This is an archive of the discontinued LLVM Phabricator instance.

Differential D114400

[SampleFDO] Compute BFI if the sample loader changes BPI in FSAFDO
ClosedPublic

Authored by xur on Nov 22 2021, 2:32 PM.

Details

Reviewers
wmi
wenlei
hoy
Commits
rGbf1138491a2b: [SampleFDO] Recompute BFI if the sample loader changes BPI
Summary

The sample loader changes the branch probability but not BFI. Here we force a recompute of BFI if the branch probabilities are changed.
No performance changes with google benchmarks.




This patch also registers FSAFDO MIR passes probably.
Diff Detail
Event Timeline
xur created this revision.Nov 22 2021, 2:32 PM
Herald added subscribers: pengfei, hiraditya.  ·  View Herald TranscriptNov 22 2021, 2:32 PM
xur requested review of this revision.Nov 22 2021, 2:32 PM
Herald added a project: Restricted Project.  ·  View Herald TranscriptNov 22 2021, 2:32 PM
hoy accepted this revision.Nov 22 2021, 3:58 PM
LGTM, thanks.
This revision is now accepted and ready to land.Nov 22 2021, 3:58 PM
Herald added a subscriber: ormris.  ·  View Herald TranscriptNov 22 2021, 3:58 PM
Harbormaster completed remote builds in B135513: Diff 389044.Nov 22 2021, 8:04 PM
Closed by commit rGbf1138491a2b: [SampleFDO] Recompute BFI if the sample loader changes BPI (authored by xur).  ·  Explain WhyNov 23 2021, 1:42 PM
This revision was automatically updated to reflect the committed changes.
xur added a commit: rGbf1138491a2b: [SampleFDO] Recompute BFI if the sample loader changes BPI.
wenlei added inline comments.Nov 23 2021, 3:19 PM
llvm/lib/CodeGen/MIRSampleProfile.cpp


318
When profile loader leads to change, we should probably recalculate BFI. But should BFI calculation be done outside of sample loader given BFI is a separate analysis? The IR BFI is done outside of sample loader.



It feels to me that updating MBFI inside sample loader kind of breaks the componentization.
hoy added inline comments.Nov 23 2021, 3:40 PM
llvm/lib/CodeGen/MIRSampleProfile.cpp


318
Normally you would mark this pass to not preserve BFI and rely on subsequent LazyBFI pass to update BFI automatically. But it's a bit special here in that we are using the BFI right after here.
wenlei added inline comments.Nov 23 2021, 3:43 PM
llvm/lib/CodeGen/MIRSampleProfile.cpp


318
But it's a bit special here in that we are using the BFI right after here.



What is the use you're referring to? I only the use of MBFI->View which is not a real use.



I don't see it as a reason to introduce this tight coupling.
Revision Contents
PathSize
llvm/
lib/
CodeGen/
2 lines
2 lines
test/
CodeGen/
X86/
307 lines
Diff 389044
llvm/lib/CodeGen/CodeGen.cpp
Show First 20 Lines • Show All 52 Lines • ▼ Show 20 Linesvoid llvm::initializeCodeGen(PassRegistry &Registry) {

  initializeLiveDebugValuesPass(Registry);
  initializeLiveDebugValuesPass(Registry);

  initializeLiveDebugVariablesPass(Registry);
  initializeLiveDebugVariablesPass(Registry);

  initializeLiveIntervalsPass(Registry);
  initializeLiveIntervalsPass(Registry);

  initializeLiveRangeShrinkPass(Registry);
  initializeLiveRangeShrinkPass(Registry);

  initializeLiveStacksPass(Registry);
  initializeLiveStacksPass(Registry);

  initializeLiveVariablesPass(Registry);
  initializeLiveVariablesPass(Registry);

  initializeLocalStackSlotPassPass(Registry);
  initializeLocalStackSlotPassPass(Registry);

  initializeLowerIntrinsicsPass(Registry);
  initializeLowerIntrinsicsPass(Registry);

  initializeMIRAddFSDiscriminatorsPass(Registry);

  initializeMIRCanonicalizerPass(Registry);
  initializeMIRCanonicalizerPass(Registry);

  initializeMIRNamerPass(Registry);
  initializeMIRNamerPass(Registry);

  initializeMIRProfileLoaderPassPass(Registry);

  initializeMachineBlockFrequencyInfoPass(Registry);
  initializeMachineBlockFrequencyInfoPass(Registry);

