[CSSPGO][llvm-profgen] Report samples for untrackable frames.

[hoy]

Summary

Fixing an issue where samples collected for an untrackable frame is not reported. An untrackable frame refers to a frame whose caller is untrackable due to missing debug info or pseudo probe. Though the frame is connected to its parent frame through the frame pointer chain at runtime, the compiler cannot build the connection without debug info or pseudo probe. In such case we just need to report the untrackable frame as the base frame and all of its child frames.

With more samples reported I'm seeing this improves the performance of an internal benchmark by 2.5%.

[wlei]

llvm/tools/llvm-profgen/PerfReader.cpp

150

Previously we always missed the context from the first frame, right? like foo @ bar, it will miss foo.

[hoy]

llvm/tools/llvm-profgen/PerfReader.cpp

150

Exactly. And if bar is inlined into foo, the issue will cause samples of bar to be ignored. For the attached regression test, without the fix, the output profile is empty.

[wlei]

LGTM, thanks for adding useful debug printer.

llvm/tools/llvm-profgen/PerfReader.cpp

150

I see, thanks for the fix!!

[wenlei]

llvm/test/tools/llvm-profgen/Inputs/truncated-pseudoprobe.ll

89

Is this the one missing probe id in discriminator?

llvm/tools/llvm-profgen/PerfReader.cpp

147

Why do we need this check? Is it ok for leaf to not find an associated call probe?

I'm thinking that the warning can be moved into ProbeStack::pushFrame, so it can be more specific, and the message can be "Truncated context due to missing call probe."

llvm/tools/llvm-profgen/PerfReader.h

611

is this removal intentional?

[hoy]

llvm/test/tools/llvm-profgen/Inputs/truncated-pseudoprobe.ll

89

Yes. The discriminator field is intentionally set empty to mimid the real case.

llvm/tools/llvm-profgen/PerfReader.cpp

147

isLeafFrame here means the address is from LBR not the stack samples. So it doesn't refer to a function frame and it'll mostly have no call probe associated.

Yes, moving the assert into ProbeStack::pushFrame sounds good.

llvm/tools/llvm-profgen/PerfReader.h

611

Yes, there's already a private preceding this.

[hoy]

Addressing Wenlei's feebacks.

[wenlei]

llvm/tools/llvm-profgen/PerfReader.h

446

nit, mention "due to missing call probe" now that this is moved into ProbeStack.

[wenlei]

lgtm with a nit, thanks for the fix.

[hoy]

Addressing Wenlei's comment.

llvm/tools/llvm-profgen/PerfReader.h

446

Done. Sorry for missing this in your previous comment.

