This is an archive of the discontinued LLVM Phabricator instance.

Differential D62408

[X86] Avoid SFB - Fix inconsistent codegen with/without debug info(2)
ClosedPublic

Authored by cdawson on May 24 2019, 9:55 AM.

Details

Reviewers
craig.topper
lsaba
RKSimon
probinson
aprantl
Commits
rGe20030f61217: [X86] Avoid SFB - Fix inconsistent codegen with/without debug info(2)
rL364829: [X86] Avoid SFB - Fix inconsistent codegen with/without debug info(2)
Summary

The function findPotentialBlockers may consider debug info instructions as potential blockers
and may stop searching for a store-load pair prematurely.

This patch corrects this and tests the cases where the store is separated from the
load by more than InspectionLimit debug instructions.

See D61680 for a previous simmilar fix in this area.

Diff Detail

Event Timeline

cdawson created this revision.May 24 2019, 9:55 AM
Herald added a project: Restricted Project. · View Herald TranscriptMay 24 2019, 9:55 AM
Herald added a subscriber: hiraditya. · View Herald Transcript
aprantl added inline comments.May 24 2019, 11:09 AM
llvm/lib/Target/X86/X86AvoidStoreForwardingBlocks.cpp
345

Do you perhapswant to use isMetaInstruction() here? There are other non-debug instructions (CFI for example) that you might want to ignore.

cdawson marked an inline comment as done.May 28 2019, 7:04 AM
cdawson added inline comments.
llvm/lib/Target/X86/X86AvoidStoreForwardingBlocks.cpp
345

This seems like a good idea, though it would be inconsistent with D61680 where skipDebugInstructionsBackward() is used. I think there's a benefit to checking for meta instrs in both these cases, but as far as I can see there isn't an equivalent skipMetaInstructionsBackward(). Do you think it would be worth adding skipMetaInstructionsBackward/Forward() and amending D61680? Or should I follow your proposed change only in this case.

cdawson marked an inline comment as not done.May 28 2019, 7:04 AM
aprantl added a comment.May 28 2019, 9:48 AM

Do you think it would be worth adding skipMetaInstructionsBackward/Forward() and amending D61680?

Yes, that sounds like a good idea!

cdawson updated this revision to Diff 201858.May 29 2019, 4:13 AM

Addressed review comments

cdawson marked an inline comment as done.May 29 2019, 4:14 AM
cdawson updated this revision to Diff 201906.May 29 2019, 7:07 AM

Updated tests to include debug and meta instructions.

cdawson updated this revision to Diff 201912.May 29 2019, 7:35 AM

Reduce test cases. Since the debug/nodebug consistency check occurs in avoid-sfb-g-no-change.mir, we can just check that debug/meta instructions are skipped in the debug case.

aprantl accepted this revision.May 29 2019, 4:14 PM
This revision is now accepted and ready to land.May 29 2019, 4:14 PM
Closed by commit rL364829: [X86] Avoid SFB - Fix inconsistent codegen with/without debug info(2) (authored by rlougher). · Explain WhyJul 1 2019, 11:29 AM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
llvm/
lib/
Target/
X86/
4 lines
test/
CodeGen/
X86/
264 lines
305 lines

Diff 201906

llvm/lib/Target/X86/X86AvoidStoreForwardingBlocks.cpp

Show First 20 Lines • Show All 336 Lines • ▼ Show 20 Lines
static SmallVector<MachineInstr *, 2> static SmallVector<MachineInstr *, 2>
findPotentialBlockers(MachineInstr *LoadInst) { findPotentialBlockers(MachineInstr *LoadInst) {
SmallVector<MachineInstr *, 2> PotentialBlockers; SmallVector<MachineInstr *, 2> PotentialBlockers;
unsigned BlockCount = 0; unsigned BlockCount = 0;
const unsigned InspectionLimit = X86AvoidSFBInspectionLimit; const unsigned InspectionLimit = X86AvoidSFBInspectionLimit;
for (auto PBInst = std::next(MachineBasicBlock::reverse_iterator(LoadInst)), for (auto PBInst = std::next(MachineBasicBlock::reverse_iterator(LoadInst)),
E = LoadInst->getParent()->rend(); E = LoadInst->getParent()->rend();
PBInst != E; ++PBInst) { PBInst != E; ++PBInst) {
if (PBInst->isMetaInstruction())
aprantlUnsubmitted
Done
ReplyInline Actions