  initializeMachineBlockPlacementPass(Registry);
  initializeMachineBlockPlacementPass(Registry);

  initializeMachineBlockPlacementStatsPass(Registry);
  initializeMachineBlockPlacementStatsPass(Registry);

  initializeMachineCSEPass(Registry);
  initializeMachineCSEPass(Registry);

  initializeMachineCombinerPass(Registry);
  initializeMachineCombinerPass(Registry);

  initializeMachineCopyPropagationPass(Registry);
  initializeMachineCopyPropagationPass(Registry);

  initializeMachineDominatorTreePass(Registry);
  initializeMachineDominatorTreePass(Registry);

  initializeMachineFunctionPrinterPassPass(Registry);
  initializeMachineFunctionPrinterPassPass(Registry);

▲ Show 20 Lines • Show All 58 Lines • Show Last 20 Lines
llvm/lib/CodeGen/MIRSampleProfile.cpp
Show First 20 Lines • Show All 308 Lines • ▼ Show 20 Linesbool MIRProfileLoaderPass::runOnMachineFunction(MachineFunction &MF) {

  MF.RenumberBlocks();
  MF.RenumberBlocks();

  if (ViewBFIBefore && ViewBlockLayoutWithBFI != GVDT_None &&
  if (ViewBFIBefore && ViewBlockLayoutWithBFI != GVDT_None &&

      (ViewBlockFreqFuncName.empty() ||
      (ViewBlockFreqFuncName.empty() ||

       MF.getFunction().getName().equals(ViewBlockFreqFuncName))) {
       MF.getFunction().getName().equals(ViewBlockFreqFuncName))) {

    MBFI->view("MIR_Prof_loader_b." + MF.getName(), false);
    MBFI->view("MIR_Prof_loader_b." + MF.getName(), false);

  }
  }




  bool Changed = MIRSampleLoader->runOnFunction(MF);
  bool Changed = MIRSampleLoader->runOnFunction(MF);

  if (Changed)

    MBFI->calculate(MF, *MBFI->getMBPI(), *&getAnalysis<MachineLoopInfo>());

wenleiUnsubmitted
Not Done
ReplyInline Actions
When profile loader leads to change, we should probably recalculate BFI. But should BFI calculation be done outside of sample loader given BFI is a separate analysis? The IR BFI is done outside of sample loader.



It feels to me that updating MBFI inside sample loader kind of breaks the componentization.
wenlei: When profile loader leads to change, we should probably recalculate BFI. But should BFI…
hoyUnsubmitted
Not Done
ReplyInline Actions
Normally you would mark this pass to not preserve BFI and rely on subsequent LazyBFI pass to update BFI automatically. But it's a bit special here in that we are using the BFI right after here.
hoy: Normally you would mark this pass to not preserve BFI and rely on subsequent LazyBFI pass to…
wenleiUnsubmitted
Not Done
ReplyInline Actions
But it's a bit special here in that we are using the BFI right after here.



What is the use you're referring to? I only the use of MBFI->View which is not a real use.



I don't see it as a reason to introduce this tight coupling.
wenlei: > But it's a bit special here in that we are using the BFI right after here. 

What is the use…



  if (ViewBFIAfter && ViewBlockLayoutWithBFI != GVDT_None &&
  if (ViewBFIAfter && ViewBlockLayoutWithBFI != GVDT_None &&

      (ViewBlockFreqFuncName.empty() ||
      (ViewBlockFreqFuncName.empty() ||

       MF.getFunction().getName().equals(ViewBlockFreqFuncName))) {
       MF.getFunction().getName().equals(ViewBlockFreqFuncName))) {

    MBFI->view("MIR_prof_loader_a." + MF.getName(), false);
    MBFI->view("MIR_prof_loader_a." + MF.getName(), false);

  }
  }




  return Changed;
  return Changed;

Show All 19 Lines
llvm/test/CodeGen/X86/fsafdo_test3.ll
  • This file was added.
; RUN: llvm-profdata merge --sample -profile-isfs -o %t.afdo %S/Inputs/fsloader.afdo

; RUN: llc -enable-fs-discriminator -fs-profile-file=%t.afdo -disable-ra-fsprofile-loader=false -disable-layout-fsprofile-loader=false -print-machine-bfi -print-bfi-func-name=foo -print-before=fs-profile-loader -stop-after=fs-profile-loader < %s 2>&1 | FileCheck %s --check-prefix=BFI

;

;;

;; C source code for the test (compiler at -O3):

;; // A test case for loop unroll.

