.f64(double) #0
declare double @llvm.fma.f64(double, double, double) #0 declare double @llvm.fma.f64(double, double, double) #0
▲ Show 20 Lines • Show All 618 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fma.f64.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
declare double @llvm.fma.f64(double, double, double) nounwind readnone declare double @llvm.fma.f64(double, double, double) nounwind readnone
declare <2 x double> @llvm.fma.v2f64(<2 x double>, <2 x double>, <2 x double>) nounwind readnone declare <2 x double> @llvm.fma.v2f64(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
declare <4 x double> @llvm.fma.v4f64(<4 x double>, <4 x double>, <4 x double>) nounwind readnone declare <4 x double> @llvm.fma.v4f64(<4 x double>, <4 x double>, <4 x double>) nounwind readnone
; FUNC-LABEL: {{^}}fma_f64: ; FUNC-LABEL: {{^}}fma_f64:
; SI: v_fma_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}} ; SI: v_fma_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
Show All 37 Lines

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

; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare float @llvm.fma.f32(float, float, float) nounwind readnone declare float @llvm.fma.f32(float, float, float) nounwind readnone
declare <2 x float> @llvm.fma.v2f32(<2 x float>, <2 x float>, <2 x float>) nounwind readnone declare <2 x float> @llvm.fma.v2f32(<2 x float>, <2 x float>, <2 x float>) nounwind readnone
declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>) nounwind readnone declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
declare i32 @llvm.r600.read.tidig.x() nounwind readnone declare i32 @llvm.r600.read.tidig.x() nounwind readnone
; FUNC-LABEL: {{^}}fma_f32: ; FUNC-LABEL: {{^}}fma_f32:
▲ Show 20 Lines • Show All 82 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fmul.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}fmul_f16 ; GCN-LABEL: {{^}}fmul_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[B_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[B_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]] ; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]]
; SI: v_mul_f32_e32 v[[R_F32:[0-9]+]], v[[B_F32]], v[[A_F32]] ; SI: v_mul_f32_e32 v[[R_F32:[0-9]+]], v[[B_F32]], v[[A_F32]]
; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]] ; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]]
▲ Show 20 Lines • Show All 138 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fmul64.ll

; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s
; FUNC-LABEL: {{^}}fmul_f64: ; FUNC-LABEL: {{^}}fmul_f64:
; SI: v_mul_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}} ; SI: v_mul_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
define amdgpu_kernel void @fmul_f64(double addrspace(1)* %out, double addrspace(1)* %in1, define amdgpu_kernel void @fmul_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) { double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1 %r0 = load double, double addrspace(1)* %in1
%r1 = load double, double addrspace(1)* %in2 %r1 = load double, double addrspace(1)* %in2
%r2 = fmul double %r0, %r1 %r2 = fmul double %r0, %r1
Show All 29 Lines

llvm/trunk/test/CodeGen/AMDGPU/fmuladd.f32.ll

; RUN: llc -verify-machineinstrs -mcpu=tahiti -mattr=-fp32-denormals,+fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-STRICT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-FASTFMA,GCN-FLUSH-FASTFMA-STRICT,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=tahiti -mattr=-fp32-denormals,+fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-STRICT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-FASTFMA,GCN-FLUSH-FASTFMA-STRICT,SI %s
; RUN: llc -verify-machineinstrs -mcpu=tahiti -mattr=+fp32-denormals,+fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-STRICT,GCN-DENORM,SI-DENORM,GCN-DENORM-FASTFMA,GCN-DENORM-FASTFMA-STRICT,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=tahiti -mattr=+fp32-denormals,+fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-STRICT,GCN-DENORM,SI-DENORM,GCN-DENORM-FASTFMA,GCN-DENORM-FASTFMA-STRICT,SI %s
; RUN: llc -verify-machineinstrs -mcpu=verde -mattr=-fp32-denormals,-fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-STRICT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-SLOWFMA,GCN-FLUSH-SLOWFMA-STRICT,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=verde -mattr=-fp32-denormals,-fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-STRICT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-SLOWFMA,GCN-FLUSH-SLOWFMA-STRICT,SI %s
; RUN: llc -verify-machineinstrs -mcpu=verde -mattr=+fp32-denormals,-fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-STRICT,GCN-DENORM,SI-DENORM,GCN-DENORM-SLOWFMA,GCN-DENORM-SLOWFMA-STRICT,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=verde -mattr=+fp32-denormals,-fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-STRICT,GCN-DENORM,SI-DENORM,GCN-DENORM-SLOWFMA,GCN-DENORM-SLOWFMA-STRICT,SI %s
; RUN: llc -verify-machineinstrs -mcpu=tahiti -mattr=-fp32-denormals,+fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-CONTRACT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-FASTFMA,GCN-FLUSH-FASTFMA-CONTRACT,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=tahiti -mattr=-fp32-denormals,+fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-CONTRACT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-FASTFMA,GCN-FLUSH-FASTFMA-CONTRACT,SI %s
; RUN: llc -verify-machineinstrs -mcpu=tahiti -mattr=+fp32-denormals,+fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-CONTRACT,GCN-DENORM,SI-DENORM,GCN-DENORM-FASTFMA,GCN-DENORM-FASTFMA-CONTRACT,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=tahiti -mattr=+fp32-denormals,+fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-CONTRACT,GCN-DENORM,SI-DENORM,GCN-DENORM-FASTFMA,GCN-DENORM-FASTFMA-CONTRACT,SI %s
; RUN: llc -verify-machineinstrs -mcpu=verde -mattr=-fp32-denormals,-fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-CONTRACT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-SLOWFMA,GCN-FLUSH-SLOWFMA-CONTRACT,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=verde -mattr=-fp32-denormals,-fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-CONTRACT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-SLOWFMA,GCN-FLUSH-SLOWFMA-CONTRACT,SI %s
; RUN: llc -verify-machineinstrs -mcpu=verde -mattr=+fp32-denormals,-fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-CONTRACT,GCN-DENORM,SI-DENORM,GCN-DENORM-SLOWFMA,GCN-DENORM-SLOWFMA-CONTRACT,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=verde -mattr=+fp32-denormals,-fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-CONTRACT,GCN-DENORM,SI-DENORM,GCN-DENORM-SLOWFMA,GCN-DENORM-SLOWFMA-CONTRACT,SI %s
; Test all permutations of: fp32 denormals, fast fp contract, fp contract enabled for fmuladd, fmaf fast/slow. ; Test all permutations of: fp32 denormals, fast fp contract, fp contract enabled for fmuladd, fmaf fast/slow.
target triple = "amdgcn--" target triple = "amdgcn--"
declare i32 @llvm.amdgcn.workitem.id.x() #1 declare i32 @llvm.amdgcn.workitem.id.x() #1
declare float @llvm.fmuladd.f32(float, float, float) #1 declare float @llvm.fmuladd.f32(float, float, float) #1
▲ Show 20 Lines • Show All 566 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fmuladd.f64.ll

; RUN: llc -march=amdgcn -mcpu=tahiti -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICTSI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICTSI %s
; RUN: llc -march=amdgcn -mcpu=verde -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,SI %s
; RUN: llc -march=amdgcn -mcpu=tahiti -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,SI %s
; RUN: llc -march=amdgcn -mcpu=verde -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,VI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,VI %s
; GCN-LABEL: {{^}}fmuladd_f64: ; GCN-LABEL: {{^}}fmuladd_f64:
; GCN: v_fma_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}} ; GCN: v_fma_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
define amdgpu_kernel void @fmuladd_f64(double addrspace(1)* %out, double addrspace(1)* %in1, define amdgpu_kernel void @fmuladd_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2, double addrspace(1)* %in3) #0 { double addrspace(1)* %in2, double addrspace(1)* %in3) #0 {
%r0 = load double, double addrspace(1)* %in1 %r0 = load double, double addrspace(1)* %in1
%r1 = load double, double addrspace(1)* %in2 %r1 = load double, double addrspace(1)* %in2
%r2 = load double, double addrspace(1)* %in3 %r2 = load double, double addrspace(1)* %in3
▲ Show 20 Lines • Show All 168 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fmuladd.v2f16.ll

; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-FLUSH,GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-FLUSH,GFX9 %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-FLUSH,GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-FLUSH,GFX9 %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-FLUSH,GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-FLUSH,GFX9 %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-FLUSH,GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-FLUSH,GFX9 %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-DENORM-STRICT,GFX9-DENORM,GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-DENORM-STRICT,GFX9-DENORM,GFX9 %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-DENORM-STRICT,GFX9-DENORM,GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-DENORM-STRICT,GFX9-DENORM,GFX9 %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-DENORM-CONTRACT,GFX9-DENORM,GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-DENORM-CONTRACT,GFX9-DENORM,GFX9 %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-DENORM-CONTRACT,GFX9-DENORM,GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-DENORM-CONTRACT,GFX9-DENORM,GFX9 %s
declare i32 @llvm.amdgcn.workitem.id.x() #1 declare i32 @llvm.amdgcn.workitem.id.x() #1
declare <2 x half> @llvm.fmuladd.v2f16(<2 x half>, <2 x half>, <2 x half>) #1 declare <2 x half> @llvm.fmuladd.v2f16(<2 x half>, <2 x half>, <2 x half>) #1
declare <2 x half> @llvm.fabs.v2f16(<2 x half>) #1 declare <2 x half> @llvm.fabs.v2f16(<2 x half>) #1
; GCN-LABEL: {{^}}fmuladd_v2f16: ; GCN-LABEL: {{^}}fmuladd_v2f16:
; GFX9-FLUSH: v_pk_mul_f16 {{v[0-9]+, v[0-9]+, v[0-9]+}} ; GFX9-FLUSH: v_pk_mul_f16 {{v[0-9]+, v[0-9]+, v[0-9]+}}
; GFX9-FLUSH: v_pk_add_f16 {{v[0-9]+, v[0-9]+, v[0-9]+}} ; GFX9-FLUSH: v_pk_add_f16 {{v[0-9]+, v[0-9]+, v[0-9]+}}
▲ Show 20 Lines • Show All 90 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fneg-fabs.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}fneg_fabs_fadd_f32: ; FUNC-LABEL: {{^}}fneg_fabs_fadd_f32:
; SI-NOT: and ; SI-NOT: and
; SI: v_sub_f32_e64 {{v[0-9]+}}, {{s[0-9]+}}, |{{v[0-9]+}}| ; SI: v_sub_f32_e64 {{v[0-9]+}}, {{s[0-9]+}}, |{{v[0-9]+}}|
define amdgpu_kernel void @fneg_fabs_fadd_f32(float addrspace(1)* %out, float %x, float %y) { define amdgpu_kernel void @fneg_fabs_fadd_f32(float addrspace(1)* %out, float %x, float %y) {
%fabs = call float @llvm.fabs.f32(float %x) %fabs = call float @llvm.fabs.f32(float %x)
%fsub = fsub float -0.000000e+00, %fabs %fsub = fsub float -0.000000e+00, %fabs
%fadd = fadd float %y, %fsub %fadd = fadd float %y, %fsub
▲ Show 20 Lines • Show All 101 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fneg.f16.ll