Do you perhapswant to use isMetaInstruction() here? There are other non-debug instructions (CFI for example) that you might want to ignore.

aprantl: Do you perhapswant to use `isMetaInstruction()` here? There are other non-debug instructions…
cdawsonAuthorUnsubmitted
Not Done
ReplyInline Actions

This seems like a good idea, though it would be inconsistent with D61680 where skipDebugInstructionsBackward() is used. I think there's a benefit to checking for meta instrs in both these cases, but as far as I can see there isn't an equivalent skipMetaInstructionsBackward(). Do you think it would be worth adding skipMetaInstructionsBackward/Forward() and amending D61680? Or should I follow your proposed change only in this case.

cdawson: This seems like a good idea, though it would be inconsistent with D61680 where…
continue;
BlockCount++; BlockCount++;
if (BlockCount >= InspectionLimit) if (BlockCount >= InspectionLimit)
break; break;
MachineInstr &MI = *PBInst; MachineInstr &MI = *PBInst;
if (MI.getDesc().isCall()) if (MI.getDesc().isCall())
return PotentialBlockers; return PotentialBlockers;
PotentialBlockers.push_back(&MI); PotentialBlockers.push_back(&MI);
} }
// If we didn't get to the instructions limit try predecessing blocks. // If we didn't get to the instructions limit try predecessing blocks.
// Ideally we should traverse the predecessor blocks in depth with some // Ideally we should traverse the predecessor blocks in depth with some
// coloring algorithm, but for now let's just look at the first order // coloring algorithm, but for now let's just look at the first order
// predecessors. // predecessors.
if (BlockCount < InspectionLimit) { if (BlockCount < InspectionLimit) {
MachineBasicBlock *MBB = LoadInst->getParent(); MachineBasicBlock *MBB = LoadInst->getParent();
int LimitLeft = InspectionLimit - BlockCount; int LimitLeft = InspectionLimit - BlockCount;
for (MachineBasicBlock::pred_iterator PB = MBB->pred_begin(), for (MachineBasicBlock::pred_iterator PB = MBB->pred_begin(),
PE = MBB->pred_end(); PE = MBB->pred_end();
PB != PE; ++PB) { PB != PE; ++PB) {
MachineBasicBlock *PMBB = *PB; MachineBasicBlock *PMBB = *PB;
int PredCount = 0; int PredCount = 0;
for (MachineBasicBlock::reverse_iterator PBInst = PMBB->rbegin(), for (MachineBasicBlock::reverse_iterator PBInst = PMBB->rbegin(),
PME = PMBB->rend(); PME = PMBB->rend();
PBInst != PME; ++PBInst) { PBInst != PME; ++PBInst) {
if (PBInst->isMetaInstruction())
continue;
PredCount++; PredCount++;
if (PredCount >= LimitLeft) if (PredCount >= LimitLeft)
break; break;
if (PBInst->getDesc().isCall()) if (PBInst->getDesc().isCall())
break; break;
PotentialBlockers.push_back(&*PBInst); PotentialBlockers.push_back(&*PBInst);
} }
} }
▲ Show 20 Lines • Show All 359 Lines • Show Last 20 Lines

