[SampleFDO] minimize performance impact when profile-sample-accurate is enabled

[wmi]

Summary

We can save memory and reduce binary size significantly by enabling ProfileSampleAccurate. However when ProfileSampleAccurate is true, function without sample will be regarded as cold and this could potentially cause performance regression.

To minimize the potential negative performance impact, we want to be a little conservative here saying if a function shows up in the profile, no matter as outline function, inline instance or call targets, treat the function as not being cold. This will handle the cases such as most callsites of a function are inlined in sampled binary (thus outline copy don't get any sample) but not inlined in current build (because of source code drift, imprecise debug information, or the callsites are all cold individually but not cold accumulatively...), so that the outline function showing up as cold in sampled binary will actually not be cold after current build. After the change, such function will be treated as not cold even profile-sample-accurate is enabled.

At the same time we lower the hot criteria of callsiteIsHot check when profile-sample-accurate is enabled. callsiteIsHot is used to determined whether a callsite is hot and qualified for early inlining. When profile-sample-accurate is enabled, functions without profile will be regarded as cold and much less inlining will happen in CGSCC inlining pass, so we can worry less about size increase and be aggressive to allow more early inlining to happen for warm callsites and it is helpful for performance overall.

[davidxl]

If we use NameTable, is ProfileSymbolList still needed? I assume NameTable is a superset ?

lib/Transforms/IPO/SampleProfile.cpp

1794

This is a linear search. What is the compile time impact?

[wmi]

In D67561#1670630, @davidxl wrote:

If we use NameTable, is ProfileSymbolList still needed? I assume NameTable is a superset ?

NameTable is not a superset. ProfileSymbolList contains all the function symbols in the sampled binary. NameTable contains all the symbols in the profile including outline copy, inline instance and call targets. The size of NameTable is usually < 1/10 of the size of ProfileSymbolList -- a lot of function symbols are cold and won't show up in the profile.

lib/Transforms/IPO/SampleProfile.cpp

1794

I will evaluate it and get back.

[wmi]

Create a StringSet for NameTable to avoid linear scanning of the table for each function.

[wmi]

lib/Transforms/IPO/SampleProfile.cpp

1794

I did the evaluation. I didn't see any compile time increase when compiling an internal server benchmark, but I saw 1~2% compile time increase when bootstrapping clang. I am not very sure about the increase. Although I run comparison twice and saw the increase in both times, the time to bootstrap clang on my desktop is not very stable so it could be noise.

Anyway, I think it is better to avoid the linear scan for every function, so I change NameTable to be a StringSet.

[davidxl]

include/llvm/ProfileData/SampleProfReader.h

334

Make it clearer that it includes all the names that have samples either in outline instance or an inline instance. It is a subset of the symbol list of the binary.

lib/Transforms/IPO/SampleProfile.cpp

481

Make this a member function to avoid passing the boolean parameter around.

1616

what are these two flags?

1792

If this is no symbol list, then it should not use this heuristic to be consistent with old behavior?

[wmi]

include/llvm/ProfileData/SampleProfReader.h

334

Will fix.

lib/Transforms/IPO/SampleProfile.cpp

481

Will fix.

1616

They are used to emit a warning when certain ratio of profile records or samples are not used during annotation. callsiteIsHot is used when computing the total samples (Only samples in hot callsite will be considered in total).

1792

I think it is beneficial to have the change when profile-sample-accurate is true no matter symbol list is there or not. symbol list is used to prevent new hot functions from being treated as cold. NamesInProfile is used to prevent functions are mistakenly treated as cold because inlining difference between sampled binary and the IR after early inlining in current build. These two changes are orthogonal to each other and are both beneficial individually.

[wmi]

include/llvm/ProfileData/SampleProfReader.h

334

Actually it is not a subset of symbol list of the binary. It is possible that a function is inlined everywhere and the outline instance gets deleted so it won't show up in the symbol list of the binary, but it will show up in the profile.

[wmi]

Address David's comment.

[davidxl]

lgtm

[davidxl]

lib/Transforms/IPO/SampleProfile.cpp

1792

Keeping the old behavior without Symbol list may be able to fix Chrome's size regression.