; RUN: llc -march=amdgcn -mcpu=kaveri -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=CIVI -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=kaveri -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=CIVI -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=tonga -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=CIVI -check-prefix=GCN -check-prefix=GFX89 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=CIVI -check-prefix=GCN -check-prefix=GFX89 %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 -check-prefix=GCN -check-prefix=GFX89 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 -check-prefix=GCN -check-prefix=GFX89 %s
; FIXME: Should be able to do scalar op ; FIXME: Should be able to do scalar op
; GCN-LABEL: {{^}}s_fneg_f16: ; GCN-LABEL: {{^}}s_fneg_f16:
define amdgpu_kernel void @s_fneg_f16(half addrspace(1)* %out, half %in) #0 { define amdgpu_kernel void @s_fneg_f16(half addrspace(1)* %out, half %in) #0 {
%fneg = fsub half -0.0, %in %fneg = fsub half -0.0, %in
store half %fneg, half addrspace(1)* %out store half %fneg, half addrspace(1)* %out
ret void ret void
} }
▲ Show 20 Lines • Show All 160 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fp32_to_fp16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i16 @llvm.convert.to.fp16.f32(float) nounwind readnone declare i16 @llvm.convert.to.fp16.f32(float) nounwind readnone
; FUNC-LABEL: {{^}}test_convert_fp32_to_fp16: ; FUNC-LABEL: {{^}}test_convert_fp32_to_fp16:
; GCN: buffer_load_dword [[VAL:v[0-9]+]] ; GCN: buffer_load_dword [[VAL:v[0-9]+]]
; GCN: v_cvt_f16_f32_e32 [[RESULT:v[0-9]+]], [[VAL]] ; GCN: v_cvt_f16_f32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_short [[RESULT]] ; GCN: buffer_store_short [[RESULT]]
Show All 9 Lines

llvm/trunk/test/CodeGen/AMDGPU/fpext.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=GFX89 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=GFX89 %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -check-prefix=GFX89 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -check-prefix=GFX89 %s
; GCN-LABEL: {{^}}fpext_f16_to_f32 ; GCN-LABEL: {{^}}fpext_f16_to_f32
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: v_cvt_f32_f16_e32 v[[R_F32:[0-9]+]], v[[A_F16]] ; GCN: v_cvt_f32_f16_e32 v[[R_F32:[0-9]+]], v[[A_F16]]
; GCN: buffer_store_dword v[[R_F32]] ; GCN: buffer_store_dword v[[R_F32]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @fpext_f16_to_f32( define amdgpu_kernel void @fpext_f16_to_f32(
float addrspace(1)* %r, float addrspace(1)* %r,
▲ Show 20 Lines • Show All 263 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fptosi.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}fptosi_f16_to_i16 ; GCN-LABEL: {{^}}fptosi_f16_to_i16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; GCN: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; GCN: v_cvt_i32_f32_e32 v[[R_I16:[0-9]+]], v[[A_F32]] ; GCN: v_cvt_i32_f32_e32 v[[R_I16:[0-9]+]], v[[A_F32]]
; GCN: buffer_store_short v[[R_I16]] ; GCN: buffer_store_short v[[R_I16]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @fptosi_f16_to_i16( define amdgpu_kernel void @fptosi_f16_to_i16(
▲ Show 20 Lines • Show All 110 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fptoui.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}fptoui_f16_to_i16 ; GCN-LABEL: {{^}}fptoui_f16_to_i16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; GCN: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_cvt_u32_f32_e32 v[[R_I16:[0-9]+]], v[[A_F32]] ; SI: v_cvt_u32_f32_e32 v[[R_I16:[0-9]+]], v[[A_F32]]
; VI: v_cvt_i32_f32_e32 v[[R_I16:[0-9]+]], v[[A_F32]] ; VI: v_cvt_i32_f32_e32 v[[R_I16:[0-9]+]], v[[A_F32]]
; GCN: buffer_store_short v[[R_I16]] ; GCN: buffer_store_short v[[R_I16]]
; GCN: s_endpgm ; GCN: s_endpgm
▲ Show 20 Lines • Show All 110 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fptrunc.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=SIVI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=SIVI %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global,-fp64-fp16-denormals -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global,-fp64-fp16-denormals -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
; GCN-LABEL: {{^}}fptrunc_f32_to_f16: ; GCN-LABEL: {{^}}fptrunc_f32_to_f16:
; GCN: buffer_load_dword v[[A_F32:[0-9]+]] ; GCN: buffer_load_dword v[[A_F32:[0-9]+]]
; GCN: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[A_F32]] ; GCN: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[A_F32]]
; GCN: buffer_store_short v[[R_F16]] ; GCN: buffer_store_short v[[R_F16]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @fptrunc_f32_to_f16( define amdgpu_kernel void @fptrunc_f32_to_f16(
half addrspace(1)* %r, half addrspace(1)* %r,
▲ Show 20 Lines • Show All 182 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fract.f64.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -enable-unsafe-fp-math -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=SI-UNSAFE -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -enable-unsafe-fp-math -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=SI-UNSAFE -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -enable-unsafe-fp-math -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=VI-UNSAFE -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -enable-unsafe-fp-math -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=VI-UNSAFE -check-prefix=FUNC %s
declare double @llvm.fabs.f64(double) #0 declare double @llvm.fabs.f64(double) #0
declare double @llvm.floor.f64(double) #0 declare double @llvm.floor.f64(double) #0
; FUNC-LABEL: {{^}}fract_f64: ; FUNC-LABEL: {{^}}fract_f64:
; SI-DAG: v_fract_f64_e32 [[FRC:v\[[0-9]+:[0-9]+\]]], v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]] ; SI-DAG: v_fract_f64_e32 [[FRC:v\[[0-9]+:[0-9]+\]]], v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]]
; SI-DAG: v_mov_b32_e32 v[[UPLO:[0-9]+]], -1 ; SI-DAG: v_mov_b32_e32 v[[UPLO:[0-9]+]], -1
; SI-DAG: v_mov_b32_e32 v[[UPHI:[0-9]+]], 0x3fefffff ; SI-DAG: v_mov_b32_e32 v[[UPHI:[0-9]+]], 0x3fefffff
▲ Show 20 Lines • Show All 97 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fract.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=CI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=CI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=GCN %s
declare float @llvm.fabs.f32(float) #0 declare float @llvm.fabs.f32(float) #0
declare float @llvm.floor.f32(float) #0 declare float @llvm.floor.f32(float) #0
; GCN-LABEL: {{^}}fract_f32: ; GCN-LABEL: {{^}}fract_f32:
; GCN-SAFE: v_floor_f32_e32 [[FLR:v[0-9]+]], [[INPUT:v[0-9]+]] ; GCN-SAFE: v_floor_f32_e32 [[FLR:v[0-9]+]], [[INPUT:v[0-9]+]]
; GCN-SAFE: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[FLR]], [[INPUT]] ; GCN-SAFE: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[FLR]], [[INPUT]]
▲ Show 20 Lines • Show All 61 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/frem.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=GCN -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}frem_f32: ; FUNC-LABEL: {{^}}frem_f32:
; GCN-DAG: buffer_load_dword [[X:v[0-9]+]], {{.*$}} ; GCN-DAG: buffer_load_dword [[X:v[0-9]+]], {{.*$}}
; GCN-DAG: buffer_load_dword [[Y:v[0-9]+]], {{.*}} offset:16 ; GCN-DAG: buffer_load_dword [[Y:v[0-9]+]], {{.*}} offset:16
; GCN: v_div_scale_f32 ; GCN: v_div_scale_f32
; GCN: v_rcp_f32_e32 ; GCN: v_rcp_f32_e32
; GCN: v_fma_f32 ; GCN: v_fma_f32
▲ Show 20 Lines • Show All 102 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fsqrt.f64.ll

; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}v_safe_fsqrt_f64: ; FUNC-LABEL: {{^}}v_safe_fsqrt_f64:
; GCN: v_sqrt_f64_e32 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}} ; GCN: v_sqrt_f64_e32 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
define amdgpu_kernel void @v_safe_fsqrt_f64(double addrspace(1)* %out, double addrspace(1)* %in) #1 { define amdgpu_kernel void @v_safe_fsqrt_f64(double addrspace(1)* %out, double addrspace(1)* %in) #1 {
%r0 = load double, double addrspace(1)* %in %r0 = load double, double addrspace(1)* %in
%r1 = call double @llvm.sqrt.f64(double %r0) %r1 = call double @llvm.sqrt.f64(double %r0)
store double %r1, double addrspace(1)* %out store double %r1, double addrspace(1)* %out
ret void ret void
Show All 16 Lines

llvm/trunk/test/CodeGen/AMDGPU/fsqrt.ll

; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
; Run with unsafe-fp-math to make sure nothing tries to turn this into 1 / rsqrt(x) ; Run with unsafe-fp-math to make sure nothing tries to turn this into 1 / rsqrt(x)
; FUNC-LABEL: {{^}}v_safe_fsqrt_f32: ; FUNC-LABEL: {{^}}v_safe_fsqrt_f32:
; GCN: v_sqrt_f32_e32 {{v[0-9]+, v[0-9]+}} ; GCN: v_sqrt_f32_e32 {{v[0-9]+, v[0-9]+}}
define amdgpu_kernel void @v_safe_fsqrt_f32(float addrspace(1)* %out, float addrspace(1)* %in) #1 { define amdgpu_kernel void @v_safe_fsqrt_f32(float addrspace(1)* %out, float addrspace(1)* %in) #1 {
%r0 = load float, float addrspace(1)* %in %r0 = load float, float addrspace(1)* %in
▲ Show 20 Lines • Show All 132 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fsub.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=VI -check-prefix=SIVI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=VI -check-prefix=SIVI %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s
; GCN-LABEL: {{^}}fsub_f16: ; GCN-LABEL: {{^}}fsub_f16:
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[B_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[B_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]] ; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]]
; SI: v_subrev_f32_e32 v[[R_F32:[0-9]+]], v[[B_F32]], v[[A_F32]] ; SI: v_subrev_f32_e32 v[[R_F32:[0-9]+]], v[[B_F32]], v[[A_F32]]
; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]] ; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]]
▲ Show 20 Lines • Show All 156 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fsub.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}v_fsub_f32: ; FUNC-LABEL: {{^}}v_fsub_f32:
; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}} ; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
define amdgpu_kernel void @v_fsub_f32(float addrspace(1)* %out, float addrspace(1)* %in) { define amdgpu_kernel void @v_fsub_f32(float addrspace(1)* %out, float addrspace(1)* %in) {
%b_ptr = getelementptr float, float addrspace(1)* %in, i32 1 %b_ptr = getelementptr float, float addrspace(1)* %in, i32 1
%a = load float, float addrspace(1)* %in, align 4 %a = load float, float addrspace(1)* %in, align 4
%b = load float, float addrspace(1)* %b_ptr, align 4 %b = load float, float addrspace(1)* %b_ptr, align 4
%result = fsub float %a, %b %result = fsub float %a, %b
▲ Show 20 Lines • Show All 123 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/fsub64.ll

; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
declare double @llvm.fabs.f64(double) #0 declare double @llvm.fabs.f64(double) #0
; SI-LABEL: {{^}}fsub_f64: ; SI-LABEL: {{^}}fsub_f64:
; SI: v_add_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], -v\[[0-9]+:[0-9]+\]}} ; SI: v_add_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], -v\[[0-9]+:[0-9]+\]}}
define amdgpu_kernel void @fsub_f64(double addrspace(1)* %out, double addrspace(1)* %in1, define amdgpu_kernel void @fsub_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) { double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1 %r0 = load double, double addrspace(1)* %in1
▲ Show 20 Lines • Show All 97 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/ftrunc.f64.ll