llvm/test/tools/llvm-profgen/Inputs/truncated-pseudoprobe.ll

+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+@.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
+
+; Function Attrs: nounwind readnone uwtable willreturn
+define dso_local i32 @bar(i32 %x, i32 %y) local_unnamed_addr #0 !dbg !10 {
+entry:
+  call void @llvm.dbg.value(metadata i32 %x, metadata !15, metadata !DIExpression()), !dbg !17
+  call void @llvm.dbg.value(metadata i32 %y, metadata !16, metadata !DIExpression()), !dbg !17
+  call void @llvm.pseudoprobe(i64 -2012135647395072713, i64 1, i32 0, i64 -1), !dbg !18
+  %rem = srem i32 %x, 3, !dbg !20
+  %tobool.not = icmp eq i32 %rem, 0, !dbg !20
+  call void @llvm.pseudoprobe(i64 -2012135647395072713, i64 2, i32 2, i64 -1), !dbg !21
+  call void @llvm.pseudoprobe(i64 -2012135647395072713, i64 3, i32 2, i64 -1), !dbg !23
+  %0 = sub i32 0, %y, !dbg !24
+  %retval.0.p = select i1 %tobool.not, i32 %y, i32 %0, !dbg !24
+  %retval.0 = add i32 %retval.0.p, %x, !dbg !24
+  call void @llvm.pseudoprobe(i64 -2012135647395072713, i64 4, i32 0, i64 -1), !dbg !25
+  ret i32 %retval.0, !dbg !25
+}
+
+; Function Attrs: noinline nounwind uwtable
+define dso_local void @foo() local_unnamed_addr #1 !dbg !26 {
+entry:
+  call void @llvm.pseudoprobe(i64 6699318081062747564, i64 1, i32 0, i64 -1), !dbg !32
+  call void @llvm.dbg.value(metadata i32 0, metadata !30, metadata !DIExpression()), !dbg !33
+  call void @llvm.dbg.value(metadata i32 0, metadata !31, metadata !DIExpression()), !dbg !33
+  call void @llvm.pseudoprobe(i64 6699318081062747564, i64 2, i32 0, i64 -1), !dbg !34
+  call void @llvm.dbg.value(metadata i32 1, metadata !31, metadata !DIExpression()), !dbg !33
+  br label %while.body, !dbg !35
+
+while.body:                                       ; preds = %entry, %if.end
+  %inc8 = phi i32 [ 1, %entry ], [ %inc, %if.end ]
+  %s.07 = phi i32 [ 0, %entry ], [ %s.1, %if.end ]
+  call void @llvm.dbg.value(metadata i32 %s.07, metadata !30, metadata !DIExpression()), !dbg !33
+  call void @llvm.pseudoprobe(i64 6699318081062747564, i64 3, i32 0, i64 -1), !dbg !36
+  %rem = urem i32 %inc8, 91, !dbg !38
+  %tobool.not = icmp eq i32 %rem, 0, !dbg !38
+  br i1 %tobool.not, label %if.else, label %if.then, !dbg !39
+
+if.then:                                          ; preds = %while.body
+  call void @llvm.pseudoprobe(i64 6699318081062747564, i64 4, i32 0, i64 -1), !dbg !40
+  call void @llvm.dbg.value(metadata i32 %inc8, metadata !15, metadata !DIExpression()) #6, !dbg !41
+  call void @llvm.dbg.value(metadata i32 %s.07, metadata !16, metadata !DIExpression()) #6, !dbg !41
+  call void @llvm.pseudoprobe(i64 -2012135647395072713, i64 1, i32 0, i64 -1) #6, !dbg !44
+  %rem.i = urem i32 %inc8, 3, !dbg !45
+  %tobool.not.i = icmp eq i32 %rem.i, 0, !dbg !45
+  call void @llvm.pseudoprobe(i64 -2012135647395072713, i64 2, i32 2, i64 -1) #6, !dbg !46
+  call void @llvm.pseudoprobe(i64 -2012135647395072713, i64 3, i32 2, i64 -1) #6, !dbg !47
+  %0 = sub i32 0, %s.07, !dbg !48
+  %retval.0.p.i = select i1 %tobool.not.i, i32 %s.07, i32 %0, !dbg !48
+  %retval.0.i = add i32 %retval.0.p.i, %inc8, !dbg !48
+  call void @llvm.pseudoprobe(i64 -2012135647395072713, i64 4, i32 0, i64 -1) #6, !dbg !49
+  call void @llvm.dbg.value(metadata i32 %retval.0.i, metadata !30, metadata !DIExpression()), !dbg !33
+  br label %if.end, !dbg !50
+
+if.else:                                          ; preds = %while.body
+  call void @llvm.pseudoprobe(i64 6699318081062747564, i64 5, i32 0, i64 -1), !dbg !51
+  %add = add nsw i32 %s.07, 30, !dbg !51
+  call void @llvm.dbg.value(metadata i32 %add, metadata !30, metadata !DIExpression()), !dbg !33
+  br label %if.end
+
+if.end:                                           ; preds = %if.else, %if.then
+  %s.1 = phi i32 [ %retval.0.i, %if.then ], [ %add, %if.else ], !dbg !52
+  call void @llvm.dbg.value(metadata i32 %s.1, metadata !30, metadata !DIExpression()), !dbg !33
+  call void @llvm.pseudoprobe(i64 6699318081062747564, i64 6, i32 0, i64 -1), !dbg !35
+  call void @llvm.dbg.value(metadata i32 %inc8, metadata !31, metadata !DIExpression()), !dbg !33
+  call void @llvm.pseudoprobe(i64 6699318081062747564, i64 2, i32 0, i64 -1), !dbg !34
+  %inc = add nuw nsw i32 %inc8, 1, !dbg !34
+  call void @llvm.dbg.value(metadata i32 %inc, metadata !31, metadata !DIExpression()), !dbg !33
+  %exitcond.not = icmp eq i32 %inc, 16000001, !dbg !53
+  br i1 %exitcond.not, label %while.end, label %while.body, !dbg !35, !llvm.loop !54
+
+while.end:                                        ; preds = %if.end
+  call void @llvm.pseudoprobe(i64 6699318081062747564, i64 7, i32 0, i64 -1), !dbg !57
+  %call1 = call i32 (i8*, ...) @printf(i8* nonnull dereferenceable(1) getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0), i32 %s.1), !dbg !58
+  ret void, !dbg !60
+}
+
+; Function Attrs: nofree nounwind
+declare dso_local noundef i32 @printf(i8* nocapture noundef readonly, ...) local_unnamed_addr #2
+
+; Function Attrs: nounwind uwtable