[wmi]

lib/Transforms/IPO/SampleProfile.cpp

1792

I am evaluating whether we need to do it for both changes: using NamesInProfile and adjusting threshold in callsiteIsHot. I feel maybe not using NamesInProfile without symbol list is enough to remove the size regression. I will evaluate and have a patch for it.

[wmi]

lib/Transforms/IPO/SampleProfile.cpp

1792

For my experiment, adjusting threshold in callsiteIsHot only increases size by 0.09% so most of the size increase comes from treating names in NamesInProfile as not cold.

include/llvm/ProfileData/SampleProfReader.h

public:
 
   /// \brief Return the profile format.
   SampleProfileFormat getFormat() { return Format; }
 
   virtual std::unique_ptr<ProfileSymbolList> getProfileSymbolList() {
     return nullptr;
   };
 
+  /// Get the pointer of name table which includes all the names used in
+  /// FunctionSamples.
+  virtual std::vector<StringRef> *getNameTable() { return nullptr; }
+
 protected:
   /// Map every function to its associated profile.
   ///
   /// The profile of every function executed at runtime is collected
   /// in the structure FunctionSamples. This maps function objects
   /// to their corresponding profiles.
   StringMap<FunctionSamples> Profiles;
 
SampleProfileReaderBinary(std::unique_ptr<MemoryBuffer> B, LLVMContext &C,
       : SampleProfileReader(std::move(B), C, Format) {}
 
   /// Read and validate the file header.
   virtual std::error_code readHeader() override;
 
   /// Read sample profiles from the associated file.
   std::error_code read() override;
 
+  /// Get name table which includes all the names used in FunctionSamples.
+  virtual std::vector<StringRef> *getNameTable() override { return &NameTable; }
+
 protected:
   /// Read a numeric value of type T from the profile.
   ///
   /// If an error occurs during decoding, a diagnostic message is emitted and
   /// EC is set.
   ///
   /// \returns the read value.
   template <typename T> ErrorOr<T> readNumber();

lib/Transforms/IPO/SampleProfile.cpp

public:
   unsigned countUsedRecords(const FunctionSamples *FS,
                             ProfileSummaryInfo *PSI) const;
   unsigned countBodyRecords(const FunctionSamples *FS,
                             ProfileSummaryInfo *PSI) const;
   uint64_t getTotalUsedSamples() const { return TotalUsedSamples; }
   uint64_t countBodySamples(const FunctionSamples *FS,
                             ProfileSummaryInfo *PSI) const;
 
+  void setProfSampleAccurateEnabled() { ProfSampleAccEnabled = true; }
+
+  bool isProfSampleAccurateEnabled() const { return ProfSampleAccEnabled; }
+
   void clear() {
     SampleCoverage.clear();
     TotalUsedSamples = 0;
   }
 
 private:
   using BodySampleCoverageMap = std::map<LineLocation, unsigned>;
   using FunctionSamplesCoverageMap =
private:
   /// based on the total number of samples available in this function and
   /// its callsites.
   ///
   /// Note that this accumulator tracks samples used from a single function
   /// and all the inlined callsites. Strictly, we should have a map of counters
   /// keyed by FunctionSamples pointers, but these stats are cleared after
   /// every function, so we just need to keep a single counter.
   uint64_t TotalUsedSamples = 0;
+
+  // Showing whether ProfileSampleAccurate is enabled for current function.
+  bool ProfSampleAccEnabled = false;
 };
 
 class GUIDToFuncNameMapper {
 public:
   GUIDToFuncNameMapper(Module &M, SampleProfileReader &Reader,
                         DenseMap<uint64_t, StringRef> &GUIDToFuncNameMap)
       : CurrentReader(Reader), CurrentModule(M),
       CurrentGUIDToFuncNameMap(GUIDToFuncNameMap) {
protected:
   struct NotInlinedProfileInfo {
     uint64_t entryCount;
   };
   DenseMap<Function *, NotInlinedProfileInfo> notInlinedCallInfo;
 
   // GUIDToFuncNameMap saves the mapping from GUID to the symbol name, for
   // all the function symbols defined or declared in current module.
   DenseMap<uint64_t, StringRef> GUIDToFuncNameMap;
+
+  // All the Names used in FunctionSamples including outline function
+  // names, inline instance names and call target names.
+  StringSet<> NamesInProfile;
+
+  // Showing whether ProfileSampleAccurate is enabled for current function.
+  bool ProfSampleAccEnabled = false;
 };
 
 class SampleProfileLoaderLegacyPass : public ModulePass {
 public:
   // Class identification, replacement for typeinfo
   static char ID;
 