; RUN: llc -march=amdgcn -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s
declare double @llvm.trunc.f64(double) nounwind readnone declare double @llvm.trunc.f64(double) nounwind readnone
declare <2 x double> @llvm.trunc.v2f64(<2 x double>) nounwind readnone declare <2 x double> @llvm.trunc.v2f64(<2 x double>) nounwind readnone
declare <3 x double> @llvm.trunc.v3f64(<3 x double>) nounwind readnone declare <3 x double> @llvm.trunc.v3f64(<3 x double>) nounwind readnone
declare <4 x double> @llvm.trunc.v4f64(<4 x double>) nounwind readnone declare <4 x double> @llvm.trunc.v4f64(<4 x double>) nounwind readnone
declare <8 x double> @llvm.trunc.v8f64(<8 x double>) nounwind readnone declare <8 x double> @llvm.trunc.v8f64(<8 x double>) nounwind readnone
declare <16 x double> @llvm.trunc.v16f64(<16 x double>) nounwind readnone declare <16 x double> @llvm.trunc.v16f64(<16 x double>) nounwind readnone
▲ Show 20 Lines • Show All 100 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/global-extload-i16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; XUN: llc -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; XUN: llc -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FIXME: cypress is broken because the bigger testcases spill and it's not implemented ; FIXME: cypress is broken because the bigger testcases spill and it's not implemented
; FUNC-LABEL: {{^}}zextload_global_i16_to_i32: ; FUNC-LABEL: {{^}}zextload_global_i16_to_i32:
; SI: buffer_load_ushort ; SI: buffer_load_ushort
; SI: buffer_store_dword ; SI: buffer_store_dword
; SI: s_endpgm ; SI: s_endpgm
define amdgpu_kernel void @zextload_global_i16_to_i32(i32 addrspace(1)* %out, i16 addrspace(1)* %in) nounwind { define amdgpu_kernel void @zextload_global_i16_to_i32(i32 addrspace(1)* %out, i16 addrspace(1)* %in) nounwind {
▲ Show 20 Lines • Show All 292 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/half.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; half args should be promoted to float for SI and lower. ; half args should be promoted to float for SI and lower.
; GCN-LABEL: {{^}}load_f16_arg: ; GCN-LABEL: {{^}}load_f16_arg:
; GCN: s_load_dword [[ARG:s[0-9]+]] ; GCN: s_load_dword [[ARG:s[0-9]+]]
; SI: v_cvt_f16_f32_e32 [[CVT:v[0-9]+]], [[ARG]] ; SI: v_cvt_f16_f32_e32 [[CVT:v[0-9]+]], [[ARG]]
; VI: v_trunc_f16_e32 [[CVT:v[0-9]+]], [[ARG]] ; VI: v_trunc_f16_e32 [[CVT:v[0-9]+]], [[ARG]]
; GCN: buffer_store_short [[CVT]] ; GCN: buffer_store_short [[CVT]]
▲ Show 20 Lines • Show All 645 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/imm.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; Use a 64-bit value with lo bits that can be represented as an inline constant ; Use a 64-bit value with lo bits that can be represented as an inline constant
; GCN-LABEL: {{^}}i64_imm_inline_lo: ; GCN-LABEL: {{^}}i64_imm_inline_lo:
; GCN: v_mov_b32_e32 v[[LO_VGPR:[0-9]+]], 5 ; GCN: v_mov_b32_e32 v[[LO_VGPR:[0-9]+]], 5
; GCN: buffer_store_dwordx2 v{{\[}}[[LO_VGPR]]: ; GCN: buffer_store_dwordx2 v{{\[}}[[LO_VGPR]]:
define amdgpu_kernel void @i64_imm_inline_lo(i64 addrspace(1) *%out) { define amdgpu_kernel void @i64_imm_inline_lo(i64 addrspace(1) *%out) {
entry: entry:
store i64 1311768464867721221, i64 addrspace(1) *%out ; 0x1234567800000005 store i64 1311768464867721221, i64 addrspace(1) *%out ; 0x1234567800000005
▲ Show 20 Lines • Show All 674 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/immv216.ll

; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=kaveri -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
; FIXME: Merge into imm.ll ; FIXME: Merge into imm.ll
; GCN-LABEL: {{^}}store_inline_imm_neg_0.0_v2i16: ; GCN-LABEL: {{^}}store_inline_imm_neg_0.0_v2i16:
; GCN: v_mov_b32_e32 [[REG:v[0-9]+]], 0x80008000{{$}} ; GCN: v_mov_b32_e32 [[REG:v[0-9]+]], 0x80008000{{$}}
; GCN: buffer_store_dword [[REG]] ; GCN: buffer_store_dword [[REG]]
define amdgpu_kernel void @store_inline_imm_neg_0.0_v2i16(<2 x i16> addrspace(1)* %out) #0 { define amdgpu_kernel void @store_inline_imm_neg_0.0_v2i16(<2 x i16> addrspace(1)* %out) #0 {
store <2 x i16> <i16 -32768, i16 -32768>, <2 x i16> addrspace(1)* %out store <2 x i16> <i16 -32768, i16 -32768>, <2 x i16> addrspace(1)* %out
ret void ret void
▲ Show 20 Lines • Show All 452 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/indirect-addressing-si.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=MOVREL %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=MOVREL %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=MOVREL %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=MOVREL %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -amdgpu-vgpr-index-mode -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=IDXMODE %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -amdgpu-vgpr-index-mode -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=IDXMODE %s
; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=IDXMODE %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=IDXMODE %s
; Tests for indirect addressing on SI, which is implemented using dynamic ; Tests for indirect addressing on SI, which is implemented using dynamic
; indexing of vectors. ; indexing of vectors.
; GCN-LABEL: {{^}}extract_w_offset: ; GCN-LABEL: {{^}}extract_w_offset:
; GCN-DAG: s_load_dword [[IN:s[0-9]+]] ; GCN-DAG: s_load_dword [[IN:s[0-9]+]]
; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, 4.0 ; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, 4.0
; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, 0x40400000 ; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, 0x40400000
▲ Show 20 Lines • Show All 742 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/invariant-load-no-alias-store.ll

; RUN: llc -march=amdgcn -amdgpu-load-store-vectorizer=0 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -amdgpu-load-store-vectorizer=0 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; GatherAllAliases gives up on trying to analyze cases where the ; GatherAllAliases gives up on trying to analyze cases where the
; pointer may have been loaded from an aliased store, so make sure ; pointer may have been loaded from an aliased store, so make sure
; that this works and allows moving the stores to a better chain to ; that this works and allows moving the stores to a better chain to
; allow them to be merged merged when it's clear the pointer is loaded ; allow them to be merged merged when it's clear the pointer is loaded
; from constant/invariant memory. ; from constant/invariant memory.
; GCN-LABEL: {{^}}test_merge_store_constant_i16_invariant_global_pointer_load: ; GCN-LABEL: {{^}}test_merge_store_constant_i16_invariant_global_pointer_load:
Show All 26 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.amdgcn.class.f16.ll

; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.fabs.f16(half %a) declare half @llvm.fabs.f16(half %a)
declare i1 @llvm.amdgcn.class.f16(half %a, i32 %b) declare i1 @llvm.amdgcn.class.f16(half %a, i32 %b)
; GCN-LABEL: {{^}}class_f16 ; GCN-LABEL: {{^}}class_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: buffer_load_dword v[[B_I32:[0-9]+]] ; GCN: buffer_load_dword v[[B_I32:[0-9]+]]
; VI: v_cmp_class_f16_e32 vcc, v[[A_F16]], v[[B_I32]] ; VI: v_cmp_class_f16_e32 vcc, v[[A_F16]], v[[B_I32]]
▲ Show 20 Lines • Show All 146 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.amdgcn.class.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
declare i1 @llvm.amdgcn.class.f32(float, i32) #1 declare i1 @llvm.amdgcn.class.f32(float, i32) #1
declare i1 @llvm.amdgcn.class.f64(double, i32) #1 declare i1 @llvm.amdgcn.class.f64(double, i32) #1
declare i32 @llvm.amdgcn.workitem.id.x() #1 declare i32 @llvm.amdgcn.workitem.id.x() #1
declare float @llvm.fabs.f32(float) #1 declare float @llvm.fabs.f32(float) #1
declare double @llvm.fabs.f64(double) #1 declare double @llvm.fabs.f64(double) #1
; SI-LABEL: {{^}}test_class_f32: ; SI-LABEL: {{^}}test_class_f32:
▲ Show 20 Lines • Show All 502 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.amdgcn.div.fmas.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -strict-whitespace -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -strict-whitespace -check-prefix=GCN -check-prefix=SI %s
; XUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -strict-whitespace -check-prefix=GCN -check-prefix=VI %s ; XUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -strict-whitespace -check-prefix=GCN -check-prefix=VI %s
; FIXME: Enable for VI. ; FIXME: Enable for VI.
declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
declare float @llvm.amdgcn.div.fmas.f32(float, float, float, i1) nounwind readnone declare float @llvm.amdgcn.div.fmas.f32(float, float, float, i1) nounwind readnone
declare double @llvm.amdgcn.div.fmas.f64(double, double, double, i1) nounwind readnone declare double @llvm.amdgcn.div.fmas.f64(double, double, double, i1) nounwind readnone
▲ Show 20 Lines • Show All 169 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.amdgcn.ldexp.f16.ll

; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.amdgcn.ldexp.f16(half %a, i32 %b) declare half @llvm.amdgcn.ldexp.f16(half %a, i32 %b)
; GCN-LABEL: {{^}}ldexp_f16 ; GCN-LABEL: {{^}}ldexp_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: buffer_load_dword v[[B_I32:[0-9]+]] ; GCN: buffer_load_dword v[[B_I32:[0-9]+]]
; VI: v_ldexp_f16_e32 v[[R_F16:[0-9]+]], v[[A_F16]], v[[B_I32]] ; VI: v_ldexp_f16_e32 v[[R_F16:[0-9]+]], v[[A_F16]], v[[B_I32]]
; GCN: buffer_store_short v[[R_F16]] ; GCN: buffer_store_short v[[R_F16]]
Show All 36 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.amdgcn.sbfe.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}bfe_i32_arg_arg_arg: ; GCN-LABEL: {{^}}bfe_i32_arg_arg_arg:
; GCN: v_bfe_i32 ; GCN: v_bfe_i32
define amdgpu_kernel void @bfe_i32_arg_arg_arg(i32 addrspace(1)* %out, i32 %src0, i32 %src1, i32 %src2) #0 { define amdgpu_kernel void @bfe_i32_arg_arg_arg(i32 addrspace(1)* %out, i32 %src0, i32 %src1, i32 %src2) #0 {
%bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 %src0, i32 %src1, i32 %src1) %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 %src0, i32 %src1, i32 %src1)
store i32 %bfe_i32, i32 addrspace(1)* %out, align 4 store i32 %bfe_i32, i32 addrspace(1)* %out, align 4
ret void ret void
} }
▲ Show 20 Lines • Show All 546 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.amdgcn.sffbh.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
declare i32 @llvm.amdgcn.sffbh.i32(i32) #1 declare i32 @llvm.amdgcn.sffbh.i32(i32) #1
; FUNC-LABEL: {{^}}s_flbit: ; FUNC-LABEL: {{^}}s_flbit:
; GCN: s_load_dword [[VAL:s[0-9]+]], ; GCN: s_load_dword [[VAL:s[0-9]+]],
; GCN: s_flbit_i32 [[SRESULT:s[0-9]+]], [[VAL]] ; GCN: s_flbit_i32 [[SRESULT:s[0-9]+]], [[VAL]]
; GCN: v_mov_b32_e32 [[VRESULT:v[0-9]+]], [[SRESULT]] ; GCN: v_mov_b32_e32 [[VRESULT:v[0-9]+]], [[SRESULT]]
; GCN: buffer_store_dword [[VRESULT]], ; GCN: buffer_store_dword [[VRESULT]],
Show All 19 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.amdgcn.trig.preop.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
declare double @llvm.amdgcn.trig.preop.f64(double, i32) nounwind readnone declare double @llvm.amdgcn.trig.preop.f64(double, i32) nounwind readnone
; SI-LABEL: {{^}}test_trig_preop_f64: ; SI-LABEL: {{^}}test_trig_preop_f64:
; SI-DAG: buffer_load_dword [[SEG:v[0-9]+]] ; SI-DAG: buffer_load_dword [[SEG:v[0-9]+]]
; SI-DAG: buffer_load_dwordx2 [[SRC:v\[[0-9]+:[0-9]+\]]], ; SI-DAG: buffer_load_dwordx2 [[SRC:v\[[0-9]+:[0-9]+\]]],
; SI: v_trig_preop_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[SRC]], [[SEG]] ; SI: v_trig_preop_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[SRC]], [[SEG]]
; SI: buffer_store_dwordx2 [[RESULT]], ; SI: buffer_store_dwordx2 [[RESULT]],
Show All 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.amdgcn.ubfe.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}bfe_u32_arg_arg_arg: ; GCN-LABEL: {{^}}bfe_u32_arg_arg_arg:
; GCN: v_bfe_u32 ; GCN: v_bfe_u32
define amdgpu_kernel void @bfe_u32_arg_arg_arg(i32 addrspace(1)* %out, i32 %src0, i32 %src1, i32 %src2) #0 { define amdgpu_kernel void @bfe_u32_arg_arg_arg(i32 addrspace(1)* %out, i32 %src0, i32 %src1, i32 %src2) #0 {
%bfe_u32 = call i32 @llvm.amdgcn.ubfe.i32(i32 %src0, i32 %src1, i32 %src1) %bfe_u32 = call i32 @llvm.amdgcn.ubfe.i32(i32 %src0, i32 %src1, i32 %src1)
store i32 %bfe_u32, i32 addrspace(1)* %out, align 4 store i32 %bfe_u32, i32 addrspace(1)* %out, align 4
ret void ret void
} }
▲ Show 20 Lines • Show All 613 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.ceil.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.ceil.f16(half %a) declare half @llvm.ceil.f16(half %a)
declare <2 x half> @llvm.ceil.v2f16(<2 x half> %a) declare <2 x half> @llvm.ceil.v2f16(<2 x half> %a)
; GCN-LABEL: {{^}}ceil_f16: ; GCN-LABEL: {{^}}ceil_f16:
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_ceil_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]] ; SI: v_ceil_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]]
▲ Show 20 Lines • Show All 43 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.cos.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.cos.f16(half %a) declare half @llvm.cos.f16(half %a)
declare <2 x half> @llvm.cos.v2f16(<2 x half> %a) declare <2 x half> @llvm.cos.v2f16(<2 x half> %a)
; GCN-LABEL: {{^}}cos_f16 ; GCN-LABEL: {{^}}cos_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; GCN: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; GCN: v_mul_f32_e32 v[[M_F32:[0-9]+]], {{0.15915494|0x3e22f983}}, v[[A_F32]] ; GCN: v_mul_f32_e32 v[[M_F32:[0-9]+]], {{0.15915494|0x3e22f983}}, v[[A_F32]]
▲ Show 20 Lines • Show All 53 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.exp2.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.exp2.f16(half %a) declare half @llvm.exp2.f16(half %a)
declare <2 x half> @llvm.exp2.v2f16(<2 x half> %a) declare <2 x half> @llvm.exp2.v2f16(<2 x half> %a)
; GCN-LABEL: {{^}}exp2_f16 ; GCN-LABEL: {{^}}exp2_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_exp_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]] ; SI: v_exp_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]]
▲ Show 20 Lines • Show All 43 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.floor.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.floor.f16(half %a) declare half @llvm.floor.f16(half %a)
declare <2 x half> @llvm.floor.v2f16(<2 x half> %a) declare <2 x half> @llvm.floor.v2f16(<2 x half> %a)
; GCN-LABEL: {{^}}floor_f16 ; GCN-LABEL: {{^}}floor_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_floor_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]] ; SI: v_floor_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]]
▲ Show 20 Lines • Show All 43 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.fma.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.fma.f16(half %a, half %b, half %c) declare half @llvm.fma.f16(half %a, half %b, half %c)
declare <2 x half> @llvm.fma.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c) declare <2 x half> @llvm.fma.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c)
; GCN-LABEL: {{^}}fma_f16 ; GCN-LABEL: {{^}}fma_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[B_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[B_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[C_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[C_F16:[0-9]+]]
▲ Show 20 Lines • Show All 258 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.fmuladd.f16.ll

; RUN: llc -march=amdgcn -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SI-FLUSH %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SI-FLUSH %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=VI-FLUSH %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=VI-FLUSH %s
; RUN: llc -march=amdgcn -mattr=+fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SI-DENORM %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=+fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SI-DENORM %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=+fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=VI-DENORM %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=+fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=VI-DENORM %s
declare half @llvm.fmuladd.f16(half %a, half %b, half %c) declare half @llvm.fmuladd.f16(half %a, half %b, half %c)
declare <2 x half> @llvm.fmuladd.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c) declare <2 x half> @llvm.fmuladd.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c)
; GCN-LABEL: {{^}}fmuladd_f16 ; GCN-LABEL: {{^}}fmuladd_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[B_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[B_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[C_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[C_F16:[0-9]+]]
▲ Show 20 Lines • Show All 138 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.log2.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.log2.f16(half %a) declare half @llvm.log2.f16(half %a)
declare <2 x half> @llvm.log2.v2f16(<2 x half> %a) declare <2 x half> @llvm.log2.v2f16(<2 x half> %a)
; GCN-LABEL: {{^}}log2_f16 ; GCN-LABEL: {{^}}log2_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_log_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]] ; SI: v_log_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]]
▲ Show 20 Lines • Show All 43 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.maxnum.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.maxnum.f16(half %a, half %b) declare half @llvm.maxnum.f16(half %a, half %b)
declare <2 x half> @llvm.maxnum.v2f16(<2 x half> %a, <2 x half> %b) declare <2 x half> @llvm.maxnum.v2f16(<2 x half> %a, <2 x half> %b)
; GCN-LABEL: {{^}}maxnum_f16: ; GCN-LABEL: {{^}}maxnum_f16:
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[B_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[B_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
▲ Show 20 Lines • Show All 147 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.minnum.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.minnum.f16(half %a, half %b) declare half @llvm.minnum.f16(half %a, half %b)
declare <2 x half> @llvm.minnum.v2f16(<2 x half> %a, <2 x half> %b) declare <2 x half> @llvm.minnum.v2f16(<2 x half> %a, <2 x half> %b)
; GCN-LABEL: {{^}}minnum_f16: ; GCN-LABEL: {{^}}minnum_f16:
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[B_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[B_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
▲ Show 20 Lines • Show All 148 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.rint.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SIVI -check-prefix=VI -check-prefix=GFX89 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SIVI -check-prefix=VI -check-prefix=GFX89 %s
; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s
declare half @llvm.rint.f16(half %a) declare half @llvm.rint.f16(half %a)
declare <2 x half> @llvm.rint.v2f16(<2 x half> %a) declare <2 x half> @llvm.rint.v2f16(<2 x half> %a)
; GCN-LABEL: {{^}}rint_f16 ; GCN-LABEL: {{^}}rint_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_rndne_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]] ; SI: v_rndne_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]]
▲ Show 20 Lines • Show All 49 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.sin.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.sin.f16(half %a) declare half @llvm.sin.f16(half %a)
declare <2 x half> @llvm.sin.v2f16(<2 x half> %a) declare <2 x half> @llvm.sin.v2f16(<2 x half> %a)
; GCN-LABEL: {{^}}sin_f16 ; GCN-LABEL: {{^}}sin_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; GCN: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; GCN: v_mul_f32_e32 v[[M_F32:[0-9]+]], {{0.15915494|0x3e22f983}}, v[[A_F32]] ; GCN: v_mul_f32_e32 v[[M_F32:[0-9]+]], {{0.15915494|0x3e22f983}}, v[[A_F32]]
▲ Show 20 Lines • Show All 55 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.sqrt.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.sqrt.f16(half %a) declare half @llvm.sqrt.f16(half %a)
declare <2 x half> @llvm.sqrt.v2f16(<2 x half> %a) declare <2 x half> @llvm.sqrt.v2f16(<2 x half> %a)
; GCN-LABEL: {{^}}sqrt_f16 ; GCN-LABEL: {{^}}sqrt_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_sqrt_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]] ; SI: v_sqrt_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]]
▲ Show 20 Lines • Show All 43 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/llvm.trunc.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.trunc.f16(half %a) declare half @llvm.trunc.f16(half %a)
declare <2 x half> @llvm.trunc.v2f16(<2 x half> %a) declare <2 x half> @llvm.trunc.v2f16(<2 x half> %a)
; GCN-LABEL: {{^}}trunc_f16 ; GCN-LABEL: {{^}}trunc_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_trunc_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]] ; SI: v_trunc_f32_e32 v[[R_F32:[0-9]+]], v[[A_F32]]
▲ Show 20 Lines • Show All 43 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/load-global-f32.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
; RUN: llc -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}global_load_f32: ; FUNC-LABEL: {{^}}global_load_f32:
; GCN-NOHSA: buffer_load_dword v{{[0-9]+}} ; GCN-NOHSA: buffer_load_dword v{{[0-9]+}}
; GCN-HSA: flat_load_dword ; GCN-HSA: flat_load_dword
; R600: VTX_READ_32 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0 ; R600: VTX_READ_32 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0
define amdgpu_kernel void @global_load_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 { define amdgpu_kernel void @global_load_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
entry: entry:
▲ Show 20 Lines • Show All 79 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/load-global-f64.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
; RUN: llc -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}global_load_f64: ; FUNC-LABEL: {{^}}global_load_f64:
; GCN-NOHSA: buffer_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]] ; GCN-NOHSA: buffer_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]]
; GCN-NOHSA: buffer_store_dwordx2 [[VAL]] ; GCN-NOHSA: buffer_store_dwordx2 [[VAL]]
; GCN-HSA: flat_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]] ; GCN-HSA: flat_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]]
; GCN-HSA: flat_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, [[VAL]] ; GCN-HSA: flat_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, [[VAL]]
define amdgpu_kernel void @global_load_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 { define amdgpu_kernel void @global_load_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
▲ Show 20 Lines • Show All 83 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/load-global-i16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,GCN-NOHSA-SI,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,GCN-NOHSA-SI,FUNC %s
; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-HSA,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-HSA,FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,GCN-NOHSA-VI,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,GCN-NOHSA-VI,FUNC %s
; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=EGCM -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=EGCM -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cayman -verify-machineinstrs < %s | FileCheck -check-prefix=CM -check-prefix=EGCM -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cayman -verify-machineinstrs < %s | FileCheck -check-prefix=CM -check-prefix=EGCM -check-prefix=FUNC %s
; FIXME: r600 is broken because the bigger testcases spill and it's not implemented ; FIXME: r600 is broken because the bigger testcases spill and it's not implemented
; FUNC-LABEL: {{^}}global_load_i16: ; FUNC-LABEL: {{^}}global_load_i16:
; GCN-NOHSA: buffer_load_ushort v{{[0-9]+}} ; GCN-NOHSA: buffer_load_ushort v{{[0-9]+}}
; GCN-HSA: flat_load_ushort ; GCN-HSA: flat_load_ushort
; EGCM: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1 ; EGCM: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
▲ Show 20 Lines • Show All 638 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/load-global-i32.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}global_load_i32: ; FUNC-LABEL: {{^}}global_load_i32:
; GCN-NOHSA: buffer_load_dword v{{[0-9]+}} ; GCN-NOHSA: buffer_load_dword v{{[0-9]+}}
; GCN-HSA: flat_load_dword ; GCN-HSA: flat_load_dword
; EG: VTX_READ_32 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0 ; EG: VTX_READ_32 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0
define amdgpu_kernel void @global_load_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 { define amdgpu_kernel void @global_load_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
▲ Show 20 Lines • Show All 509 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/load-global-i64.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}global_load_i64: ; FUNC-LABEL: {{^}}global_load_i64:
; GCN-NOHSA: buffer_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]] ; GCN-NOHSA: buffer_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]]
; GCN-NOHSA: buffer_store_dwordx2 [[VAL]] ; GCN-NOHSA: buffer_store_dwordx2 [[VAL]]
; GCN-HSA: flat_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]] ; GCN-HSA: flat_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]]
; GCN-HSA: flat_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, [[VAL]] ; GCN-HSA: flat_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, [[VAL]]
▲ Show 20 Lines • Show All 108 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/load-global-i8.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,SI,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,SI,FUNC %s
; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-HSA,SI,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-HSA,SI,FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,VI,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,VI,FUNC %s
; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cayman -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cayman -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}global_load_i8: ; FUNC-LABEL: {{^}}global_load_i8:
; GCN-NOHSA: buffer_load_ubyte v{{[0-9]+}} ; GCN-NOHSA: buffer_load_ubyte v{{[0-9]+}}
; GCN-HSA: flat_load_ubyte ; GCN-HSA: flat_load_ubyte
; EG: VTX_READ_8 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1 ; EG: VTX_READ_8 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
; TODO: NOT AND ; TODO: NOT AND
▲ Show 20 Lines • Show All 966 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/load-weird-sizes.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NOHSA -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NOHSA -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=CI-HSA -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=CI-HSA -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NOHSA -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NOHSA -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=EG -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=R600 -check-prefix=CM -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=R600 -check-prefix=CM -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}load_i24: ; FUNC-LABEL: {{^}}load_i24:
; SI: {{flat|buffer}}_load_ubyte ; SI: {{flat|buffer}}_load_ubyte
; SI: {{flat|buffer}}_load_ushort ; SI: {{flat|buffer}}_load_ushort
; SI: {{flat|buffer}}_store_dword ; SI: {{flat|buffer}}_store_dword
define amdgpu_kernel void @load_i24(i32 addrspace(1)* %out, i24 addrspace(1)* %in) #0 { define amdgpu_kernel void @load_i24(i32 addrspace(1)* %out, i24 addrspace(1)* %in) #0 {
%1 = load i24, i24 addrspace(1)* %in %1 = load i24, i24 addrspace(1)* %in
%2 = zext i24 %1 to i32 %2 = zext i24 %1 to i32
Show All 18 Lines