llvm/test/CodeGen/X86/avoid-sfb-g-no-change2.mir

  • This file was added.
# RUN: llc %s -run-pass x86-avoid-SFB -mtriple=x86_64-unknown-linux-gnu -o - | FileCheck %s -check-prefixes DEBUG-LABEL,CHECK
# RUN: llc %s -run-pass x86-avoid-SFB -mtriple=x86_64-unknown-linux-gnu -o - | FileCheck %s -check-prefixes NODEBUG-LABEL,CHECK
#
# This was generated from:
#
# using alpha = float __attribute__((ext_vector_type(4)));
#
# void bravo(alpha * __restrict__ p1, alpha * __restrict__ p2) {
# char *p3 = (char *)p1;
# *p3 = 0;
# alpha t = *p1;
# *p2 = t;
# }
#
# Using the command line:
# clang -g -c 1.cpp -O2 -S -emit-llvm -fno-strict-aliasing --target=x86_64-unknown-unknown -o test.ll
# llc -stop-before=x86-avoid-SFB test.ll -o before.mir
#
# The IR has been modified to include a number of debug/meta instructions between the initial store and load.
# If the code treats debug/meta instructions as normal instructions, the separation between the store and load
# will be too great to enable the optimization.
--- |
; ModuleID = 'test.ll'
source_filename = "1.cpp"
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-unknown"
; Function Attrs: norecurse nounwind uwtable
define dso_local void @debug(<4 x float>* noalias nocapture %p1, <4 x float>* noalias nocapture %p2) local_unnamed_addr #0 !dbg !10 {
entry:
call void @llvm.dbg.value(metadata <4 x float>* %p1, metadata !21, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata <4 x float>* %p2, metadata !22, metadata !DIExpression()), !dbg !25
%0 = bitcast <4 x float>* %p1 to i8*, !dbg !26
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
store i8 0, i8* %0, align 1, !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !25
%1 = load <4 x float>, <4 x float>* %p1, align 16, !dbg !28
call void @llvm.dbg.value(metadata <4 x float> %1, metadata !24, metadata !DIExpression()), !dbg !25
store <4 x float> %1, <4 x float>* %p2, align 16, !dbg !29
ret void, !dbg !30
}
; Function Attrs: norecurse nounwind uwtable
define dso_local void @nodebug(<4 x float>* noalias nocapture %p1, <4 x float>* noalias nocapture %p2) local_unnamed_addr #0 {
entry:
%0 = bitcast <4 x float>* %p1 to i8*
store i8 0, i8* %0, align 1
%1 = load <4 x float>, <4 x float>* %p1, align 16
store <4 x float> %1, <4 x float>* %p2, align 16
ret void
}
; Function Attrs: nounwind readnone speculatable
declare void @llvm.dbg.value(metadata, metadata, metadata) #1
; Function Attrs: nounwind
declare void @llvm.stackprotector(i8*, i8**) #2
attributes #0 = { norecurse nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #1 = { nounwind readnone speculatable }
attributes #2 = { nounwind }
!llvm.dbg.cu = !{!0}
!llvm.module.flags = !{!6, !7, !8}
!llvm.ident = !{!9}
!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, file: !1, producer: "clang version 9.0.0 (https://github.com/llvm/llvm-project.git 9afc4764dd24bd2f23c44e51ad33f8e58234a8b6)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, retainedTypes: !3, nameTableKind: None)
!1 = !DIFile(filename: "1.cpp", directory: "C:\5CUsers\5Cgbdawsoc\5CDocuments\5Cllvm\5Cbg40969")
!2 = !{}
!3 = !{!4}
!4 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !5, size: 64)
!5 = !DIBasicType(name: "char", size: 8, encoding: DW_ATE_signed_char)
!6 = !{i32 2, !"Dwarf Version", i32 4}
!7 = !{i32 2, !"Debug Info Version", i32 3}
!8 = !{i32 1, !"wchar_size", i32 4}
!9 = !{!"clang version 9.0.0 (https://github.com/llvm/llvm-project.git 9afc4764dd24bd2f23c44e51ad33f8e58234a8b6)"}
!10 = distinct !DISubprogram(name: "bravo", linkageName: "_Z5bravoPDv4_fS0_", scope: !1, file: !1, line: 4, type: !11, scopeLine: 4, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !20)
!11 = !DISubroutineType(types: !12)
!12 = !{null, !13, !13}
!13 = !DIDerivedType(tag: DW_TAG_restrict_type, baseType: !14)
!14 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !15, size: 64)
