only use flattened profile for O0 mode

Author: wlei-llvm

llvm/include/llvm/Transforms/IPO/SampleProfile.h

@@ -42,7 +42,8 @@ class SampleProfileLoaderPass : public PassInfoMixin<SampleProfileLoaderPass> {
       std::string File = "", std::string RemappingFile = "",
       ThinOrFullLTOPhase LTOPhase = ThinOrFullLTOPhase::None,
       IntrusiveRefCntPtr<vfs::FileSystem> FS = nullptr,
-      bool DisableSampleProfileInlining = false);
+      bool DisableSampleProfileInlining = false,
+      bool UseFlattenedProfile = false);
 
   PreservedAnalyses run(Module &M, ModuleAnalysisManager &AM);
 
@@ -52,6 +53,7 @@ class SampleProfileLoaderPass : public PassInfoMixin<SampleProfileLoaderPass> {
   const ThinOrFullLTOPhase LTOPhase;
   IntrusiveRefCntPtr<vfs::FileSystem> FS;
   bool DisableSampleProfileInlining;
+  bool UseFlattenedProfile;
 };
 
 } // end namespace llvm

llvm/lib/Passes/PassBuilderPipelines.cpp

@@ -2154,11 +2154,13 @@ PassBuilder::buildO0DefaultPipeline(OptimizationLevel Level,
     MPM.addPass(createModuleToFunctionPassAdaptor(AddDiscriminatorsPass()));
 
   if (PGOOpt && PGOOpt->Action == PGOOptions::SampleUse) {
-    // Explicitly disable sample loader inlining in O0 pipeline.
+    // Explicitly disable sample loader inlining and use flattened profile in O0
+    // pipeline.
     MPM.addPass(SampleProfileLoaderPass(PGOOpt->ProfileFile,
                                         PGOOpt->ProfileRemappingFile,
                                         ThinOrFullLTOPhase::None, nullptr,
-                                        /*DisableSampleProfileInlining=*/true));
+                                        /*DisableSampleProfileInlining=*/true,
+                                        /*UseFlattenedProfile=*/true));
     // Cache ProfileSummaryAnalysis once to avoid the potential need to insert
     // RequireAnalysisPass for PSI before subsequent non-module passes.
     MPM.addPass(RequireAnalysisPass<ProfileSummaryAnalysis, Module>());

llvm/lib/Transforms/IPO/SampleProfile.cpp

@@ -195,7 +195,8 @@ static cl::opt<bool> ProfileSizeInline(
 static cl::opt<bool> DisableSampleLoaderInlining(
     "disable-sample-loader-inlining", cl::Hidden, cl::init(false),
     cl::desc("If true, artifically skip inline transformation in sample-loader "
-             "pass, and use flattened profiles to emit annotation."));
+             "pass, and merge (or scale) profiles (as configured by "
+             "--sample-profile-merge-inlinee)."));
 
 namespace llvm {
 cl::opt<bool>
@@ -469,7 +470,8 @@ class SampleProfileLoader final : public SampleProfileLoaderBaseImpl<Function> {
       std::function<AssumptionCache &(Function &)> GetAssumptionCache,
       std::function<TargetTransformInfo &(Function &)> GetTargetTransformInfo,
       std::function<const TargetLibraryInfo &(Function &)> GetTLI,
-      LazyCallGraph &CG, bool DisableSampleProfileInlining)
+      LazyCallGraph &CG, bool DisableSampleProfileInlining,
+      bool UseFlattenedProfile)
       : SampleProfileLoaderBaseImpl(std::string(Name), std::string(RemapName),
                                     std::move(FS)),
         GetAC(std::move(GetAssumptionCache)),
@@ -479,7 +481,8 @@ class SampleProfileLoader final : public SampleProfileLoaderBaseImpl<Function> {
                               ? llvm::AnnotateInlinePassName(InlineContext{
                                     LTOPhase, InlinePass::SampleProfileInliner})
                               : CSINLINE_DEBUG),
-        DisableSampleProfileInlining(DisableSampleProfileInlining) {}
+        DisableSampleProfileInlining(DisableSampleProfileInlining),
+        UseFlattenedProfile(UseFlattenedProfile) {}
 
   bool doInitialization(Module &M, FunctionAnalysisManager *FAM = nullptr);
   bool runOnModule(Module &M, ModuleAnalysisManager *AM,
@@ -595,6 +598,8 @@ class SampleProfileLoader final : public SampleProfileLoaderBaseImpl<Function> {
 
   bool DisableSampleProfileInlining;
 