llvm/trunk/test/CodeGen/AMDGPU/max.ll

; RUN: llc -march=amdgcn < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}v_test_imax_sge_i32: ; FUNC-LABEL: {{^}}v_test_imax_sge_i32:
; SI: v_max_i32_e32 ; SI: v_max_i32_e32
; EG: MAX_INT ; EG: MAX_INT
define amdgpu_kernel void @v_test_imax_sge_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr, i32 addrspace(1)* %bptr) nounwind { define amdgpu_kernel void @v_test_imax_sge_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr, i32 addrspace(1)* %bptr) nounwind {
%a = load i32, i32 addrspace(1)* %aptr, align 4 %a = load i32, i32 addrspace(1)* %aptr, align 4
▲ Show 20 Lines • Show All 309 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/merge-stores.ll

; RUN: llc -march=amdgcn -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
; This test is mostly to test DAG store merging, so disable the vectorizer. ; This test is mostly to test DAG store merging, so disable the vectorizer.
; Run with devices with different unaligned load restrictions. ; Run with devices with different unaligned load restrictions.
; TODO: Vector element tests ; TODO: Vector element tests
; TODO: Non-zero base offset for load and store combinations ; TODO: Non-zero base offset for load and store combinations
; TODO: Same base addrspacecasted ; TODO: Same base addrspacecasted
▲ Show 20 Lines • Show All 682 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/mubuf.ll

; RUN: llc -march=amdgcn -show-mc-encoding -verify-machineinstrs < %s | FileCheck %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -show-mc-encoding -verify-machineinstrs < %s | FileCheck %s
declare i32 @llvm.amdgcn.workitem.id.x() readnone declare i32 @llvm.amdgcn.workitem.id.x() readnone
;;;==========================================================================;;; ;;;==========================================================================;;;
;;; MUBUF LOAD TESTS ;;; MUBUF LOAD TESTS
;;;==========================================================================;;; ;;;==========================================================================;;;
; MUBUF load with an immediate byte offset that fits into 12-bits ; MUBUF load with an immediate byte offset that fits into 12-bits
▲ Show 20 Lines • Show All 173 Lines • Show Last 20 Lines

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

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG %s -check-prefix=FUNC ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG %s -check-prefix=FUNC
; mul24 and mad24 are affected ; mul24 and mad24 are affected
; FUNC-LABEL: {{^}}test_mul_v2i32: ; FUNC-LABEL: {{^}}test_mul_v2i32:
; EG: MULLO_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: MULLO_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; EG: MULLO_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: MULLO_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; SI: v_mul_lo_i32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: v_mul_lo_i32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
▲ Show 20 Lines • Show All 260 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/no-shrink-extloads.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
; Make sure we don't turn the 32-bit argument load into a 16-bit ; Make sure we don't turn the 32-bit argument load into a 16-bit
; load. There aren't extending scalar lods, so that would require ; load. There aren't extending scalar lods, so that would require
; using a buffer_load instruction. ; using a buffer_load instruction.
; FUNC-LABEL: {{^}}truncate_kernarg_i32_to_i16: ; FUNC-LABEL: {{^}}truncate_kernarg_i32_to_i16:
▲ Show 20 Lines • Show All 206 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/or.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}or_v2i32: ; FUNC-LABEL: {{^}}or_v2i32:
; EG: OR_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: OR_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; EG: OR_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: OR_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; SI: v_or_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: v_or_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
; SI: v_or_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: v_or_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
▲ Show 20 Lines • Show All 261 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/promote-alloca-invariant-markers.ll

; RUN: llc -march=amdgcn -mattr=+promote-alloca -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=+promote-alloca -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
declare {}* @llvm.invariant.start.p0i8(i64, i8* nocapture) #0 declare {}* @llvm.invariant.start.p0i8(i64, i8* nocapture) #0
declare void @llvm.invariant.end.p0i8({}*, i64, i8* nocapture) #0 declare void @llvm.invariant.end.p0i8({}*, i64, i8* nocapture) #0
declare i8* @llvm.invariant.group.barrier(i8*) #1 declare i8* @llvm.invariant.group.barrier(i8*) #1
; GCN-LABEL: {{^}}use_invariant_promotable_lds: ; GCN-LABEL: {{^}}use_invariant_promotable_lds:
; GCN: buffer_load_dword ; GCN: buffer_load_dword
; GCN: ds_write_b32 ; GCN: ds_write_b32
Show All 16 Lines

llvm/trunk/test/CodeGen/AMDGPU/reduce-load-width-alignment.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}reduce_i64_load_align_4_width_to_i32: ; GCN-LABEL: {{^}}reduce_i64_load_align_4_width_to_i32:
; GCN: buffer_load_dword [[VAL:v[0-9]+]] ; GCN: buffer_load_dword [[VAL:v[0-9]+]]
; GCN: v_and_b32_e32 v{{[0-9]+}}, 0x12d687, [[VAL]] ; GCN: v_and_b32_e32 v{{[0-9]+}}, 0x12d687, [[VAL]]
; GCN: buffer_store_dwordx2 ; GCN: buffer_store_dwordx2
define amdgpu_kernel void @reduce_i64_load_align_4_width_to_i32(i64 addrspace(1)* %out, i64 addrspace(1)* %in) #0 { define amdgpu_kernel void @reduce_i64_load_align_4_width_to_i32(i64 addrspace(1)* %out, i64 addrspace(1)* %in) #0 {
%a = load i64, i64 addrspace(1)* %in, align 4 %a = load i64, i64 addrspace(1)* %in, align 4
%and = and i64 %a, 1234567 %and = and i64 %a, 1234567
Show All 27 Lines

llvm/trunk/test/CodeGen/AMDGPU/reorder-stores.ll

; RUN: llc -march=amdgcn < %s | FileCheck -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn < %s | FileCheck -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI %s
; SI-LABEL: {{^}}no_reorder_v2f64_global_load_store: ; SI-LABEL: {{^}}no_reorder_v2f64_global_load_store:
; SI: buffer_load_dwordx4 ; SI: buffer_load_dwordx4
; SI: buffer_load_dwordx4 ; SI: buffer_load_dwordx4
; SI: buffer_store_dwordx4 ; SI: buffer_store_dwordx4
; SI: buffer_store_dwordx4 ; SI: buffer_store_dwordx4
; SI: s_endpgm ; SI: s_endpgm
define amdgpu_kernel void @no_reorder_v2f64_global_load_store(<2 x double> addrspace(1)* nocapture %x, <2 x double> addrspace(1)* nocapture %y) nounwind { define amdgpu_kernel void @no_reorder_v2f64_global_load_store(<2 x double> addrspace(1)* nocapture %x, <2 x double> addrspace(1)* nocapture %y) nounwind {
▲ Show 20 Lines • Show All 59 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/rotl.i64.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=BOTH %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=BOTH %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=BOTH %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=BOTH %s
; BOTH-LABEL: {{^}}s_rotl_i64: ; BOTH-LABEL: {{^}}s_rotl_i64:
; BOTH-DAG: s_lshl_b64 ; BOTH-DAG: s_lshl_b64
; BOTH-DAG: s_sub_i32 ; BOTH-DAG: s_sub_i32
; BOTH-DAG: s_lshr_b64 ; BOTH-DAG: s_lshr_b64
; BOTH: s_or_b64 ; BOTH: s_or_b64
; BOTH: s_endpgm ; BOTH: s_endpgm
define amdgpu_kernel void @s_rotl_i64(i64 addrspace(1)* %in, i64 %x, i64 %y) { define amdgpu_kernel void @s_rotl_i64(i64 addrspace(1)* %in, i64 %x, i64 %y) {
Show All 29 Lines

llvm/trunk/test/CodeGen/AMDGPU/rotr.i64.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=BOTH %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=BOTH %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=BOTH %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=BOTH %s
; BOTH-LABEL: {{^}}s_rotr_i64: ; BOTH-LABEL: {{^}}s_rotr_i64:
; BOTH-DAG: s_sub_i32 ; BOTH-DAG: s_sub_i32
; BOTH-DAG: s_lshr_b64 ; BOTH-DAG: s_lshr_b64
; BOTH-DAG: s_lshl_b64 ; BOTH-DAG: s_lshl_b64
; BOTH: s_or_b64 ; BOTH: s_or_b64
define amdgpu_kernel void @s_rotr_i64(i64 addrspace(1)* %in, i64 %x, i64 %y) { define amdgpu_kernel void @s_rotr_i64(i64 addrspace(1)* %in, i64 %x, i64 %y) {
entry: entry:
▲ Show 20 Lines • Show All 51 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/rsq.ll

; RUN: llc -march=amdgcn -mattr=-fp32-denormals -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=SI-UNSAFE -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=-fp32-denormals -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=SI-UNSAFE -check-prefix=SI %s
; RUN: llc -march=amdgcn -mattr=-fp32-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=SI-SAFE -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=-fp32-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=SI-SAFE -check-prefix=SI %s
declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
declare float @llvm.sqrt.f32(float) nounwind readnone declare float @llvm.sqrt.f32(float) nounwind readnone
declare double @llvm.sqrt.f64(double) nounwind readnone declare double @llvm.sqrt.f64(double) nounwind readnone
; SI-LABEL: {{^}}rsq_f32: ; SI-LABEL: {{^}}rsq_f32:
; SI: v_rsq_f32_e32 ; SI: v_rsq_f32_e32
; SI: s_endpgm ; SI: s_endpgm
▲ Show 20 Lines • Show All 128 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/s_movk_i32.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
; SI-LABEL: {{^}}s_movk_i32_k0: ; SI-LABEL: {{^}}s_movk_i32_k0:
; SI-DAG: s_mov_b32 [[LO_S_IMM:s[0-9]+]], 0xffff{{$}} ; SI-DAG: s_mov_b32 [[LO_S_IMM:s[0-9]+]], 0xffff{{$}}
; SI-DAG: buffer_load_dwordx2 v{{\[}}[[LO_VREG:[0-9]+]]:[[HI_VREG:[0-9]+]]{{\]}}, ; SI-DAG: buffer_load_dwordx2 v{{\[}}[[LO_VREG:[0-9]+]]:[[HI_VREG:[0-9]+]]{{\]}},
; SI-DAG: v_or_b32_e32 {{v[0-9]+}}, [[LO_S_IMM]], v[[LO_VREG]] ; SI-DAG: v_or_b32_e32 {{v[0-9]+}}, [[LO_S_IMM]], v[[LO_VREG]]
; SI-DAG: v_or_b32_e32 {{v[0-9]+}}, 1, v[[HI_VREG]] ; SI-DAG: v_or_b32_e32 {{v[0-9]+}}, 1, v[[HI_VREG]]
; SI: s_endpgm ; SI: s_endpgm
define amdgpu_kernel void @s_movk_i32_k0(i64 addrspace(1)* %out, i64 addrspace(1)* %a, i64 addrspace(1)* %b) { define amdgpu_kernel void @s_movk_i32_k0(i64 addrspace(1)* %out, i64 addrspace(1)* %a, i64 addrspace(1)* %b) {
▲ Show 20 Lines • Show All 181 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/saddo.ll

; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=SI -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs< %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs< %s
declare { i32, i1 } @llvm.sadd.with.overflow.i32(i32, i32) nounwind readnone declare { i32, i1 } @llvm.sadd.with.overflow.i32(i32, i32) nounwind readnone
declare { i64, i1 } @llvm.sadd.with.overflow.i64(i64, i64) nounwind readnone declare { i64, i1 } @llvm.sadd.with.overflow.i64(i64, i64) nounwind readnone
; FUNC-LABEL: {{^}}saddo_i64_zext: ; FUNC-LABEL: {{^}}saddo_i64_zext:
define amdgpu_kernel void @saddo_i64_zext(i64 addrspace(1)* %out, i64 %a, i64 %b) nounwind { define amdgpu_kernel void @saddo_i64_zext(i64 addrspace(1)* %out, i64 %a, i64 %b) nounwind {
%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
▲ Show 20 Lines • Show All 52 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/salu-to-valu.ll

; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=CI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=CI %s
; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI --check-prefix=GCN-HSA %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI --check-prefix=GCN-HSA %s
declare i32 @llvm.amdgcn.workitem.id.x() #0 declare i32 @llvm.amdgcn.workitem.id.x() #0
declare i32 @llvm.amdgcn.workitem.id.y() #0 declare i32 @llvm.amdgcn.workitem.id.y() #0
; In this test both the pointer and the offset operands to the ; In this test both the pointer and the offset operands to the
; BUFFER_LOAD instructions end up being stored in vgprs. This ; BUFFER_LOAD instructions end up being stored in vgprs. This
; requires us to add the pointer and offset together, store the ; requires us to add the pointer and offset together, store the
; result in the offset operand (vaddr), and then store 0 in an ; result in the offset operand (vaddr), and then store 0 in an
▲ Show 20 Lines • Show All 496 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/scalar_to_vector.ll

; RUN: llc -march=amdgcn -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; XXX - Why the packing? ; XXX - Why the packing?
; GCN-LABEL: {{^}}scalar_to_vector_v2i32: ; GCN-LABEL: {{^}}scalar_to_vector_v2i32:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], ; GCN: buffer_load_dword [[VAL:v[0-9]+]],
; GCN: v_lshrrev_b32_e32 [[SHR:v[0-9]+]], 16, [[VAL]] ; GCN: v_lshrrev_b32_e32 [[SHR:v[0-9]+]], 16, [[VAL]]
; GCN: v_lshlrev_b32_e32 [[SHL:v[0-9]+]], 16, [[SHR]] ; GCN: v_lshlrev_b32_e32 [[SHL:v[0-9]+]], 16, [[SHR]]
; GCN: v_or_b32_e32 v[[OR:[0-9]+]], [[SHL]], [[SHR]] ; GCN: v_or_b32_e32 v[[OR:[0-9]+]], [[SHL]], [[SHR]]
; GCN: v_mov_b32_e32 v[[COPY:[0-9]+]], v[[OR]] ; GCN: v_mov_b32_e32 v[[COPY:[0-9]+]], v[[OR]]
▲ Show 20 Lines • Show All 65 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/schedule-global-loads.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s
; FIXME: This currently doesn't do a great job of clustering the ; FIXME: This currently doesn't do a great job of clustering the
; loads, which end up with extra moves between them. Right now, it ; loads, which end up with extra moves between them. Right now, it
; seems the only things areLoadsFromSameBasePtr is accomplishing is ; seems the only things areLoadsFromSameBasePtr is accomplishing is
; ordering the loads so that the lower address loads come first. ; ordering the loads so that the lower address loads come first.
; FUNC-LABEL: {{^}}cluster_global_arg_loads: ; FUNC-LABEL: {{^}}cluster_global_arg_loads:
; SI-DAG: buffer_load_dword [[REG0:v[0-9]+]], off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}} ; SI-DAG: buffer_load_dword [[REG0:v[0-9]+]], off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
Show All 29 Lines

llvm/trunk/test/CodeGen/AMDGPU/scratch-buffer.ll

; RUN: llc -verify-machineinstrs -march=amdgcn < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -march=amdgcn < %s | FileCheck -check-prefix=GCN %s
; RUN: llc -verify-machineinstrs -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=GCN %s
; When a frame index offset is more than 12-bits, make sure we don't store ; When a frame index offset is more than 12-bits, make sure we don't store
; it in mubuf's offset field. ; it in mubuf's offset field.
; Also, make sure we use the same register for storing the scratch buffer addresss ; Also, make sure we use the same register for storing the scratch buffer addresss
; for both stores. This register is allocated by the register scavenger, so we ; for both stores. This register is allocated by the register scavenger, so we
; should be able to reuse the same regiser for each scratch buffer access. ; should be able to reuse the same regiser for each scratch buffer access.
▲ Show 20 Lines • Show All 101 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/sdiv.ll

; RUN: llc -march=amdgcn < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; The code generated by sdiv is long and complex and may frequently change. ; The code generated by sdiv is long and complex and may frequently change.
; The goal of this test is to make sure the ISel doesn't fail. ; The goal of this test is to make sure the ISel doesn't fail.
; ;
; This program was previously failing to compile when one of the selectcc ; This program was previously failing to compile when one of the selectcc
; opcodes generated by the sdiv lowering was being legalized and optimized to: ; opcodes generated by the sdiv lowering was being legalized and optimized to:
; selectcc Remainder -1, 0, -1, SETGT ; selectcc Remainder -1, 0, -1, SETGT
; This was fixed by adding an additional pattern in R600Instructions.td to ; This was fixed by adding an additional pattern in R600Instructions.td to
▲ Show 20 Lines • Show All 160 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/sdwa-peephole.ll

; RUN: llc -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole=0 -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=NOSDWA -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole=0 -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=NOSDWA -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=SDWA -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=SDWA -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-sdwa-peephole -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 -check-prefix=SDWA -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -amdgpu-sdwa-peephole -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 -check-prefix=SDWA -check-prefix=GCN %s
; GCN-LABEL: {{^}}add_shr_i32: ; GCN-LABEL: {{^}}add_shr_i32:
; NOSDWA: v_lshrrev_b32_e32 v[[DST:[0-9]+]], 16, v{{[0-9]+}} ; NOSDWA: v_lshrrev_b32_e32 v[[DST:[0-9]+]], 16, v{{[0-9]+}}
; NOSDWA: v_add_i32_e32 v{{[0-9]+}}, vcc, v{{[0-9]+}}, v[[DST]] ; NOSDWA: v_add_i32_e32 v{{[0-9]+}}, vcc, v{{[0-9]+}}, v[[DST]]
; NOSDWA-NOT: v_add_i32_sdwa ; NOSDWA-NOT: v_add_i32_sdwa
; SDWA: v_add_i32_sdwa v{{[0-9]+}}, vcc, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; SDWA: v_add_i32_sdwa v{{[0-9]+}}, vcc, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
▲ Show 20 Lines • Show All 487 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/select-vectors.ll

; RUN: llc -verify-machineinstrs -march=amdgcn < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -march=amdgcn < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -verify-machineinstrs -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; RUN: llc -verify-machineinstrs -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s ; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
; Test expansion of scalar selects on vectors. ; Test expansion of scalar selects on vectors.
; Evergreen not enabled since it seems to be having problems with doubles. ; Evergreen not enabled since it seems to be having problems with doubles.
; GCN-LABEL: {{^}}v_select_v2i8: ; GCN-LABEL: {{^}}v_select_v2i8:
; SI: v_cndmask_b32 ; SI: v_cndmask_b32
; SI-NOT: cndmask ; SI-NOT: cndmask
▲ Show 20 Lines • Show All 363 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/select.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}select_f16: ; GCN-LABEL: {{^}}select_f16:
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[B_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[B_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[C_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[C_F16:[0-9]+]]
; GCN: buffer_load_ushort v[[D_F16:[0-9]+]] ; GCN: buffer_load_ushort v[[D_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]] ; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]]
▲ Show 20 Lines • Show All 314 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/sext-in-reg.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=GFX89 -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=GFX89 -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -check-prefix=GFX89 -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -check-prefix=GFX89 -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FIXME: i16 promotion pass ruins the scalar cases when legal. ; FIXME: i16 promotion pass ruins the scalar cases when legal.
; FIXME: r600 fails verifier ; FIXME: r600 fails verifier
; FUNC-LABEL: {{^}}sext_in_reg_i1_i32: ; FUNC-LABEL: {{^}}sext_in_reg_i1_i32:
; GCN: s_load_dword [[ARG:s[0-9]+]], ; GCN: s_load_dword [[ARG:s[0-9]+]],
; GCN: s_bfe_i32 [[SEXTRACT:s[0-9]+]], [[ARG]], 0x10000 ; GCN: s_bfe_i32 [[SEXTRACT:s[0-9]+]], [[ARG]], 0x10000
; GCN: v_mov_b32_e32 [[EXTRACT:v[0-9]+]], [[SEXTRACT]] ; GCN: v_mov_b32_e32 [[EXTRACT:v[0-9]+]], [[SEXTRACT]]
▲ Show 20 Lines • Show All 709 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/sgpr-copy-duplicate-operand.ll

; RUN: llc -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI %s
; Copy VGPR -> SGPR used twice as an instruction operand, which is then ; Copy VGPR -> SGPR used twice as an instruction operand, which is then
; used in an REG_SEQUENCE that also needs to be handled. ; used in an REG_SEQUENCE that also needs to be handled.
; SI-LABEL: {{^}}test_dup_operands: ; SI-LABEL: {{^}}test_dup_operands:
; SI: v_add_i32_e32 ; SI: v_add_i32_e32
define amdgpu_kernel void @test_dup_operands(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %in) { define amdgpu_kernel void @test_dup_operands(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %in) {
%a = load <2 x i32>, <2 x i32> addrspace(1)* %in %a = load <2 x i32>, <2 x i32> addrspace(1)* %in
Show All 9 Lines