+define dso_local i32 @main() local_unnamed_addr #3 !dbg !61 {
+entry:
+  call void @llvm.pseudoprobe(i64 -2624081020897602054, i64 1, i32 0, i64 -1), !dbg !64
+  call void @foo(), !dbg !65
+  ret i32 0, !dbg !67
+}
+
+; Function Attrs: inaccessiblememonly nounwind willreturn
+declare void @llvm.pseudoprobe(i64, i64, i32, i64) #4
+
+; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
+declare void @llvm.dbg.value(metadata, metadata, metadata) #5
+
+attributes #0 = { nounwind readnone uwtable willreturn "disable-tail-calls"="true" "frame-pointer"="all" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { noinline nounwind uwtable "disable-tail-calls"="true" "frame-pointer"="all" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #2 = { nofree nounwind "disable-tail-calls"="true" "frame-pointer"="all" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #3 = { nounwind uwtable "disable-tail-calls"="true" "frame-pointer"="all" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #4 = { inaccessiblememonly nounwind willreturn }
+attributes #5 = { nofree nosync nounwind readnone speculatable willreturn }
+attributes #6 = { nounwind }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5}
+!llvm.ident = !{!6}
+!llvm.pseudo_probe_desc = !{!7, !8, !9}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 12.0.0", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "test.c", directory: "test")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{!"clang version 12.0.0"}
+!7 = !{i64 -2012135647395072713, i64 72617220756, !"bar", null}
+!8 = !{i64 6699318081062747564, i64 563088904013236, !"foo", null}
+!9 = !{i64 -2624081020897602054, i64 281479271677951, !"main", null}
+!10 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 3, type: !11, scopeLine: 3, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !14)
+!11 = !DISubroutineType(types: !12)
+!12 = !{!13, !13, !13}
+!13 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!14 = !{!15, !16}
+!15 = !DILocalVariable(name: "x", arg: 1, scope: !10, file: !1, line: 3, type: !13)
+!16 = !DILocalVariable(name: "y", arg: 2, scope: !10, file: !1, line: 3, type: !13)
+!17 = !DILocation(line: 0, scope: !10)
+!18 = !DILocation(line: 4, column: 9, scope: !19)
+!19 = distinct !DILexicalBlock(scope: !10, file: !1, line: 4, column: 9)
+!20 = !DILocation(line: 4, column: 11, scope: !19)
+!21 = !DILocation(line: 5, column: 18, scope: !22)
+!22 = distinct !DILexicalBlock(scope: !19, file: !1, line: 4, column: 16)
+!23 = !DILocation(line: 7, column: 14, scope: !10)
+!24 = !DILocation(line: 4, column: 9, scope: !10)
+!25 = !DILocation(line: 8, column: 1, scope: !10)
+!26 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 11, type: !27, scopeLine: 11, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !29)
+!27 = !DISubroutineType(types: !28)
+!28 = !{null}
+!29 = !{!30, !31}
+!30 = !DILocalVariable(name: "s", scope: !26, file: !1, line: 12, type: !13)
+!31 = !DILocalVariable(name: "i", scope: !26, file: !1, line: 12, type: !13)
+!32 = !DILocation(line: 12, column: 5, scope: !26)
+!33 = !DILocation(line: 0, scope: !26)
+!34 = !DILocation(line: 13, column: 15, scope: !26)
+!35 = !DILocation(line: 13, column: 7, scope: !26)
+!36 = !DILocation(line: 14, column: 17, scope: !37)
+!37 = distinct !DILexicalBlock(scope: !26, file: !1, line: 14, column: 17)
+!38 = !DILocation(line: 14, column: 19, scope: !37)
+!39 = !DILocation(line: 14, column: 17, scope: !26)
+!40 = !DILocation(line: 14, column: 33, scope: !37)
+!41 = !DILocation(line: 0, scope: !10, inlinedAt: !42)
+!42 = distinct !DILocation(line: 14, column: 29, scope: !43)
+!43 = !DILexicalBlockFile(scope: !37, file: !1, discriminator: 186646599)
+!44 = !DILocation(line: 4, column: 9, scope: !19, inlinedAt: !42)
+!45 = !DILocation(line: 4, column: 11, scope: !19, inlinedAt: !42)
+!46 = !DILocation(line: 5, column: 18, scope: !22, inlinedAt: !42)
+!47 = !DILocation(line: 7, column: 14, scope: !10, inlinedAt: !42)
+!48 = !DILocation(line: 4, column: 9, scope: !10, inlinedAt: !42)
+!49 = !DILocation(line: 8, column: 1, scope: !10, inlinedAt: !42)
+!50 = !DILocation(line: 14, column: 25, scope: !37)
+!51 = !DILocation(line: 14, column: 47, scope: !37)
+!52 = !DILocation(line: 0, scope: !37)
+!53 = !DILocation(line: 13, column: 18, scope: !26)
+!54 = distinct !{!54, !35, !55, !56}
+!55 = !DILocation(line: 14, column: 50, scope: !26)
+!56 = !{!"llvm.loop.mustprogress"}
+!57 = !DILocation(line: 15, column: 31, scope: !26)
+!58 = !DILocation(line: 15, column: 9, scope: !59)
+!59 = !DILexicalBlockFile(scope: !26, file: !1, discriminator: 186646607)
+!60 = !DILocation(line: 16, column: 1, scope: !26)
+!61 = distinct !DISubprogram(name: "main", scope: !1, file: !1, line: 18, type: !62, scopeLine: 18, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
+!62 = !DISubroutineType(types: !63)
+!63 = !{!13}
+!64 = !DILocation(line: 19, column: 5, scope: !61)
+!65 = !DILocation(line: 19, column: 5, scope: !66)
+!66 = !DILexicalBlockFile(scope: !61, file: !1, discriminator: 7)
+!67 = !DILocation(line: 20, column: 7, scope: !61)