   SampleProfileLoaderLegacyPass(StringRef Name = SampleProfileFile,

 /// in the inline stack in the sample profile. If the profile shows that
 /// the original inline decision was "good" (i.e., the callsite is executed
 /// frequently), then we will recreate the inline decision and apply the
 /// profile from the inlined callsite.
 ///
 /// To decide whether an inlined callsite is hot, we compare the callsite
 /// sample count with the hot cutoff computed by ProfileSummaryInfo, it is
 /// regarded as hot if the count is above the cutoff value.
+///
+/// When profile-sample-accurate is enabled, functions without profile will
+/// be regarded as cold and much less inlining will happen in CGSCC inlining
+/// pass, so we tend to lower the hot criteria here to allow more early
+/// inlining to happen for warm callsites and it is helpful for performance.
 static bool callsiteIsHot(const FunctionSamples *CallsiteFS,
-                          ProfileSummaryInfo *PSI) {
+                          ProfileSummaryInfo *PSI, bool ProfSampleAccEnabled) {
   if (!CallsiteFS)
     return false; // The callsite was not inlined in the original binary.
 
   assert(PSI && "PSI is expected to be non null");
   uint64_t CallsiteTotalSamples = CallsiteFS->getTotalSamples();
+  if (ProfSampleAccEnabled)
+    return !PSI->isColdCount(CallsiteTotalSamples);
+  else
     return PSI->isHotCount(CallsiteTotalSamples);
 }
 
 /// Mark as used the sample record for the given function samples at
 /// (LineOffset, Discriminator).
 ///
 /// \returns true if this is the first time we mark the given record.
 bool SampleCoverageTracker::markSamplesUsed(const FunctionSamples *FS,
                                             uint32_t LineOffset,
SampleCoverageTracker::countUsedRecords(const FunctionSamples *FS,
   unsigned Count = (I != SampleCoverage.end()) ? I->second.size() : 0;
 
   // If there are inlined callsites in this function, count the samples found
   // in the respective bodies. However, do not bother counting callees with 0
   // total samples, these are callees that were never invoked at runtime.
   for (const auto &I : FS->getCallsiteSamples())
     for (const auto &J : I.second) {
       const FunctionSamples *CalleeSamples = &J.second;
-      if (callsiteIsHot(CalleeSamples, PSI))
+      if (callsiteIsHot(CalleeSamples, PSI, ProfSampleAccEnabled))
         Count += countUsedRecords(CalleeSamples, PSI);
     }
 
   return Count;
 }
 
 /// Return the number of sample records in the body of this profile.
 ///
 /// This count does not include records from cold inlined callsites.
 unsigned
 SampleCoverageTracker::countBodyRecords(const FunctionSamples *FS,
                                         ProfileSummaryInfo *PSI) const {
   unsigned Count = FS->getBodySamples().size();
 
   // Only count records in hot callsites.
   for (const auto &I : FS->getCallsiteSamples())
     for (const auto &J : I.second) {
       const FunctionSamples *CalleeSamples = &J.second;
-      if (callsiteIsHot(CalleeSamples, PSI))
+      if (callsiteIsHot(CalleeSamples, PSI, ProfSampleAccEnabled))
         Count += countBodyRecords(CalleeSamples, PSI);
     }
 
   return Count;
 }
 
 /// Return the number of samples collected in the body of this profile.
 ///
 /// This count does not include samples from cold inlined callsites.
 uint64_t
 SampleCoverageTracker::countBodySamples(const FunctionSamples *FS,
                                         ProfileSummaryInfo *PSI) const {
   uint64_t Total = 0;
   for (const auto &I : FS->getBodySamples())
     Total += I.second.getSamples();
 