;;

;; __attribute__((noinline)) int bar(int i){

;;   volatile int j;

;;   j = i;

;;   return j;

;; }

;;

;; unsigned sum;

;; __attribute__((noinline)) void work(int i){

;;   if (sum % 7)

;;     sum += i;

;;   else

;;     sum -= i;

;; }

;;

;; __attribute__((noinline)) void foo(){

;;   int i, j;

;;   for (j = 0; j < 48; j++)

;;     for (i = 0; i < 4; i++) {

;;       int ii = bar(i+j*48);

;;       if (ii % 2)

;;         work(ii*2);

;;       if (ii % 4)

;;         work(ii*3);

;;   }

;; }

;;

;; int main() {

;;   int i;

;;   for (i = 0; i < 10000000; i++) {

;;     foo();

;;   }

;; }

;;

;; Check BFI before and after



; BFI: block-frequency-info: foo

; BFI:  - BB0[entry]: float = 1.0, int = 8, count = 4268

; BFI:  - BB1[for.cond1.preheader]: float = 59.967, int = 479, count = 255547

; BFI:  - BB2[if.then]: float = 2.5405, int = 20, count = 10670

; BFI:  - BB3[if.end]: float = 59.967, int = 479, count = 255547

; BFI:  - BB4[if.then7]: float = 2.5405, int = 20, count = 10670

; BFI:  - BB5[if.end9]: float = 59.967, int = 479, count = 255547

; BFI:  - BB6[if.then.1]: float = 2.5405, int = 20, count = 10670

; BFI:  - BB7[if.end.1]: float = 59.967, int = 479, count = 255547

; BFI:  - BB8[if.then7.1]: float = 2.5405, int = 20, count = 10670

; BFI:  - BB9[if.end9.1]: float = 59.967, int = 479, count = 255547

; BFI:  - BB10[if.then.2]: float = 2.5405, int = 20, count = 10670

; BFI:  - BB11[if.end.2]: float = 59.967, int = 479, count = 255547

; BFI:  - BB12[if.then7.2]: float = 2.5405, int = 20, count = 10670

; BFI:  - BB13[if.end9.2]: float = 59.967, int = 479, count = 255547

; BFI:  - BB14[if.then.3]: float = 2.5405, int = 20, count = 10670

; BFI:  - BB15[if.end.3]: float = 59.967, int = 479, count = 255547

; BFI:  - BB16[if.then7.3]: float = 2.5405, int = 20, count = 10670

; BFI:  - BB17[if.end9.3]: float = 59.967, int = 479, count = 255547

; BFI:  - BB18[for.end12]: float = 1.0, int = 8, count = 4268

;

; BFI: # *** IR Dump Before SampleFDO loader in MIR (fs-profile-loader) ***:

; BFI: # End machine code for function foo.

;

; BFI: block-frequency-info: foo

; BFI:  - BB0[entry]: float = 1.0, int = 8, count = 4268

; BFI:  - BB1[for.cond1.preheader]: float = 66.446, int = 531, count = 283289

; BFI:  - BB2[if.then]: float = 2.7041, int = 21, count = 11204

; BFI:  - BB3[if.end]: float = 66.446, int = 531, count = 283289

; BFI:  - BB4[if.then7]: float = 2.7041, int = 21, count = 11204

; BFI:  - BB5[if.end9]: float = 66.446, int = 531, count = 283289

; BFI:  - BB6[if.then.1]: float = 65.351, int = 522, count = 278487

; BFI:  - BB7[if.end.1]: float = 66.446, int = 531, count = 283289

; BFI:  - BB8[if.then7.1]: float = 66.446, int = 531, count = 283289

; BFI:  - BB9[if.end9.1]: float = 66.446, int = 531, count = 283289

; BFI:  - BB10[if.then.2]: float = 2.7041, int = 21, count = 11204

; BFI:  - BB11[if.end.2]: float = 66.446, int = 531, count = 283289

; BFI:  - BB12[if.then7.2]: float = 65.405, int = 523, count = 279021

; BFI:  - BB13[if.end9.2]: float = 66.446, int = 531, count = 283289

; BFI:  - BB14[if.then.3]: float = 61.075, int = 488, count = 260348

; BFI:  - BB15[if.end.3]: float = 66.446, int = 531, count = 283289

; BFI:  - BB16[if.then7.3]: float = 54.846, int = 438, count = 233673

; BFI:  - BB17[if.end9.3]: float = 66.446, int = 531, count = 283289

; BFI:  - BB18[for.end12]: float = 1.0, int = 8, count = 4268



target triple = "x86_64-unknown-linux-gnu"