llvm/test/tools/llvm-profgen/Inputs/truncated-pseudoprobe.perfscript

+PERF_RECORD_MMAP2 1664112/1664112: [0x400000(0x1000) @ 0 08:11 806256818 82060973]: r-xp truncated-pseudoprobe.perfbin
+            40057d
+            4005b9
+      7f67469af555
+ 0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0  0x40059f/0x400553/P/-/-/0 

llvm/test/tools/llvm-profgen/truncated-pseudoprobe.test

+; RUN: llvm-profgen --perfscript=%S/Inputs/truncated-pseudoprobe.perfscript --binary=%S/Inputs/truncated-pseudoprobe.perfbin --output=%t
+; RUN: FileCheck %s --input-file %t
+
+; CHECK:     [foo]:75:0
+; CHECK-NEXT:  2: 15
+; CHECK-NEXT:  3: 15
+; CHECK-NEXT:  4: 15
+; CHECK-NEXT:  6: 15
+; CHECK-NEXT:  8: 15 bar:15
+; CHECK-NEXT:  !CFGChecksum: 563088904013236
+; CHECK-NEXT:  !Attributes: 0
+; CHECK:     [foo:8 @ bar]:30:15
+; CHECK-NEXT:  1: 15
+; CHECK-NEXT:  2: 18446744073709551615
+; CHECK-NEXT:  3: 18446744073709551615
+; CHECK-NEXT:  4: 15
+; CHECK-NEXT:  !CFGChecksum: 72617220756
+; CHECK-NEXT:  !Attributes: 1
+
+; truncated-pseudoprobe.perfbin is from the following compile commands:
+; llc -pseudo-probe-for-profiling truncated-pseudoprobe.ll -filetype=obj -o truncated-pseudoprobe.o
+; clang truncated-pseudoprobe.o -o truncated-pseudoprobe.perfbin

llvm/tools/llvm-profgen/PerfReader.h

struct LBREntry {
   uint64_t Source = 0;
   uint64_t Target = 0;
   // An artificial branch stands for a series of consecutive branches starting
   // from the current binary with a transition through external code and
   // eventually landing back in the current binary.
   bool IsArtificial = false;
   LBREntry(uint64_t S, uint64_t T, bool I)
       : Source(S), Target(T), IsArtificial(I) {}
+
+#ifndef NDEBUG
+  void print() const {
+    dbgs() << "from " << format("%#010x", Source) << " to "
+           << format("%#010x", Target);
+    if (IsArtificial)
+      dbgs() << " Artificial";
+  }
+#endif
 };
 