llvm/trunk/test/CodeGen/AMDGPU/sgpr-copy.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck %s
; CHECK-LABEL: {{^}}phi1: ; CHECK-LABEL: {{^}}phi1:
; CHECK: s_buffer_load_dword [[DST:s[0-9]]], {{s\[[0-9]+:[0-9]+\]}}, 0x0 ; CHECK: s_buffer_load_dword [[DST:s[0-9]]], {{s\[[0-9]+:[0-9]+\]}}, 0x0
; CHECK: v_mov_b32_e32 v{{[0-9]}}, [[DST]] ; CHECK: v_mov_b32_e32 v{{[0-9]}}, [[DST]]
define amdgpu_ps void @phi1(<4 x i32> addrspace(2)* inreg %arg, <4 x i32> addrspace(2)* inreg %arg1, <8 x i32> addrspace(2)* inreg %arg2, i32 inreg %arg3, <2 x i32> %arg4, <2 x i32> %arg5, <2 x i32> %arg6, <3 x i32> %arg7, <2 x i32> %arg8, <2 x i32> %arg9, <2 x i32> %arg10, float %arg11, float %arg12, float %arg13, float %arg14, float %arg15, float %arg16, float %arg17, float %arg18, float %arg19) #0 { define amdgpu_ps void @phi1(<4 x i32> addrspace(2)* inreg %arg, <4 x i32> addrspace(2)* inreg %arg1, <8 x i32> addrspace(2)* inreg %arg2, i32 inreg %arg3, <2 x i32> %arg4, <2 x i32> %arg5, <2 x i32> %arg6, <3 x i32> %arg7, <2 x i32> %arg8, <2 x i32> %arg9, <2 x i32> %arg10, float %arg11, float %arg12, float %arg13, float %arg14, float %arg15, float %arg16, float %arg17, float %arg18, float %arg19) #0 {
main_body: main_body:
%tmp = getelementptr <4 x i32>, <4 x i32> addrspace(2)* %arg, i32 0 %tmp = getelementptr <4 x i32>, <4 x i32> addrspace(2)* %arg, i32 0
%tmp20 = load <4 x i32>, <4 x i32> addrspace(2)* %tmp, !tbaa !0 %tmp20 = load <4 x i32>, <4 x i32> addrspace(2)* %tmp, !tbaa !0
▲ Show 20 Lines • Show All 422 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/shift-and-i128-ubfe.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; Extract the high bit of the 1st quarter ; Extract the high bit of the 1st quarter
; GCN-LABEL: {{^}}v_uextract_bit_31_i128: ; GCN-LABEL: {{^}}v_uextract_bit_31_i128:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}} ; GCN: buffer_load_dword [[VAL:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
; GCN: v_mov_b32_e32 v[[ZERO0:[0-9]+]], 0{{$}} ; GCN: v_mov_b32_e32 v[[ZERO0:[0-9]+]], 0{{$}}
; GCN: v_mov_b32_e32 v[[ZERO1:[0-9]+]], v[[ZERO0]]{{$}} ; GCN: v_mov_b32_e32 v[[ZERO1:[0-9]+]], v[[ZERO0]]{{$}}
; GCN: v_mov_b32_e32 v[[ZERO2:[0-9]+]], v[[ZERO0]]{{$}} ; GCN: v_mov_b32_e32 v[[ZERO2:[0-9]+]], v[[ZERO0]]{{$}}
▲ Show 20 Lines • Show All 113 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/shift-and-i64-ubfe.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; Make sure 64-bit BFE pattern does a 32-bit BFE on the relevant half. ; Make sure 64-bit BFE pattern does a 32-bit BFE on the relevant half.
; Extract the high bit of the low half ; Extract the high bit of the low half
; GCN-LABEL: {{^}}v_uextract_bit_31_i64: ; GCN-LABEL: {{^}}v_uextract_bit_31_i64:
; GCN: buffer_load_dword [[VAL:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}} ; GCN: buffer_load_dword [[VAL:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
; GCN-DAG: v_lshrrev_b32_e32 v[[SHIFT:[0-9]+]], 31, [[VAL]] ; GCN-DAG: v_lshrrev_b32_e32 v[[SHIFT:[0-9]+]], 31, [[VAL]]
; GCN-DAG: v_mov_b32_e32 v[[ZERO:[0-9]+]], 0{{$}} ; GCN-DAG: v_mov_b32_e32 v[[ZERO:[0-9]+]], 0{{$}}
▲ Show 20 Lines • Show All 388 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/shift-i64-opts.ll

; RUN: llc -march=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefix=FAST64 -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefix=FAST64 -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefix=SLOW64 -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefix=SLOW64 -check-prefix=GCN %s
; lshr (i64 x), c: c > 32 => reg_sequence lshr (i32 hi_32(x)), (c - 32), 0 ; lshr (i64 x), c: c > 32 => reg_sequence lshr (i32 hi_32(x)), (c - 32), 0
; GCN-LABEL: {{^}}lshr_i64_35: ; GCN-LABEL: {{^}}lshr_i64_35:
; GCN-DAG: buffer_load_dword [[VAL:v[0-9]+]] ; GCN-DAG: buffer_load_dword [[VAL:v[0-9]+]]
; GCN-DAG: v_lshrrev_b32_e32 v[[LO:[0-9]+]], 3, [[VAL]] ; GCN-DAG: v_lshrrev_b32_e32 v[[LO:[0-9]+]], 3, [[VAL]]
; GCN-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], 0{{$}} ; GCN-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], 0{{$}}
; GCN: buffer_store_dwordx2 v{{\[}}[[LO]]:[[HI]]{{\]}} ; GCN: buffer_store_dwordx2 v{{\[}}[[LO]]:[[HI]]{{\]}}
▲ Show 20 Lines • Show All 309 Lines • Show Last 20 Lines

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

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
; XUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=FUNC %s ; XUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i32 @llvm.r600.read.tidig.x() #0 declare i32 @llvm.r600.read.tidig.x() #0
declare i32 @llvm.r600.read.tgid.x() #0 declare i32 @llvm.r600.read.tgid.x() #0
;EG: {{^}}shl_v2i32: ;EG: {{^}}shl_v2i32:
;EG: LSHL {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ;EG: LSHL {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
▲ Show 20 Lines • Show All 468 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/si-triv-disjoint-mem-access.ll

; RUN: llc -march=amdgcn -mcpu=bonaire -enable-amdgpu-aa=0 -verify-machineinstrs -enable-misched -enable-aa-sched-mi < %s | FileCheck -check-prefix=FUNC -check-prefix=CI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -enable-amdgpu-aa=0 -verify-machineinstrs -enable-misched -enable-aa-sched-mi < %s | FileCheck -check-prefix=FUNC -check-prefix=CI %s
declare void @llvm.amdgcn.tbuffer.store.i32(i32, <4 x i32>, i32, i32, i32, i32, i32, i32, i1, i1) declare void @llvm.amdgcn.tbuffer.store.i32(i32, <4 x i32>, i32, i32, i32, i32, i32, i32, i1, i1)
declare void @llvm.amdgcn.tbuffer.store.v4i32(<4 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i1, i1) declare void @llvm.amdgcn.tbuffer.store.v4i32(<4 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i1, i1)
declare void @llvm.amdgcn.s.barrier() #1 declare void @llvm.amdgcn.s.barrier() #1
declare i32 @llvm.amdgcn.workitem.id.x() #2 declare i32 @llvm.amdgcn.workitem.id.x() #2
@stored_lds_ptr = addrspace(3) global i32 addrspace(3)* undef, align 4 @stored_lds_ptr = addrspace(3) global i32 addrspace(3)* undef, align 4
▲ Show 20 Lines • Show All 265 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/sign_extend.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI %s
; GCN-LABEL: {{^}}s_sext_i1_to_i32: ; GCN-LABEL: {{^}}s_sext_i1_to_i32:
; GCN: v_cndmask_b32_e64 ; GCN: v_cndmask_b32_e64
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @s_sext_i1_to_i32(i32 addrspace(1)* %out, i32 %a, i32 %b) nounwind { define amdgpu_kernel void @s_sext_i1_to_i32(i32 addrspace(1)* %out, i32 %a, i32 %b) nounwind {
%cmp = icmp eq i32 %a, %b %cmp = icmp eq i32 %a, %b
%sext = sext i1 %cmp to i32 %sext = sext i1 %cmp to i32
store i32 %sext, i32 addrspace(1)* %out, align 4 store i32 %sext, i32 addrspace(1)* %out, align 4
▲ Show 20 Lines • Show All 216 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/sitofp.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}sitofp_i16_to_f16 ; GCN-LABEL: {{^}}sitofp_i16_to_f16
; GCN: buffer_load_{{sshort|ushort}} v[[A_I16:[0-9]+]] ; GCN: buffer_load_{{sshort|ushort}} v[[A_I16:[0-9]+]]
; GCN: v_cvt_f32_i32_e32 v[[A_F32:[0-9]+]], v[[A_I16]] ; GCN: v_cvt_f32_i32_e32 v[[A_F32:[0-9]+]], v[[A_I16]]
; GCN: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[A_F32]] ; GCN: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[A_F32]]
; GCN: buffer_store_short v[[R_F16]] ; GCN: buffer_store_short v[[R_F16]]
; GCN: s_endpgm ; GCN: s_endpgm
define amdgpu_kernel void @sitofp_i16_to_f16( define amdgpu_kernel void @sitofp_i16_to_f16(
▲ Show 20 Lines • Show All 85 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/sminmax.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}s_abs_i32: ; FUNC-LABEL: {{^}}s_abs_i32:
; GCN: s_abs_i32 ; GCN: s_abs_i32
; GCN: s_add_i32 ; GCN: s_add_i32
; EG: MAX_INT ; EG: MAX_INT
define amdgpu_kernel void @s_abs_i32(i32 addrspace(1)* %out, i32 %val) nounwind { define amdgpu_kernel void @s_abs_i32(i32 addrspace(1)* %out, i32 %val) nounwind {
%neg = sub i32 0, %val %neg = sub i32 0, %val
▲ Show 20 Lines • Show All 227 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/sminmax.v2i16.ll

; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GFX9 -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GFX9 -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=CIVI -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=CIVI -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=CIVI -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=CIVI -check-prefix=GCN %s
; GCN-LABEL: {{^}}s_abs_v2i16: ; GCN-LABEL: {{^}}s_abs_v2i16:
; GFX9: s_load_dword [[VAL:s[0-9]+]] ; GFX9: s_load_dword [[VAL:s[0-9]+]]
; GFX9: v_pk_sub_i16 [[SUB:v[0-9]+]], 0, [[VAL]] ; GFX9: v_pk_sub_i16 [[SUB:v[0-9]+]], 0, [[VAL]]
; GFX9: v_pk_max_i16 [[MAX:v[0-9]+]], [[VAL]], [[SUB]] ; GFX9: v_pk_max_i16 [[MAX:v[0-9]+]], [[VAL]], [[SUB]]
; GFX9: v_pk_add_u16 [[ADD:v[0-9]+]], [[MAX]], 2 ; GFX9: v_pk_add_u16 [[ADD:v[0-9]+]], [[MAX]], 2
; VI: v_sub_i32_e32 ; VI: v_sub_i32_e32
▲ Show 20 Lines • Show All 214 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/spill-cfg-position.ll

; RUN: llc -march=amdgcn -verify-machineinstrs -stress-regalloc=6 < %s | FileCheck %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -stress-regalloc=6 < %s | FileCheck %s
; Inline spiller can decide to move a spill as early as possible in the basic block. ; Inline spiller can decide to move a spill as early as possible in the basic block.
; It will skip phis and label, but we also need to make sure it skips instructions ; It will skip phis and label, but we also need to make sure it skips instructions
; in the basic block prologue which restore exec mask. ; in the basic block prologue which restore exec mask.
; Make sure instruction to restore exec mask immediately follows label ; Make sure instruction to restore exec mask immediately follows label
; CHECK-LABEL: {{^}}spill_cfg_position: ; CHECK-LABEL: {{^}}spill_cfg_position:
; CHECK: s_cbranch_execz [[LABEL1:BB[0-9_]+]] ; CHECK: s_cbranch_execz [[LABEL1:BB[0-9_]+]]
▲ Show 20 Lines • Show All 69 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/sra.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i32 @llvm.r600.read.tidig.x() #0 declare i32 @llvm.r600.read.tidig.x() #0
; FUNC-LABEL: {{^}}ashr_v2i32: ; FUNC-LABEL: {{^}}ashr_v2i32:
; SI: v_ashr_i32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: v_ashr_i32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
; SI: v_ashr_i32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: v_ashr_i32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
; VI: v_ashrrev_i32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; VI: v_ashrrev_i32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
▲ Show 20 Lines • Show All 277 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/srem.ll