!15 = !DIDerivedType(tag: DW_TAG_typedef, name: "alpha", file: !1, line: 2, baseType: !16)
!16 = !DICompositeType(tag: DW_TAG_array_type, baseType: !17, size: 128, flags: DIFlagVector, elements: !18)
!17 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
!18 = !{!19}
!19 = !DISubrange(count: 4)
!20 = !{!21, !22, !23, !24}
!21 = !DILocalVariable(name: "p1", arg: 1, scope: !10, file: !1, line: 4, type: !13)
!22 = !DILocalVariable(name: "p2", arg: 2, scope: !10, file: !1, line: 4, type: !13)
!23 = !DILocalVariable(name: "p3", scope: !10, file: !1, line: 5, type: !4)
!24 = !DILocalVariable(name: "t", scope: !10, file: !1, line: 7, type: !15)
!25 = !DILocation(line: 0, scope: !10)
!26 = !DILocation(line: 5, column: 14, scope: !10)
!27 = !DILocation(line: 6, column: 7, scope: !10)
!28 = !DILocation(line: 7, column: 13, scope: !10)
!29 = !DILocation(line: 8, column: 7, scope: !10)
!30 = !DILocation(line: 9, column: 1, scope: !10)
...
---
name: debug
alignment: 4
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
failedISel: false
tracksRegLiveness: true
hasWinCFI: false
registers:
- { id: 0, class: gr64, preferred-register: '' }
- { id: 1, class: gr64, preferred-register: '' }
- { id: 2, class: vr128, preferred-register: '' }
liveins:
- { reg: '$rdi', virtual-reg: '%0' }
- { reg: '$rsi', virtual-reg: '%1' }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
stackProtector: ''
maxCallFrameSize: 4294967295
cvBytesOfCalleeSavedRegisters: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
localFrameSize: 0
savePoint: ''
restorePoint: ''
fixedStack: []
stack: []
constants: []
machineFunctionInfo: {}
body: |
bb.0.entry:
liveins: $rdi, $rsi
DBG_VALUE $rdi, $noreg, !21, !DIExpression(), debug-location !25
DBG_VALUE $rsi, $noreg, !22, !DIExpression(), debug-location !25
%1:gr64 = COPY $rsi
DBG_VALUE %1, $noreg, !22, !DIExpression(), debug-location !25
%0:gr64 = COPY $rdi
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !25
MOV8mi %0, 1, $noreg, 0, $noreg, 0, debug-location !27 :: (store 1 into %ir.0)
CFI_INSTRUCTION offset $r13, -123
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !25
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !25
%2:vr128 = MOVAPSrm %0, 1, $noreg, 0, $noreg, debug-location !28 :: (load 16 from %ir.p1)
DBG_VALUE %2, $noreg, !24, !DIExpression(), debug-location !25
MOVAPSmr %1, 1, $noreg, 0, $noreg, killed %2, debug-location !29 :: (store 16 into %ir.p2)
RET 0, debug-location !30
...
---
name: nodebug
alignment: 4
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
failedISel: false
tracksRegLiveness: true
hasWinCFI: false
registers:
- { id: 0, class: gr64, preferred-register: '' }
- { id: 1, class: gr64, preferred-register: '' }
- { id: 2, class: vr128, preferred-register: '' }
liveins:
- { reg: '$rdi', virtual-reg: '%0' }
- { reg: '$rsi', virtual-reg: '%1' }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
stackProtector: ''
maxCallFrameSize: 4294967295
cvBytesOfCalleeSavedRegisters: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
localFrameSize: 0
savePoint: ''
restorePoint: ''
fixedStack: []
stack: []
constants: []
machineFunctionInfo: {}
body: |
bb.0.entry:
liveins: $rdi, $rsi
%1:gr64 = COPY $rsi
%0:gr64 = COPY $rdi
MOV8mi %0, 1, $noreg, 0, $noreg, 0 :: (store 1 into %ir.0)
%2:vr128 = MOVAPSrm %0, 1, $noreg, 0, $noreg :: (load 16 from %ir.p1)
MOVAPSmr %1, 1, $noreg, 0, $noreg, killed %2 :: (store 16 into %ir.p2)
RET 0
; DEBUG-LABEL: name: debug
; NODEBUG-LABEL: name: nodebug
; CHECK: %1:gr64 = COPY
; CHECK: %0:gr64 = COPY
; CHECK: MOV8mi
; CHECK: %3:gr8 = MOV8rm
; CHECK: MOV8mr
; CHECK: %4:gr64 = MOV64rm
; CHECK: MOV64mr
; CHECK: %5:gr32 = MOV32rm
; CHECK: MOV32mr
; CHECK: %6:gr16 = MOV16rm
; CHECK: MOV16mr
; CHECK: %7:gr8 = MOV8rm
; CHECK: MOV8mr
; CHECK: RET 0
; DEBUG-LABEL: name: nodebug
...