+#ifndef NDEBUG
+static inline void printLBRStack(const SmallVectorImpl<LBREntry> &LBRStack) {
+  for (size_t I = 0; I < LBRStack.size(); I++) {
+    dbgs() << "[" << I << "] ";
+    LBRStack[I].print();
+    dbgs() << "\n";
+  }
+}
+
+static inline void printCallStack(const SmallVectorImpl<uint64_t> &CallStack) {
+  for (size_t I = 0; I < CallStack.size(); I++) {
+    dbgs() << "[" << I << "] " << format("%#010x", CallStack[I]) << "\n";
+  }
+}
+#endif
+
 // Hash interface for generic data of type T
 // Data should implement a \fn getHashCode and a \fn isEqual
 // Currently getHashCode is non-virtual to avoid the overhead of calling vtable,
 // i.e we explicitly calculate hash of derived class, assign to base class's
 // HashCode. This also provides the flexibility for calculating the hash code
 // incrementally(like rolling hash) during frame stack unwinding since unwinding
 // only changes the leaf of frame stack. \fn isEqual is a virtual function,
 // which will have perf overhead. In the future, if we redesign a better hash
for (const auto &Value : CallStack) {
       Hash = HashCombine(Hash, Value);
     }
     for (const auto &Entry : LBRStack) {
       Hash = HashCombine(Hash, Entry.Source);
       Hash = HashCombine(Hash, Entry.Target);
     }
     HashCode = Hash;
   }
+
+#ifndef NDEBUG
+  __attribute__((used)) void print() const {
+    dbgs() << "LBR stack\n";
+    printLBRStack(LBRStack);
+    dbgs() << "Call stack\n";
+    printCallStack(CallStack);
+  }
+#endif
 };
 
 // After parsing the sample, we record the samples by aggregating them
 // into this counter. The key stores the sample data and the value is
 // the sample repeat times.
 using AggregatedCounter =
     std::unordered_map<Hashable<PerfSample>, uint64_t,
                        Hashable<PerfSample>::Hash, Hashable<PerfSample>::Equal>;
struct ProfiledFrame {
     }
     void recordRangeCount(uint64_t Start, uint64_t End, uint64_t Count) {
       RangeSamples.emplace_back(std::make_tuple(Start, End, Count));
     }
     void recordBranchCount(uint64_t Source, uint64_t Target, uint64_t Count) {
       BranchSamples.emplace_back(std::make_tuple(Source, Target, Count));
     }
     bool isDummyRoot() { return Address == 0; }
+    bool isLeafFrame() { return Children.empty(); }
   };
 
   ProfiledFrame DummyTrieRoot;
   ProfiledFrame *CurrentLeafFrame;
   // Used to fall through the LBR stack
   uint32_t LBRIndex = 0;
   // Reference to HybridSample.LBRStack
   const SmallVector<LBREntry, 16> &LBRStack;
struct ProbeStack {
   const ProfiledBinary *Binary;
   ProbeStack(const ProfiledBinary *B) : Binary(B) {}
   bool pushFrame(UnwindState::ProfiledFrame *Cur) {
     const PseudoProbe *CallProbe = Binary->getCallProbeForAddr(Cur->Address);
     // We may not find a probe for a merged or external callsite.
     // Callsite merging may cause the loss of original probe IDs.
     // Cutting off the context from here since the inliner will
     // not know how to consume a context with unknown callsites.
-    if (!CallProbe)
+    if (!CallProbe) {
+      if (!Cur->isLeafFrame())
+        WithColor::warning()
+            << "Untracked frame at " << format("%" PRIx64, Cur->Address)
+            << " due to missing call probe\n";
       return false;
+    }
     Stack.push_back(CallProbe);
     return true;
   }
 
   void popFrame() {
     if (!Stack.empty())
       Stack.pop_back();
   }
private:
   void unwindSamples();
   void printUnwinderOutput();
   // Helper function for looking up binary in AddressBinaryMap
   ProfiledBinary *getBinary(uint64_t Address);
 
   BinaryMap BinaryTable;
   AddressBinaryMap AddrToBinaryMap; // Used by address-based lookup.
 
-private:
   BinarySampleCounterMap BinarySampleCounters;
   // Samples with the repeating time generated by the perf reader
   AggregatedCounter AggregatedSamples;
   PerfScriptType PerfType = PERF_UNKNOWN;
 };
 
 } // end namespace sampleprof
 } // end namespace llvm
 
 #endif

llvm/tools/llvm-profgen/PerfReader.cpp

 }
 
 template <typename T>
 void VirtualUnwinder::collectSamplesFromFrameTrie(
     UnwindState::ProfiledFrame *Cur, T &Stack) {
   if (!Cur->isDummyRoot()) {
     if (!Stack.pushFrame(Cur)) {
       // Process truncated context
-      for (const auto &Item : Cur->Children) {
-        // Start a new traversal ignoring its bottom context
-        collectSamplesFromFrameTrie(Item.second.get());
+      T EmptyStack(Binary);
+      collectSamplesFromFrame(Cur, EmptyStack);
+      for (const auto &Item : Cur->Children) {
+        collectSamplesFromFrameTrie(Item.second.get(), EmptyStack);
       }
       return;
     }
   }
 
   collectSamplesFromFrame(Cur, Stack);
   // Process children frame
   for (const auto &Item : Cur->Children) {