   // Only count samples in hot callsites.
   for (const auto &I : FS->getCallsiteSamples())
     for (const auto &J : I.second) {
       const FunctionSamples *CalleeSamples = &J.second;
-      if (callsiteIsHot(CalleeSamples, PSI))
+      if (callsiteIsHot(CalleeSamples, PSI, ProfSampleAccEnabled))
         Total += countBodySamples(CalleeSamples, PSI);
     }
 
   return Total;
 }
 
 /// Return the fraction of sample records used in this profile.
 ///
for (auto &BB : F) {
       SmallVector<Instruction *, 10> Candidates;
       for (auto &I : BB.getInstList()) {
         const FunctionSamples *FS = nullptr;
         if ((isa<CallInst>(I) || isa<InvokeInst>(I)) &&
             !isa<IntrinsicInst>(I) && (FS = findCalleeFunctionSamples(I))) {
           Candidates.push_back(&I);
           if (FS->getEntrySamples() > 0)
             localNotInlinedCallSites.try_emplace(&I, FS);
-          if (callsiteIsHot(FS, PSI))
+          if (callsiteIsHot(FS, PSI, ProfSampleAccEnabled))
             Hot = true;
         }
       }
       if (Hot) {
         CIS.insert(CIS.begin(), Candidates.begin(), Candidates.end());
       }
     }
     for (auto I : CIS) {
for (auto I : CIS) {
           // If it is a recursive call, we do not inline it as it could bloat
           // the code exponentially. There is way to better handle this, e.g.
           // clone the caller first, and inline the cloned caller if it is
           // recursive. As llvm does not inline recursive calls, we will
           // simply ignore it instead of handling it explicitly.
           if (CalleeFunctionName == F.getName())
             continue;
 
-          if (!callsiteIsHot(FS, PSI))
+          if (!callsiteIsHot(FS, PSI, ProfSampleAccEnabled))
             continue;
 
           const char *Reason = "Callee function not available";
           auto R = SymbolMap.find(CalleeFunctionName);
           if (R != SymbolMap.end() && R->getValue() &&
               !R->getValue()->isDeclaration() &&
               R->getValue()->getSubprogram() &&
               isLegalToPromote(CallSite(I), R->getValue(), &Reason)) {
if (Changed) {
 
     // Find equivalence classes.
     findEquivalenceClasses(F);
 
     // Propagate weights to all edges.
     propagateWeights(F);
   }
 
+  if (SampleProfileRecordCoverage || SampleProfileSampleCoverage)
+    CoverageTracker.setProfSampleAccurateEnabled();
+
   // If coverage checking was requested, compute it now.
   if (SampleProfileRecordCoverage) {
     unsigned Used = CoverageTracker.countUsedRecords(Samples, PSI);
     unsigned Total = CoverageTracker.countBodyRecords(Samples, PSI);
     unsigned Coverage = CoverageTracker.computeCoverage(Used, Total);
     if (Coverage < SampleProfileRecordCoverage) {
       F.getContext().diagnose(DiagnosticInfoSampleProfile(
           F.getSubprogram()->getFilename(), getFunctionLoc(F),
if (std::error_code EC = ReaderOrErr.getError()) {
     Ctx.diagnose(DiagnosticInfoSampleProfile(Filename, Msg));
     return false;
   }
   Reader = std::move(ReaderOrErr.get());
   Reader->collectFuncsToUse(M);
   ProfileIsValid = (Reader->read() == sampleprof_error::success);
   PSL = Reader->getProfileSymbolList();
 
+  if (ProfileSampleAccurate) {
+    NamesInProfile.clear();
+    if (auto NameTable = Reader->getNameTable())
+      NamesInProfile.insert(NameTable->begin(), NameTable->end());
+  }
+
   if (!RemappingFilename.empty()) {
     // Apply profile remappings to the loaded profile data if requested.
     // For now, we only support remapping symbols encoded using the Itanium
     // C++ ABI's name mangling scheme.
     ReaderOrErr = SampleProfileReaderItaniumRemapper::create(
         RemappingFilename, Ctx, std::move(Reader));
     if (std::error_code EC = ReaderOrErr.getError()) {
       std::string Msg = "Could not open profile remapping file: " + EC.message();

 
 bool SampleProfileLoader::runOnFunction(Function &F, ModuleAnalysisManager *AM) {
 