llvm/test/CodeGen/X86/avoid-sfb-g-no-change3.mir

  • This file was added.
# RUN: llc %s -run-pass x86-avoid-SFB -mtriple=x86_64-unknown-linux-gnu -o - | FileCheck %s -check-prefixes DEBUG-LABEL,CHECK
# RUN: llc %s -run-pass x86-avoid-SFB -mtriple=x86_64-unknown-linux-gnu -o - | FileCheck %s -check-prefixes NODEBUG-LABEL,CHECK
#
# This was generated from:
#
# using alpha = float __attribute__((ext_vector_type(4)));
#
# void bravo(alpha * __restrict__ p1, alpha * __restrict__ p2) {
# char *p3 = (char *)p1;
# if (p1) {
# *p3 = 0;
# alpha t = *p1;
# *p2 = t;
# }
# }
#
# Using the command line:
# clang -g -c 1.cpp -O2 -S -emit-llvm -fno-strict-aliasing --target=x86_64-unknown-unknown -o test.ll
# llc -stop-before=x86-avoid-SFB test.ll -o before.mir
#
# The IR has been modified to include a number of debug/meta instructions between the initial store and load.
# If the code treats debug/meta instructions as normal instructions, the separation between the store and load
# will be too great to enable the optimization.
--- |
; ModuleID = 'test.ll'
source_filename = "1.cpp"
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-unknown"
; Function Attrs: norecurse nounwind uwtable
define dso_local void @debug(<4 x float>* noalias %p1, <4 x float>* noalias nocapture %p2) local_unnamed_addr #0 !dbg !10 {
entry:
call void @llvm.dbg.value(metadata <4 x float>* %p1, metadata !21, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata <4 x float>* %p2, metadata !22, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata <4 x float>* %p1, metadata !23, metadata !DIExpression()), !dbg !27
%tobool = icmp eq <4 x float>* %p1, null, !dbg !28
%0 = bitcast <4 x float>* %p1 to i8*, !dbg !29
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
store i8 0, i8* %0, align 1, !dbg !30
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
call void @llvm.dbg.value(metadata i8* %0, metadata !23, metadata !DIExpression()), !dbg !27
br i1 %tobool, label %if.end, label %if.then, !dbg !31
if.then: ; preds = %entry
%1 = load <4 x float>, <4 x float>* %p1, align 16, !dbg !32
call void @llvm.dbg.value(metadata <4 x float> %1, metadata !24, metadata !DIExpression()), !dbg !33
store <4 x float> %1, <4 x float>* %p2, align 16, !dbg !34
br label %if.end, !dbg !35
if.end: ; preds = %if.then, %entry
ret void, !dbg !36
}
; Function Attrs: norecurse nounwind uwtable
define dso_local void @nodebug(<4 x float>* noalias %p1, <4 x float>* noalias nocapture %p2) local_unnamed_addr #0 {
entry:
%tobool = icmp eq <4 x float>* %p1, null
br i1 %tobool, label %if.end, label %if.then
if.then: ; preds = %entry
%0 = bitcast <4 x float>* %p1 to i8*
store i8 0, i8* %0, align 1
%1 = load <4 x float>, <4 x float>* %p1, align 16
store <4 x float> %1, <4 x float>* %p2, align 16
br label %if.end
if.end: ; preds = %if.then, %entry
ret void
}
; Function Attrs: nounwind readnone speculatable
declare void @llvm.dbg.value(metadata, metadata, metadata) #1
; Function Attrs: nounwind
declare void @llvm.stackprotector(i8*, i8**) #2