@sum = dso_local local_unnamed_addr global i32 0, align 4



declare i32 @bar(i32 %i) #0

declare void @work(i32 %i) #2

declare i32 @main() #3



define dso_local void @foo() local_unnamed_addr #0 !dbg !61 !prof !62 {

entry:

  br label %for.cond1.preheader, !dbg !63



for.cond1.preheader:

  %j.024 = phi i32 [ 0, %entry ], [ %inc11, %if.end9.3 ]

  %mul = mul nuw nsw i32 %j.024, 48

  %call = tail call i32 @bar(i32 %mul), !dbg !65, !prof !66

  %0 = and i32 %call, 1, !dbg !67

  %tobool.not = icmp eq i32 %0, 0, !dbg !67

  br i1 %tobool.not, label %if.end, label %if.then, !dbg !68, !prof !69



if.then:

  %mul4 = shl nsw i32 %call, 1, !dbg !70

  tail call void @work(i32 %mul4), !dbg !71, !prof !72

  br label %if.end, !dbg !71



if.end:

  %1 = and i32 %call, 3, !dbg !73

  %tobool6.not = icmp eq i32 %1, 0, !dbg !73

  br i1 %tobool6.not, label %if.end9, label %if.then7, !dbg !74, !prof !69



if.then7:

  %mul8 = mul nsw i32 %call, 3, !dbg !75

  tail call void @work(i32 %mul8), !dbg !76, !prof !72

  br label %if.end9, !dbg !76



if.end9:

  %add.1 = or i32 %mul, 1, !dbg !77

  %call.1 = tail call i32 @bar(i32 %add.1), !dbg !65, !prof !66

  %2 = and i32 %call.1, 1, !dbg !67

  %tobool.not.1 = icmp eq i32 %2, 0, !dbg !67

  br i1 %tobool.not.1, label %if.end.1, label %if.then.1, !dbg !68, !prof !69



if.then.1:

  %mul4.1 = shl nsw i32 %call.1, 1, !dbg !70

  tail call void @work(i32 %mul4.1), !dbg !71, !prof !72

  br label %if.end.1, !dbg !71



if.end.1:

  %3 = and i32 %call.1, 3, !dbg !73

  %tobool6.not.1 = icmp eq i32 %3, 0, !dbg !73

  br i1 %tobool6.not.1, label %if.end9.1, label %if.then7.1, !dbg !74, !prof !69



if.then7.1:

  %mul8.1 = mul nsw i32 %call.1, 3, !dbg !75

  tail call void @work(i32 %mul8.1), !dbg !76, !prof !72

  br label %if.end9.1, !dbg !76



if.end9.1:

  %add.2 = or i32 %mul, 2, !dbg !77

  %call.2 = tail call i32 @bar(i32 %add.2), !dbg !65, !prof !66

  %4 = and i32 %call.2, 1, !dbg !67

  %tobool.not.2 = icmp eq i32 %4, 0, !dbg !67

  br i1 %tobool.not.2, label %if.end.2, label %if.then.2, !dbg !68, !prof !69



if.then.2:

  %mul4.2 = shl nsw i32 %call.2, 1, !dbg !70

  tail call void @work(i32 %mul4.2), !dbg !71, !prof !72

  br label %if.end.2, !dbg !71



if.end.2:

  %5 = and i32 %call.2, 3, !dbg !73

  %tobool6.not.2 = icmp eq i32 %5, 0, !dbg !73

  br i1 %tobool6.not.2, label %if.end9.2, label %if.then7.2, !dbg !74, !prof !69



if.then7.2:

  %mul8.2 = mul nsw i32 %call.2, 3, !dbg !75

  tail call void @work(i32 %mul8.2), !dbg !76, !prof !72

  br label %if.end9.2, !dbg !76



if.end9.2:

  %add.3 = or i32 %mul, 3, !dbg !77

  %call.3 = tail call i32 @bar(i32 %add.3), !dbg !65, !prof !66

  %6 = and i32 %call.3, 1, !dbg !67

  %tobool.not.3 = icmp eq i32 %6, 0, !dbg !67

  br i1 %tobool.not.3, label %if.end.3, label %if.then.3, !dbg !68, !prof !69



if.then.3:

  %mul4.3 = shl nsw i32 %call.3, 1, !dbg !70

  tail call void @work(i32 %mul4.3), !dbg !71, !prof !72

  br label %if.end.3, !dbg !71



if.end.3:

  %7 = and i32 %call.3, 3, !dbg !73

  %tobool6.not.3 = icmp eq i32 %7, 0, !dbg !73

  br i1 %tobool6.not.3, label %if.end9.3, label %if.then7.3, !dbg !74, !prof !69



if.then7.3:

  %mul8.3 = mul nsw i32 %call.3, 3, !dbg !75

  tail call void @work(i32 %mul8.3), !dbg !76, !prof !72

  br label %if.end9.3, !dbg !76



if.end9.3:

  %inc11 = add nuw nsw i32 %j.024, 1, !dbg !78

  %exitcond.not = icmp eq i32 %inc11, 48, !dbg !80

  br i1 %exitcond.not, label %for.end12, label %for.cond1.preheader, !dbg !63, !prof !81, !llvm.loop !82



for.end12:

  ret void, !dbg !86

}



attributes #0 = { nofree noinline nounwind uwtable "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "use-sample-profile" }

attributes #1 = { argmemonly mustprogress nofree nosync nounwind willreturn }

attributes #2 = { mustprogress nofree noinline norecurse nosync nounwind uwtable willreturn "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "use-sample-profile" }

attributes #3 = { nofree nounwind uwtable "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "use-sample-profile" }



!llvm.dbg.cu = !{!0}

!llvm.module.flags = !{!2, !3, !4, !5, !6, !35}

!llvm.ident = !{!40}



!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 14.0.0 (https://github.com/llvm/llvm-project.git 755f5e23159796d727c3d95d60894a52eb675b1b)", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, splitDebugInlining: false, debugInfoForProfiling: true, nameTableKind: None)

!1 = !DIFile(filename: "/tmp/aaa.c", directory: "/mnt/ssd/xur/llvm_dev/gitwork/llvm-project/rel")

!2 = !{i32 7, !"Dwarf Version", i32 4}

!3 = !{i32 2, !"Debug Info Version", i32 3}

!4 = !{i32 1, !"wchar_size", i32 4}

!5 = !{i32 7, !"uwtable", i32 1}

!6 = !{i32 1, !"ProfileSummary", !7}

!7 = !{!8, !9, !10, !11, !12, !13, !14, !15, !16, !17}

!8 = !{!"ProfileFormat", !"SampleProfile"}

!9 = !{!"TotalCount", i64 9484871}

!10 = !{!"MaxCount", i64 1246499}

!11 = !{!"MaxInternalCount", i64 0}

!12 = !{!"MaxFunctionCount", i64 1346190}

!13 = !{!"NumCounts", i64 31}

!14 = !{!"NumFunctions", i64 4}

!15 = !{!"IsPartialProfile", i64 0}

!16 = !{!"PartialProfileRatio", double 0.000000e+00}

!17 = !{!"DetailedSummary", !18}

!18 = !{!19, !20, !21, !22, !23, !24, !25, !26, !27, !28, !29, !30, !31, !32, !33, !34}

!19 = !{i32 10000, i64 1246499, i32 2}

!20 = !{i32 100000, i64 1246499, i32 2}

!21 = !{i32 200000, i64 1246499, i32 2}

!22 = !{i32 300000, i64 1056020, i32 4}

!23 = !{i32 400000, i64 1056020, i32 4}

!24 = !{i32 500000, i64 283590, i32 6}

!25 = !{i32 600000, i64 279149, i32 9}

!26 = !{i32 700000, i64 278916, i32 12}

!27 = !{i32 800000, i64 269485, i32 15}

!28 = !{i32 900000, i64 260670, i32 19}

!29 = !{i32 950000, i64 234082, i32 22}

!30 = !{i32 990000, i64 234082, i32 22}

!31 = !{i32 999000, i64 4156, i32 27}

!32 = !{i32 999900, i64 4045, i32 29}

!33 = !{i32 999990, i64 4045, i32 29}

!34 = !{i32 999999, i64 4045, i32 29}

!35 = !{i32 5, !"CG Profile", !36}

!36 = !{!37, !38, !39}

!37 = !{void ()* @foo, i32 (i32)* @bar, i64 1022188}

!38 = !{void ()* @foo, void (i32)* @work, i64 85360}

!39 = !{i32 ()* @main, void ()* @foo, i64 2080}

!40 = !{!"clang version 14.0.0 (https://github.com/llvm/llvm-project.git 755f5e23159796d727c3d95d60894a52eb675b1b)"}