   DILocation2SampleMap.clear();
   // By default the entry count is initialized to -1, which will be treated
   // conservatively by getEntryCount as the same as unknown (None). This is
   // to avoid newly added code to be treated as cold. If we have samples
   // this will be overwritten in emitAnnotations.
-  //
+  uint64_t initialEntryCount = -1;
+
+  ProfSampleAccEnabled =
+      ProfileSampleAccurate || F.hasFnAttribute("profile-sample-accurate");
+  if (ProfSampleAccEnabled) {
     // PSL -- profile symbol list include all the symbols in sampled binary.
+    // It is used to prevent new functions to be treated as cold.
     // If ProfileSampleAccurate is true or F has profile-sample-accurate
     // attribute, and if there is no profile symbol list read in, initialize
     // all the function entry counts to 0; if there is profile symbol list, only
     // initialize the entry count to 0 when current function is in the list.
-  uint64_t initialEntryCount =
-      ((ProfileSampleAccurate || F.hasFnAttribute("profile-sample-accurate")) &&
-       (!PSL || PSL->contains(F.getName())))
-          ? 0
-          : -1;
+    if (!PSL || PSL->contains(F.getName()))
+      initialEntryCount = 0;
+
+    // When ProfileSampleAccurate is true, function without sample will be
+    // regarded as cold. To minimize the potential negative performance
+    // impact it could have, we want to be a little conservative here
+    // saying if a function shows up in the profile, no matter as outline
+    // function, inline instance or call targets, treat the function as not
+    // being cold. This will handle the cases such as most callsites of a
+    // function are inlined in sampled binary but not inlined in current
+    // build (because of source code drift, imprecise debug information, or
+    // the callsites are all cold individually but not cold accumulatively...),
+    // so the outline function showing up as cold in sampled binary will
+    // actually not be cold after current build.
+    StringRef CanonName = FunctionSamples::getCanonicalFnName(F);
+    if (NamesInProfile.count(CanonName))
+      initialEntryCount = -1;
+  }
+
   F.setEntryCount(ProfileCount(initialEntryCount, Function::PCT_Real));
   std::unique_ptr<OptimizationRemarkEmitter> OwnedORE;
   if (AM) {
     auto &FAM =
         AM->getResult<FunctionAnalysisManagerModuleProxy>(*F.getParent())
             .getManager();
     ORE = &FAM.getResult<OptimizationRemarkEmitterAnalysis>(F);
   } else {