; RUN: llc -march=amdgcn -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s
define amdgpu_kernel void @srem_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) { define amdgpu_kernel void @srem_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
%den_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1 %den_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
%num = load i32, i32 addrspace(1) * %in %num = load i32, i32 addrspace(1) * %in
%den = load i32, i32 addrspace(1) * %den_ptr %den = load i32, i32 addrspace(1) * %den_ptr
%result = srem i32 %num, %den %result = srem i32 %num, %den
store i32 %result, i32 addrspace(1)* %out store i32 %result, i32 addrspace(1)* %out
ret void ret void
▲ Show 20 Lines • Show All 101 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/srl.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s
; XUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=GCN -check-prefix=FUNC %s ; XUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i32 @llvm.r600.read.tidig.x() #0 declare i32 @llvm.r600.read.tidig.x() #0
; FUNC-LABEL: {{^}}lshr_i32: ; FUNC-LABEL: {{^}}lshr_i32:
; SI: v_lshrrev_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: v_lshrrev_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
; VI: v_lshrrev_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; VI: v_lshrrev_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
; EG: LSHR {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: LSHR {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
define amdgpu_kernel void @lshr_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) { define amdgpu_kernel void @lshr_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
▲ Show 20 Lines • Show All 206 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/ssubo.ll

; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=SI -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs< %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs< %s
declare { i32, i1 } @llvm.ssub.with.overflow.i32(i32, i32) nounwind readnone declare { i32, i1 } @llvm.ssub.with.overflow.i32(i32, i32) nounwind readnone
declare { i64, i1 } @llvm.ssub.with.overflow.i64(i64, i64) nounwind readnone declare { i64, i1 } @llvm.ssub.with.overflow.i64(i64, i64) nounwind readnone
; FUNC-LABEL: {{^}}ssubo_i64_zext: ; FUNC-LABEL: {{^}}ssubo_i64_zext:
define amdgpu_kernel void @ssubo_i64_zext(i64 addrspace(1)* %out, i64 %a, i64 %b) nounwind { define amdgpu_kernel void @ssubo_i64_zext(i64 addrspace(1)* %out, i64 %a, i64 %b) nounwind {
%ssub = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) nounwind %ssub = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) nounwind
%val = extractvalue { i64, i1 } %ssub, 0 %val = extractvalue { i64, i1 } %ssub, 0
▲ Show 20 Lines • Show All 54 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/sub.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i32 @llvm.r600.read.tidig.x() readnone declare i32 @llvm.r600.read.tidig.x() readnone
; FUNC-LABEL: {{^}}test_sub_i32: ; FUNC-LABEL: {{^}}test_sub_i32:
; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; SI: v_subrev_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}} ; SI: v_subrev_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
▲ Show 20 Lines • Show All 156 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/trunc-bitcast-vector.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck --check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck --check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefix=VI %s
; CHECK-LABEL: {{^}}trunc_i64_bitcast_v2i32: ; CHECK-LABEL: {{^}}trunc_i64_bitcast_v2i32:
; CHECK: buffer_load_dword v ; CHECK: buffer_load_dword v
; CHECK: buffer_store_dword v ; CHECK: buffer_store_dword v
define amdgpu_kernel void @trunc_i64_bitcast_v2i32(i32 addrspace(1)* %out, <2 x i32> addrspace(1)* %in) { define amdgpu_kernel void @trunc_i64_bitcast_v2i32(i32 addrspace(1)* %out, <2 x i32> addrspace(1)* %in) {
%ld = load <2 x i32>, <2 x i32> addrspace(1)* %in %ld = load <2 x i32>, <2 x i32> addrspace(1)* %in
%bc = bitcast <2 x i32> %ld to i64 %bc = bitcast <2 x i32> %ld to i64
%trunc = trunc i64 %bc to i32 %trunc = trunc i64 %bc to i32
▲ Show 20 Lines • Show All 87 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/trunc.ll

; RUN: llc -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs< %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -verify-machineinstrs< %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; RUN: llc -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG %s
declare i32 @llvm.r600.read.tidig.x() nounwind readnone declare i32 @llvm.r600.read.tidig.x() nounwind readnone
define amdgpu_kernel void @trunc_i64_to_i32_store(i32 addrspace(1)* %out, i64 %in) { define amdgpu_kernel void @trunc_i64_to_i32_store(i32 addrspace(1)* %out, i64 %in) {
; GCN-LABEL: {{^}}trunc_i64_to_i32_store: ; GCN-LABEL: {{^}}trunc_i64_to_i32_store:
; GCN: s_load_dword [[SLOAD:s[0-9]+]], s[0:1], ; GCN: s_load_dword [[SLOAD:s[0-9]+]], s[0:1],
; GCN: v_mov_b32_e32 [[VLOAD:v[0-9]+]], [[SLOAD]] ; GCN: v_mov_b32_e32 [[VLOAD:v[0-9]+]], [[SLOAD]]
; SI: buffer_store_dword [[VLOAD]] ; SI: buffer_store_dword [[VLOAD]]
▲ Show 20 Lines • Show All 114 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/uaddo.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI,FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,FUNC %s
; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefixes=EG,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefixes=EG,FUNC %s
; FUNC-LABEL: {{^}}s_uaddo_i64_zext: ; FUNC-LABEL: {{^}}s_uaddo_i64_zext:
; GCN: s_add_u32 ; GCN: s_add_u32
; GCN: s_addc_u32 ; GCN: s_addc_u32
; GCN: v_cmp_lt_u64_e32 vcc ; GCN: v_cmp_lt_u64_e32 vcc
; EG: ADDC_UINT ; EG: ADDC_UINT
; EG: ADDC_UINT ; EG: ADDC_UINT
▲ Show 20 Lines • Show All 131 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/udiv.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs -mattr=-fp32-denormals < %s | FileCheck -check-prefix=SI -check-prefix=FUNC -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs -mattr=-fp32-denormals < %s | FileCheck -check-prefix=SI -check-prefix=FUNC -check-prefix=VI %s
; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=+fp32-denormals < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=+fp32-denormals < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}udiv_i32: ; FUNC-LABEL: {{^}}udiv_i32:
; EG-NOT: SETGE_INT ; EG-NOT: SETGE_INT
; EG: CF_END ; EG: CF_END
; SI: v_rcp_iflag_f32_e32 ; SI: v_rcp_iflag_f32_e32
define amdgpu_kernel void @udiv_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) { define amdgpu_kernel void @udiv_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
%b_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1 %b_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
▲ Show 20 Lines • Show All 186 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/uitofp.f16.ll

; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}uitofp_i16_to_f16 ; GCN-LABEL: {{^}}uitofp_i16_to_f16
; GCN: buffer_load_ushort v[[A_I16:[0-9]+]] ; GCN: buffer_load_ushort v[[A_I16:[0-9]+]]
; SI: v_cvt_f32_u32_e32 v[[A_F32:[0-9]+]], v[[A_I16]] ; SI: v_cvt_f32_u32_e32 v[[A_F32:[0-9]+]], v[[A_I16]]
; VI: v_cvt_f32_i32_e32 v[[A_F32:[0-9]+]], v[[A_I16]] ; VI: v_cvt_f32_i32_e32 v[[A_F32:[0-9]+]], v[[A_I16]]
; GCN: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[A_F32]] ; GCN: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[A_F32]]
; GCN: buffer_store_short v[[R_F16]] ; GCN: buffer_store_short v[[R_F16]]
; GCN: s_endpgm ; GCN: s_endpgm
▲ Show 20 Lines • Show All 85 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/urem.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; The code generated by urem is long and complex and may frequently ; The code generated by urem is long and complex and may frequently
; change. The goal of this test is to make sure the ISel doesn't fail ; change. The goal of this test is to make sure the ISel doesn't fail
; when it gets a v2i32/v4i32 urem ; when it gets a v2i32/v4i32 urem
; FUNC-LABEL: {{^}}test_urem_i32: ; FUNC-LABEL: {{^}}test_urem_i32:
; SI: s_endpgm ; SI: s_endpgm
; EG: CF_END ; EG: CF_END
▲ Show 20 Lines • Show All 83 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/usubo.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI,FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,FUNC %s
; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefixes=EG,FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefixes=EG,FUNC %s
; FUNC-LABEL: {{^}}s_usubo_i64_zext: ; FUNC-LABEL: {{^}}s_usubo_i64_zext:
; GCN: s_sub_u32 ; GCN: s_sub_u32
; GCN: s_subb_u32 ; GCN: s_subb_u32
; GCN: v_cmp_gt_u64_e32 vcc ; GCN: v_cmp_gt_u64_e32 vcc
; EG: SUBB_UINT ; EG: SUBB_UINT
; EG: ADDC_UINT ; EG: ADDC_UINT
▲ Show 20 Lines • Show All 135 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/v_mac.ll

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=VI-FLUSH -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=VI-FLUSH -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=+fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=VI-DENORM -check-prefix=GCN %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=+fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=VI-DENORM -check-prefix=GCN %s
; GCN-LABEL: {{^}}mac_vvv: ; GCN-LABEL: {{^}}mac_vvv:
; GCN: buffer_load_dword [[A:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], 0{{$}} ; GCN: buffer_load_dword [[A:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], 0{{$}}
; GCN: buffer_load_dword [[B:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], 0 offset:4 ; GCN: buffer_load_dword [[B:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], 0 offset:4
; GCN: buffer_load_dword [[C:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], 0 offset:8 ; GCN: buffer_load_dword [[C:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], 0 offset:8
; GCN: v_mac_f32_e32 [[C]], [[B]], [[A]] ; GCN: v_mac_f32_e32 [[C]], [[B]], [[A]]
; GCN: buffer_store_dword [[C]] ; GCN: buffer_store_dword [[C]]
define amdgpu_kernel void @mac_vvv(float addrspace(1)* %out, float addrspace(1)* %in) #0 { define amdgpu_kernel void @mac_vvv(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
▲ Show 20 Lines • Show All 276 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/v_mac_f16.ll

; RUN: llc -march=amdgcn -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}mac_f16: ; GCN-LABEL: {{^}}mac_f16:
; GCN: {{buffer|flat}}_load_ushort v[[A_F16:[0-9]+]] ; GCN: {{buffer|flat}}_load_ushort v[[A_F16:[0-9]+]]
; GCN: {{buffer|flat}}_load_ushort v[[B_F16:[0-9]+]] ; GCN: {{buffer|flat}}_load_ushort v[[B_F16:[0-9]+]]
; GCN: {{buffer|flat}}_load_ushort v[[C_F16:[0-9]+]] ; GCN: {{buffer|flat}}_load_ushort v[[C_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]] ; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]] ; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]]
; SI: v_cvt_f32_f16_e32 v[[C_F32:[0-9]+]], v[[C_F16]] ; SI: v_cvt_f32_f16_e32 v[[C_F32:[0-9]+]], v[[C_F16]]
▲ Show 20 Lines • Show All 669 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/vectorize-global-local.ll

; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck %s
; CHECK-DAG: flat_load_dwordx4 ; CHECK-DAG: flat_load_dwordx4
; CHECK-DAG: flat_load_dwordx4 ; CHECK-DAG: flat_load_dwordx4
; CHECK-DAG: flat_load_dwordx4 ; CHECK-DAG: flat_load_dwordx4
; CHECK-DAG: flat_load_dwordx4 ; CHECK-DAG: flat_load_dwordx4
; CHECK-DAG: ds_write2_b32 ; CHECK-DAG: ds_write2_b32
; CHECK-DAG: ds_write2_b32 ; CHECK-DAG: ds_write2_b32
; CHECK-DAG: ds_write2_b32 ; CHECK-DAG: ds_write2_b32
; CHECK-DAG: ds_write2_b32 ; CHECK-DAG: ds_write2_b32
▲ Show 20 Lines • Show All 71 Lines • Show Last 20 Lines

llvm/trunk/test/CodeGen/AMDGPU/xor.ll

; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}xor_v2i32: ; FUNC-LABEL: {{^}}xor_v2i32:
; EG: XOR_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: XOR_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; EG: XOR_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} ; EG: XOR_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; SI: v_xor_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: v_xor_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
; SI: v_xor_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}} ; SI: v_xor_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
▲ Show 20 Lines • Show All 240 Lines • Show Last 20 Lines