attributes #0 = { norecurse nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #1 = { nounwind readnone speculatable }
attributes #2 = { nounwind }
!llvm.dbg.cu = !{!0}
!llvm.module.flags = !{!6, !7, !8}
!llvm.ident = !{!9}
!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, file: !1, producer: "clang version 9.0.0 (https://github.com/llvm/llvm-project.git 1a0312ca0b20d16edb859065bbace75f6701c92e)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, retainedTypes: !3, nameTableKind: None)
!1 = !DIFile(filename: "1.cpp", directory: "/mnt/c/Users/gbdawsoc/Documents/llvm/bg40969")
!2 = !{}
!3 = !{!4}
!4 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !5, size: 64)
!5 = !DIBasicType(name: "char", size: 8, encoding: DW_ATE_signed_char)
!6 = !{i32 2, !"Dwarf Version", i32 4}
!7 = !{i32 2, !"Debug Info Version", i32 3}
!8 = !{i32 1, !"wchar_size", i32 4}
!9 = !{!"clang version 9.0.0 (https://github.com/llvm/llvm-project.git 1a0312ca0b20d16edb859065bbace75f6701c92e)"}
!10 = distinct !DISubprogram(name: "bravo", linkageName: "debug", scope: !1, file: !1, line: 4, type: !11, scopeLine: 4, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !20)
!11 = !DISubroutineType(types: !12)
!12 = !{null, !13, !13}
!13 = !DIDerivedType(tag: DW_TAG_restrict_type, baseType: !14)
!14 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !15, size: 64)
!15 = !DIDerivedType(tag: DW_TAG_typedef, name: "alpha", file: !1, line: 2, baseType: !16)
!16 = !DICompositeType(tag: DW_TAG_array_type, baseType: !17, size: 128, flags: DIFlagVector, elements: !18)
!17 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
!18 = !{!19}
!19 = !DISubrange(count: 4)
!20 = !{!21, !22, !23, !24}
!21 = !DILocalVariable(name: "p1", arg: 1, scope: !10, file: !1, line: 4, type: !13)
!22 = !DILocalVariable(name: "p2", arg: 2, scope: !10, file: !1, line: 4, type: !13)
!23 = !DILocalVariable(name: "p3", scope: !10, file: !1, line: 5, type: !4)
!24 = !DILocalVariable(name: "t", scope: !25, file: !1, line: 8, type: !15)
!25 = distinct !DILexicalBlock(scope: !26, file: !1, line: 6, column: 11)
!26 = distinct !DILexicalBlock(scope: !10, file: !1, line: 6, column: 7)
!27 = !DILocation(line: 0, scope: !10)
!28 = !DILocation(line: 6, column: 7, scope: !26)
!29 = !DILocation(line: 5, column: 14, scope: !10)
!30 = !DILocation(line: 7, column: 9, scope: !25)
!31 = !DILocation(line: 6, column: 7, scope: !10)
!32 = !DILocation(line: 8, column: 15, scope: !25)
!33 = !DILocation(line: 0, scope: !25)
!34 = !DILocation(line: 9, column: 9, scope: !25)
!35 = !DILocation(line: 10, column: 3, scope: !25)
!36 = !DILocation(line: 11, column: 1, scope: !10)
...
---
name: debug
alignment: 4
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
failedISel: false
tracksRegLiveness: true
hasWinCFI: false
registers:
- { id: 0, class: gr64, preferred-register: '' }
- { id: 1, class: gr64, preferred-register: '' }
- { id: 2, class: vr128, preferred-register: '' }
liveins:
- { reg: '$rdi', virtual-reg: '%0' }