test/Transforms/SampleProfile/profile-sample-accurate.ll

+; RUN: opt < %s -sample-profile -sample-profile-file=%S/Inputs/profsampleacc.extbinary.afdo -profile-summary-cutoff-hot=600000 -profile-sample-accurate -S | FileCheck %s
+; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/profsampleacc.extbinary.afdo -profile-summary-cutoff-hot=600000 -profile-sample-accurate -S | FileCheck %s
+; With the hot cutoff being set to 600000, the inline instance of _Z3sumii
+; in main is neither hot nor cold. Check it will still be inlined when
+; profile-sample-accurate is enabled, and check _Z3sumii's function entry
+; count won't be initialized to 0 because it shows up in the profile as
+; inline instance.
+
+; Original C++ test case
+;
+; #include <stdio.h>
+;
+; int sum(int x, int y) {
+;   return x + y;
+; }
+;
+; int main() {
+;   int s, i = 0;
+;   while (i++ < 20000 * 20000)
+;     if (i != 100) s = sum(i, s); else s = 30;
+;   printf("sum is %d\n", s);
+;   return 0;
+; }
+;
+@.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
+
+; CHECK: define i32 @_Z3sumii{{.*}}!prof ![[UNKNOWN_ID:[0-9]+]]
+; Function Attrs: nounwind uwtable
+define i32 @_Z3sumii(i32 %x, i32 %y) !dbg !4 {
+entry:
+  %x.addr = alloca i32, align 4
+  %y.addr = alloca i32, align 4
+  store i32 %x, i32* %x.addr, align 4
+  store i32 %y, i32* %y.addr, align 4
+  %0 = load i32, i32* %x.addr, align 4, !dbg !11
+  %1 = load i32, i32* %y.addr, align 4, !dbg !11
+  %add = add nsw i32 %0, %1, !dbg !11
+  ret i32 %add, !dbg !11
+}
+
+; Function Attrs: uwtable
+define i32 @main() !dbg !7 {
+entry:
+  %retval = alloca i32, align 4
+  %s = alloca i32, align 4
+  %i = alloca i32, align 4
+  store i32 0, i32* %retval
+  store i32 0, i32* %i, align 4, !dbg !12
+  br label %while.cond, !dbg !13
+
+while.cond:                                       ; preds = %if.end, %entry
+  %0 = load i32, i32* %i, align 4, !dbg !14
+  %inc = add nsw i32 %0, 1, !dbg !14
+  store i32 %inc, i32* %i, align 4, !dbg !14
+  %cmp = icmp slt i32 %0, 400000000, !dbg !14
+  br i1 %cmp, label %while.body, label %while.end, !dbg !14
+
+while.body:                                       ; preds = %while.cond
+  %1 = load i32, i32* %i, align 4, !dbg !16
+  %cmp1 = icmp ne i32 %1, 100, !dbg !16
+  br i1 %cmp1, label %if.then, label %if.else, !dbg !16
+
+; Check _Z3sumii is inlined at this callsite.
+; CHECK: if.then:
+; CHECK-NOT: call i32 @_Z3sumii
+; CHECK: if.else:
+if.then:                                          ; preds = %while.body
+  %2 = load i32, i32* %i, align 4, !dbg !18
+  %3 = load i32, i32* %s, align 4, !dbg !18
+  %call = call i32 @_Z3sumii(i32 %2, i32 %3), !dbg !18
+  store i32 %call, i32* %s, align 4, !dbg !18
+  br label %if.end, !dbg !18
+
+if.else:                                          ; preds = %while.body
+  store i32 30, i32* %s, align 4, !dbg !20
+  br label %if.end
+
+if.end:                                           ; preds = %if.else, %if.then
+  br label %while.cond, !dbg !22
+
+while.end:                                        ; preds = %while.cond
+  %4 = load i32, i32* %s, align 4, !dbg !24
+  %call2 = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i32 0, i32 0), i32 %4), !dbg !24
+  ret i32 0, !dbg !25
+}
+
+declare i32 @printf(i8*, ...) #2
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!8, !9}
+!llvm.ident = !{!10}
+
+; CHECK: ![[UNKNOWN_ID]] = !{!"function_entry_count", i64 -1}
+!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, producer: "clang version 3.5 ", isOptimized: false, emissionKind: NoDebug, file: !1, enums: !2, retainedTypes: !2, globals: !2, imports: !2)
+!1 = !DIFile(filename: "calls.cc", directory: ".")
+!2 = !{}
+!4 = distinct !DISubprogram(name: "sum", line: 3, isLocal: false, isDefinition: true, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: false, unit: !0, scopeLine: 3, file: !1, scope: !5, type: !6, retainedNodes: !2)
+!5 = !DIFile(filename: "calls.cc", directory: ".")
+!6 = !DISubroutineType(types: !2)
+!7 = distinct !DISubprogram(name: "main", line: 7, isLocal: false, isDefinition: true, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: false, unit: !0, scopeLine: 7, file: !1, scope: !5, type: !6, retainedNodes: !2)
+!8 = !{i32 2, !"Dwarf Version", i32 4}
+!9 = !{i32 1, !"Debug Info Version", i32 3}
+!10 = !{!"clang version 3.5 "}
+!11 = !DILocation(line: 4, scope: !4)
+!12 = !DILocation(line: 8, scope: !7)
+!13 = !DILocation(line: 9, scope: !7)
+!14 = !DILocation(line: 9, scope: !15)
+!15 = !DILexicalBlockFile(discriminator: 2, file: !1, scope: !7)
+!16 = !DILocation(line: 10, scope: !17)
+!17 = distinct !DILexicalBlock(line: 10, column: 0, file: !1, scope: !7)
+!18 = !DILocation(line: 10, scope: !19)
+!19 = !DILexicalBlockFile(discriminator: 2, file: !1, scope: !17)
+!20 = !DILocation(line: 10, scope: !21)
+!21 = !DILexicalBlockFile(discriminator: 4, file: !1, scope: !17)
+!22 = !DILocation(line: 10, scope: !23)
+!23 = !DILexicalBlockFile(discriminator: 6, file: !1, scope: !17)
+!24 = !DILocation(line: 11, scope: !7)
+!25 = !DILocation(line: 12, scope: !7)