!41 = distinct !DISubprogram(name: "bar", scope: !42, file: !42, line: 3, type: !43, scopeLine: 3, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !44)

!42 = !DIFile(filename: "/tmp/aaa.c", directory: "")

!43 = !DISubroutineType(types: !44)

!44 = !{}

!45 = !{!"function_entry_count", i64 1076806}

!46 = !DILocation(line: 4, column: 3, scope: !41)

!47 = !DILocation(line: 5, column: 5, scope: !41)

!48 = !{!49, !49, i64 0}

!49 = !{!"int", !50, i64 0}

!50 = !{!"omnipotent char", !51, i64 0}

!51 = !{!"Simple C/C++ TBAA"}

!52 = !DILocation(line: 6, column: 10, scope: !41)

!53 = !DILocation(line: 7, column: 1, scope: !41)

!54 = !DILocation(line: 6, column: 3, scope: !41)

!55 = distinct !DISubprogram(name: "work", scope: !42, file: !42, line: 10, type: !43, scopeLine: 10, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !44)

!56 = !{!"function_entry_count", i64 1346191}

!57 = !DILocation(line: 11, column: 7, scope: !55)

!58 = !DILocation(line: 11, column: 11, scope: !55)

!59 = !DILocation(line: 0, scope: !55)

!60 = !DILocation(line: 15, column: 1, scope: !55)

!61 = distinct !DISubprogram(name: "foo", scope: !42, file: !42, line: 17, type: !43, scopeLine: 17, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !44)

!62 = !{!"function_entry_count", i64 4268}

!63 = !DILocation(line: 19, column: 3, scope: !64)

!64 = !DILexicalBlockFile(scope: !61, file: !42, discriminator: 1)

!65 = !DILocation(line: 21, column: 16, scope: !61)

!66 = !{!"branch_weights", i32 272442}

!67 = !DILocation(line: 22, column: 14, scope: !61)

!68 = !DILocation(line: 22, column: 11, scope: !61)

!69 = !{!"branch_weights", i32 260902, i32 11542}

!70 = !DILocation(line: 23, column: 16, scope: !61)

!71 = !DILocation(line: 23, column: 9, scope: !61)

!72 = !{!"branch_weights", i32 11541}

!73 = !DILocation(line: 24, column: 14, scope: !61)

!74 = !DILocation(line: 24, column: 11, scope: !61)

!75 = !DILocation(line: 25, column: 16, scope: !61)

!76 = !DILocation(line: 25, column: 9, scope: !61)

!77 = !DILocation(line: 21, column: 21, scope: !61)

!78 = !DILocation(line: 19, column: 24, scope: !79)

!79 = !DILexicalBlockFile(scope: !61, file: !42, discriminator: 2)

!80 = !DILocation(line: 19, column: 17, scope: !64)

!81 = !{!"branch_weights", i32 4269, i32 251732}

!82 = distinct !{!82, !83, !84, !85}

!83 = !DILocation(line: 19, column: 3, scope: !61)

!84 = !DILocation(line: 26, column: 3, scope: !61)

!85 = !{!"llvm.loop.mustprogress"}

!86 = !DILocation(line: 27, column: 1, scope: !61)

!87 = distinct !DISubprogram(name: "main", scope: !42, file: !42, line: 29, type: !43, scopeLine: 29, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !44)

!88 = !{!"function_entry_count", i64 1}

!89 = !DILocation(line: 31, column: 3, scope: !90)

!90 = !DILexicalBlockFile(scope: !87, file: !42, discriminator: 1)

!91 = !DILocation(line: 32, column: 5, scope: !87)

!92 = !{!"branch_weights", i32 4156}

!93 = !DILocation(line: 31, column: 30, scope: !94)

!94 = !DILexicalBlockFile(scope: !87, file: !42, discriminator: 2)

!95 = !DILocation(line: 31, column: 17, scope: !90)

!96 = !{!"branch_weights", i32 2, i32 4157}

!97 = distinct !{!97, !98, !99, !85}

!98 = !DILocation(line: 31, column: 3, scope: !87)

!99 = !DILocation(line: 33, column: 3, scope: !87)

!100 = !DILocation(line: 34, column: 1, scope: !87)