- { reg: '$rsi', virtual-reg: '%1' }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
stackProtector: ''
maxCallFrameSize: 4294967295
cvBytesOfCalleeSavedRegisters: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
localFrameSize: 0
savePoint: ''
restorePoint: ''
fixedStack: []
stack: []
constants: []
machineFunctionInfo: {}
body: |
bb.0.entry:
successors: %bb.2(0x30000000), %bb.1(0x50000000)
liveins: $rdi, $rsi
DBG_VALUE $rdi, $noreg, !21, !DIExpression(), debug-location !27
DBG_VALUE $rsi, $noreg, !22, !DIExpression(), debug-location !27
DBG_VALUE $rdi, $noreg, !23, !DIExpression(), debug-location !27
%1:gr64 = COPY $rsi
DBG_VALUE %1, $noreg, !22, !DIExpression(), debug-location !27
%0:gr64 = COPY $rdi
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !27
TEST64rr %0, %0, implicit-def $eflags, debug-location !28
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !27
MOV8mi %0, 1, $noreg, 0, $noreg, 0, debug-location !30 :: (store 1 into %ir.0)
CFI_INSTRUCTION offset $r13, -123
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !21, !DIExpression(), debug-location !27
DBG_VALUE %0, $noreg, !23, !DIExpression(), debug-location !27
JCC_1 %bb.2, 4, implicit $eflags, debug-location !31
JMP_1 %bb.1, debug-location !31
bb.1.if.then:
successors: %bb.2(0x80000000)
%2:vr128 = MOVAPSrm %0, 1, $noreg, 0, $noreg, debug-location !32 :: (load 16 from %ir.p1)
DBG_VALUE %2, $noreg, !24, !DIExpression(), debug-location !33
MOVAPSmr %1, 1, $noreg, 0, $noreg, killed %2, debug-location !34 :: (store 16 into %ir.p2)
bb.2.if.end:
RET 0, debug-location !36
...
---
name: nodebug
alignment: 4
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
failedISel: false
tracksRegLiveness: true
hasWinCFI: false
registers:
- { id: 0, class: gr64, preferred-register: '' }
- { id: 1, class: gr64, preferred-register: '' }
- { id: 2, class: vr128, preferred-register: '' }
liveins:
- { reg: '$rdi', virtual-reg: '%0' }
- { reg: '$rsi', virtual-reg: '%1' }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
stackProtector: ''
maxCallFrameSize: 4294967295
cvBytesOfCalleeSavedRegisters: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
localFrameSize: 0
savePoint: ''
restorePoint: ''
fixedStack: []
stack: []
constants: []
machineFunctionInfo: {}
body: |
bb.0.entry:
successors: %bb.2(0x30000000), %bb.1(0x50000000)
liveins: $rdi, $rsi
%1:gr64 = COPY $rsi
%0:gr64 = COPY $rdi
TEST64rr %0, %0, implicit-def $eflags
MOV8mi %0, 1, $noreg, 0, $noreg, 0 :: (store 1 into %ir.0)
JCC_1 %bb.2, 4, implicit $eflags
JMP_1 %bb.1
bb.1.if.then:
successors: %bb.2(0x80000000)
%2:vr128 = MOVAPSrm %0, 1, $noreg, 0, $noreg :: (load 16 from %ir.p1)
MOVAPSmr %1, 1, $noreg, 0, $noreg, killed %2 :: (store 16 into %ir.p2)
bb.2.if.end:
RET 0
; DEBUG-LABEL: name: debug
; NODEBUG-LABEL: name: nodebug
; CHECK: %3:gr8 = MOV8rm
; CHECK: MOV8mr
; CHECK: %4:gr64 = MOV64rm
; CHECK: MOV64mr
; CHECK: %5:gr32 = MOV32rm
; CHECK: MOV32mr
; CHECK: %6:gr16 = MOV16rm
; CHECK: MOV16mr
; CHECK: %7:gr8 = MOV8rm
; CHECK: MOV8mr
; CHECK: RET 0
; DEBUG-LABEL: name: nodebug
...