diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td index 002f60350543d9..0ac289089839d2 100644 --- a/clang/include/clang/Driver/Options.td +++ b/clang/include/clang/Driver/Options.td @@ -1784,6 +1784,12 @@ defm debug_info_for_profiling : BoolFOption<"debug-info-for-profiling", PosFlag, NegFlag>; +def fprofile_sample_cold_function : Flag<["-"], "fprofile-sample-cold-function">, + Group, Visibility<[ClangOption, CLOption]>, + HelpText<"Generate instrumented code to cold functions guided by sampling-based profile into default.profraw file (overridden by '=' form of option or LLVM_PROFILE_FILE env var)">; +def fprofile_sample_cold_function_EQ : Joined<["-"], "fprofile-sample-cold-function=">, + Group, Visibility<[ClangOption, CLOption]>, MetaVarName<"">, + HelpText<"Generate instrumented code to cold functions guided by sampling-based profile into (overridden by LLVM_PROFILE_FILE env var)">; def fprofile_instr_generate : Flag<["-"], "fprofile-instr-generate">, Group, Visibility<[ClangOption, CLOption]>, HelpText<"Generate instrumented code to collect execution counts into default.profraw file (overridden by '=' form of option or LLVM_PROFILE_FILE env var)">; diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp index 16f9b629fc538c..5cd821cfe780c4 100644 --- a/clang/lib/Driver/ToolChain.cpp +++ b/clang/lib/Driver/ToolChain.cpp @@ -889,7 +889,9 @@ bool ToolChain::needsProfileRT(const ArgList &Args) { Args.hasArg(options::OPT_fprofile_instr_generate) || Args.hasArg(options::OPT_fprofile_instr_generate_EQ) || Args.hasArg(options::OPT_fcreate_profile) || - Args.hasArg(options::OPT_forder_file_instrumentation); + Args.hasArg(options::OPT_forder_file_instrumentation) || + Args.hasArg(options::OPT_fprofile_sample_cold_function) || + Args.hasArg(options::OPT_fprofile_sample_cold_function_EQ); } bool ToolChain::needsGCovInstrumentation(const llvm::opt::ArgList &Args) { diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp index 0bab48caf1a5e2..6e128dfc888ab4 100644 --- a/clang/lib/Driver/ToolChains/Clang.cpp +++ b/clang/lib/Driver/ToolChains/Clang.cpp @@ -649,6 +649,18 @@ static void addPGOAndCoverageFlags(const ToolChain &TC, Compilation &C, } } + if (auto *SampleColdArg = + Args.getLastArg(options::OPT_fprofile_sample_cold_function, + options::OPT_fprofile_sample_cold_function_EQ)) { + SmallString<128> Path(SampleColdArg->getOption().matches( + options::OPT_fprofile_sample_cold_function_EQ) + ? SampleColdArg->getValue() + : "default_%m.profraw"); + CmdArgs.push_back("-mllvm"); + CmdArgs.push_back(Args.MakeArgString( + Twine("--instrument-sample-cold-function-path=") + Path)); + } + Arg *PGOGenArg = nullptr; if (PGOGenerateArg) { assert(!CSPGOGenerateArg); diff --git a/llvm/lib/Passes/PassBuilderPipelines.cpp b/llvm/lib/Passes/PassBuilderPipelines.cpp index 8f151a99b11709..0514d17db20721 100644 --- a/llvm/lib/Passes/PassBuilderPipelines.cpp +++ b/llvm/lib/Passes/PassBuilderPipelines.cpp @@ -296,7 +296,13 @@ static cl::opt UseLoopVersioningLICM( "enable-loop-versioning-licm", cl::init(false), cl::Hidden, cl::desc("Enable the experimental Loop Versioning LICM pass")); +static cl::opt InstrumentSampleColdFuncPath( + "instrument-sample-cold-function-path", cl::init(""), + cl::desc("File path for instrumenting sampling PGO guided cold functions"), + cl::Hidden); + extern cl::opt UseCtxProfile; +extern cl::opt InstrumentColdFunction; namespace llvm { extern cl::opt EnableMemProfContextDisambiguation; @@ -1119,6 +1125,17 @@ PassBuilder::buildModuleSimplificationPipeline(OptimizationLevel Level, // removed. MPM.addPass( PGOIndirectCallPromotion(true /* IsInLTO */, true /* SamplePGO */)); + + if (InstrumentSampleColdFuncPath.getNumOccurrences() && + Phase != ThinOrFullLTOPhase::ThinLTOPostLink) { + assert(!InstrumentSampleColdFuncPath.empty() && + "File path is requeired for instrumentation generation"); + InstrumentColdFunction = true; + addPreInlinerPasses(MPM, Level, Phase); + addPGOInstrPasses(MPM, Level, /* RunProfileGen */ true, + /* IsCS */ false, /* AtomicCounterUpdate */ false, + InstrumentSampleColdFuncPath, "", PGOOpt->FS); + } } // Try to perform OpenMP specific optimizations on the module. This is a diff --git a/llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp b/llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp index 10442fa0bb9003..cbd14567d97c95 100644 --- a/llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp +++ b/llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp @@ -319,6 +319,16 @@ static cl::opt PGOFunctionCriticalEdgeThreshold( cl::desc("Do not instrument functions with the number of critical edges " " greater than this threshold.")); +cl::opt InstrumentColdFunction( + "instrument-cold-function", cl::init(false), cl::Hidden, + cl::desc("Instrument cold functions (currently only used under sampling " + " PGO pipeline))")); + +static cl::opt InstrumentColdFuncMaxEntryCount( + "instrument-cold-function-max-entry-count", cl::init(0), cl::Hidden, + cl::desc("When using --instrument-cold-function, skip instrumenting the " + "function whose entry count is above the given value")); + extern cl::opt MaxNumVTableAnnotations; namespace llvm { @@ -1891,6 +1901,10 @@ static bool skipPGOGen(const Function &F) { return true; if (F.getInstructionCount() < PGOFunctionSizeThreshold) return true; + if (InstrumentColdFunction && + (!F.getEntryCount() || + F.getEntryCount()->getCount() > InstrumentColdFuncMaxEntryCount)) + return true; return false; } diff --git a/llvm/test/Transforms/PGOProfile/instr-gen-cold-function.ll b/llvm/test/Transforms/PGOProfile/instr-gen-cold-function.ll new file mode 100644 index 00000000000000..14d3b3c7e9b6fa --- /dev/null +++ b/llvm/test/Transforms/PGOProfile/instr-gen-cold-function.ll @@ -0,0 +1,24 @@ +; RUN: opt < %s --passes=pgo-instr-gen -instrument-cold-function -S | FileCheck --check-prefixes=COLD %s +; RUN: opt < %s --passes=pgo-instr-gen -instrument-cold-function -instrument-cold-function-max-entry-count=1 -S | FileCheck --check-prefixes=ENTRY-COUNT %s + +; COLD: call void @llvm.instrprof.increment(ptr @__profn_foo, i64 [[#]], i32 1, i32 0) +; COLD-NOT: __profn_main + +; ENTRY-COUNT: call void @llvm.instrprof.increment(ptr @__profn_foo, i64 [[#]], i32 1, i32 0) +; ENTRY-COUNT: call void @llvm.instrprof.increment(ptr @__profn_main, i64 [[#]], i32 1, i32 0) + +target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128" +target triple = "x86_64-unknown-linux-gnu" + +define void @foo() !prof !0 { +entry: + ret void +} + +define i32 @main() !prof !1 { +entry: + ret i32 0 +} + +!0 = !{!"function_entry_count", i64 0} +!1 = !{!"function_entry_count", i64 1}