+  bool UseFlattenedProfile;
+
   // External inline advisor used to replay inline decision from remarks.
   std::unique_ptr<InlineAdvisor> ExternalInlineAdvisor;
 
@@ -1980,9 +1985,9 @@ bool SampleProfileLoader::doInitialization(Module &M,
   if (DisableSampleLoaderInlining.getNumOccurrences())
     DisableSampleProfileInlining = DisableSampleLoaderInlining;
 
-  // Use flattened profile if inlining is disabled.
-  if (DisableSampleProfileInlining && !Reader->profileIsCS())
-    ProfileConverter::flattenProfile(Reader->getProfiles());
+  if (UseFlattenedProfile)
+    ProfileConverter::flattenProfile(Reader->getProfiles(),
+                                     Reader->profileIsCS());
 
   // While profile-sample-accurate is on, ignore symbol list.
   ProfAccForSymsInList =
@@ -2314,10 +2319,12 @@ bool SampleProfileLoader::runOnFunction(Function &F, ModuleAnalysisManager *AM)
 }
 SampleProfileLoaderPass::SampleProfileLoaderPass(
     std::string File, std::string RemappingFile, ThinOrFullLTOPhase LTOPhase,
-    IntrusiveRefCntPtr<vfs::FileSystem> FS, bool DisableSampleProfileInlining)
+    IntrusiveRefCntPtr<vfs::FileSystem> FS, bool DisableSampleProfileInlining,
+    bool UseFlattenedProfile)
     : ProfileFileName(File), ProfileRemappingFileName(RemappingFile),
       LTOPhase(LTOPhase), FS(std::move(FS)),
-      DisableSampleProfileInlining(DisableSampleProfileInlining) {}
+      DisableSampleProfileInlining(DisableSampleProfileInlining),
+      UseFlattenedProfile(UseFlattenedProfile) {}
 
 PreservedAnalyses SampleProfileLoaderPass::run(Module &M,
                                                ModuleAnalysisManager &AM) {
@@ -2343,7 +2350,7 @@ PreservedAnalyses SampleProfileLoaderPass::run(Module &M,
       ProfileRemappingFileName.empty() ? SampleProfileRemappingFile
                                        : ProfileRemappingFileName,
       LTOPhase, FS, GetAssumptionCache, GetTTI, GetTLI, CG,
-      DisableSampleProfileInlining);
+      DisableSampleProfileInlining, UseFlattenedProfile);
   if (!SampleLoader.doInitialization(M, &FAM))
     return PreservedAnalyses::all();
 

llvm/test/Transforms/SampleProfile/inline-mergeprof.ll

@@ -9,9 +9,8 @@
 ; when the profile uses md5.
 ; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/inline-mergeprof.md5.prof -sample-profile-merge-inlinee=true -use-profiled-call-graph=0 -S | FileCheck -check-prefix=MERGE  %s
 
-; Test we properly use flattened profile when using '--disable-sample-loader-inlining'.
-
-; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/inline-mergeprof.prof -sample-profile-merge-inlinee=true --disable-sample-loader-inlining -use-profiled-call-graph=0 -S | FileCheck -check-prefix=FLATTEN  %s
+; Test we properly merge not inlined profile with '--sample-profile-merge-inlinee' even if '--disable-sample-loader-inlining' is true
+; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/inline-mergeprof.md5.prof -sample-profile-merge-inlinee=true --disable-sample-loader-inlining -use-profiled-call-graph=0 -S | FileCheck -check-prefix=MERGE  %s
 
 @.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
 
@@ -104,10 +103,3 @@ declare i32 @printf(ptr, ...)
 ; MERGE: !{!"branch_weights", i32 10}
 ; MERGE: name: "sub"
 ; MERGE-NEXT: {!"function_entry_count", i64 3}
-
-; FLATTEN: name: "sum"
-; FLATTEN-NEXT: {!"function_entry_count", i64 35}
-; FLATTEN: !{!"branch_weights", i32 13, i32 23}
-; FLATTEN: !{!"branch_weights", i32 12}
-; FLATTEN: name: "sub"
-; FLATTEN-NEXT: {!"function_entry_count", i64 3}