Index: include/llvm/Analysis/OptimizationDiagnosticInfo.h =================================================================== --- /dev/null +++ include/llvm/Analysis/OptimizationDiagnosticInfo.h @@ -0,0 +1,43 @@ +//===- OptimizationDiagnosticInfo.h - Optimization Diagnostic ---*- C++ -*-===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// Optimization diagnostic interfaces. Since these use BFI they have to reside +// in Analysis. +//===----------------------------------------------------------------------===// + +#ifndef LLVM_IR_OPTIMIZATIONDIAGNOSTICINFO_H +#define LLVM_IR_OPTIMIZATIONDIAGNOSTICINFO_H + +namespace llvm { +class DebugLoc; +class Function; +class LLVMContext; +class Pass; +class Twine; +class Value; + +/// Emit an optimization-missed message. +/// +/// \p FromPass is the pass emitting the message. If -Rpass-missed= is given and +/// the name of pass matches the regular expression in -Rpass, then the remark +/// will be emitted. \p Fn is the function triggering the remark, \p DLoc is the +/// debug location where the diagnostic is generated. \p V is the IR Value that +/// identifies the code region. \p Msg is the message string to use. +void emitOptimizationRemarkMissed(LLVMContext &Ctx, const Pass &FromPass, + const Function &Fn, const DebugLoc &DLoc, + Value *V, const Twine &Msg); + +/// \brief Same as above but derives the IR Value for the code region and the +/// debug location from the Loop parameter \p L. +void emitOptimizationRemarkMissed(LLVMContext &Ctx, const Pass &FromPass, + const Function &Fn, Loop *L, + const Twine &Msg); +} + +#endif // LLVM_IR_OPTIMIZATIONDIAGNOSTICINFO_H Index: include/llvm/IR/DiagnosticInfo.h =================================================================== --- include/llvm/IR/DiagnosticInfo.h +++ include/llvm/IR/DiagnosticInfo.h @@ -31,6 +31,7 @@ class Instruction; class LLVMContext; class Module; +class Pass; class SMDiagnostic; /// \brief Defines the different supported severity of a diagnostic. @@ -385,7 +386,13 @@ const char *PassName, const Function &Fn, const DebugLoc &DLoc, const Twine &Msg) : DiagnosticInfoWithDebugLocBase(Kind, Severity, Fn, DLoc), - PassName(PassName), Msg(Msg) {} + PassName(PassName), Msg(Msg), Hotness(0) {} + + DiagnosticInfoOptimizationBase(enum DiagnosticKind Kind, + enum DiagnosticSeverity Severity, + const Pass &FromPass, const Function &Fn, + const DebugLoc &DLoc, const Twine &Msg, + uint64_t Hotness); /// \see DiagnosticInfo::print. void print(DiagnosticPrinter &DP) const override; @@ -413,6 +420,10 @@ /// Message to report. const Twine &Msg; + + /// The number of times the corresponding code was executed in a profile + /// instrumentation run. If profile informatin is not avaiable, this zero. + uint64_t Hotness; }; /// Diagnostic information for applied optimization remarks. @@ -457,6 +468,13 @@ : DiagnosticInfoOptimizationBase(DK_OptimizationRemarkMissed, DS_Remark, PassName, Fn, DLoc, Msg) {} + DiagnosticInfoOptimizationRemarkMissed(const Pass &FromPass, + const Function &Fn, + const DebugLoc &DLoc, const Twine &Msg, + uint64_t Hotness) + : DiagnosticInfoOptimizationBase(DK_OptimizationRemarkMissed, DS_Remark, + FromPass, Fn, DLoc, Msg, Hotness) {} + static bool classof(const DiagnosticInfo *DI) { return DI->getKind() == DK_OptimizationRemarkMissed; } Index: lib/Analysis/CMakeLists.txt =================================================================== --- lib/Analysis/CMakeLists.txt +++ lib/Analysis/CMakeLists.txt @@ -53,6 +53,7 @@ ObjCARCAliasAnalysis.cpp ObjCARCAnalysisUtils.cpp ObjCARCInstKind.cpp + OptimizationDiagnosticInfo.cpp OrderedBasicBlock.cpp PHITransAddr.cpp PostDominators.cpp Index: lib/Analysis/OptimizationDiagnosticInfo.cpp =================================================================== --- /dev/null +++ lib/Analysis/OptimizationDiagnosticInfo.cpp @@ -0,0 +1,40 @@ +//===- OptimizationDiagnosticInfo.cpp - Optimization Diagnostic -*- C++ -*-===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// Optimization diagnostic interfaces. Since these use BFI they have to reside +// in Analysis. +//===----------------------------------------------------------------------===// + +#include "llvm/Analysis/OptimizationDiagnosticInfo.h" +#include "llvm/Analysis/BlockFrequencyInfo.h" +#include "llvm/IR/DiagnosticInfo.h" +#include "llvm/IR/LLVMContext.h" + +using namespace llvm; + +static uint64_t computeHotness(const Pass &FromPass, const Function &Fn, + Value *V) { + uint64_t Hotness = 0; + + auto *BFI = &FromPass.getAnalysis().getBFI(); + auto OptionalBlockCount = BFI->getBlockProfileCount(cast(V)); + auto OptionalFunctionCount = Fn.getEntryCount(); + if (OptionalBlockCount && OptionalFunctionCount) + Hotness = *OptionalFunctionCount * *OptionalBlockCount; + + return Hotness; +} + +void llvm::emitOptimizationRemarkMissed(LLVMContext &Ctx, const Pass &FromPass, + const Function &Fn, + const DebugLoc &DLoc, Value *V, + const Twine &Msg) { + Ctx.diagnose(DiagnosticInfoOptimizationRemarkMissed( + FromPass, Fn, DLoc, Msg, computeHotness(FromPass, Fn, V))); +} Index: lib/IR/DiagnosticInfo.cpp =================================================================== --- lib/IR/DiagnosticInfo.cpp +++ lib/IR/DiagnosticInfo.cpp @@ -170,10 +170,29 @@ return (Filename + ":" + Twine(Line) + ":" + Twine(Column)).str(); } +DiagnosticInfoOptimizationBase::DiagnosticInfoOptimizationBase( + enum DiagnosticKind Kind, enum DiagnosticSeverity Severity, + const Pass &FromPass, const Function &Fn, const DebugLoc &DLoc, + const Twine &Msg, uint64_t Hotness) + : DiagnosticInfoWithDebugLocBase(Kind, Severity, Fn, DLoc), Msg(Msg), + Hotness(Hotness) { + const PassInfo *PI = Pass::lookupPassInfo(FromPass.getPassID()); + assert(PI && "Pass not registered"); + PassName = PI->getPassArgument(); +} + void DiagnosticInfoOptimizationBase::print(DiagnosticPrinter &DP) const { DP << getLocationStr() << ": " << getMsg(); + if (Hotness) + DP << " (hotness: " << Hotness << ")"; } +DiagnosticInfoOptimizationRemarkMissed::DiagnosticInfoOptimizationRemarkMissed( + const Pass &FromPass, const Function &Fn, const DebugLoc &DLoc, + const Twine &Msg, uint64_t Hotness) + : DiagnosticInfoOptimizationBase(DK_OptimizationRemarkMissed, DS_Remark, + FromPass, Fn, DLoc, Msg, Hotness) {} + bool DiagnosticInfoOptimizationRemark::isEnabled() const { return PassRemarksOptLoc.Pattern && PassRemarksOptLoc.Pattern->match(getPassName()); Index: lib/Transforms/Vectorize/LoopVectorize.cpp =================================================================== --- lib/Transforms/Vectorize/LoopVectorize.cpp +++ lib/Transforms/Vectorize/LoopVectorize.cpp @@ -66,6 +66,7 @@ #include "llvm/Analysis/LoopInfo.h" #include "llvm/Analysis/LoopIterator.h" #include "llvm/Analysis/LoopPass.h" +#include "llvm/Analysis/OptimizationDiagnosticInfo.h" #include "llvm/Analysis/ScalarEvolution.h" #include "llvm/Analysis/ScalarEvolutionExpander.h" #include "llvm/Analysis/ScalarEvolutionExpressions.h" @@ -1310,23 +1311,6 @@ LoopAccessReport::emitAnalysis(Message, TheFunction, TheLoop, Name); } -static void emitMissedWarning(Function *F, Loop *L, - const LoopVectorizeHints &LH) { - emitOptimizationRemarkMissed(F->getContext(), LV_NAME, *F, L->getStartLoc(), - LH.emitRemark()); - - if (LH.getForce() == LoopVectorizeHints::FK_Enabled) { - if (LH.getWidth() != 1) - emitLoopVectorizeWarning( - F->getContext(), *F, L->getStartLoc(), - "failed explicitly specified loop vectorization"); - else if (LH.getInterleave() != 1) - emitLoopInterleaveWarning( - F->getContext(), *F, L->getStartLoc(), - "failed explicitly specified loop interleaving"); - } -} - /// LoopVectorizationLegality checks if it is legal to vectorize a loop, and /// to what vectorization factor. /// This class does not look at the profitability of vectorization, only the @@ -2144,6 +2128,22 @@ AU.addPreserved(); AU.addPreserved(); } + + void emitMissedWarning(Function *F, Loop *L, const LoopVectorizeHints &LH) { + emitOptimizationRemarkMissed(F->getContext(), *this, *F, L, + LH.emitRemark()); + + if (LH.getForce() == LoopVectorizeHints::FK_Enabled) { + if (LH.getWidth() != 1) + emitLoopVectorizeWarning( + F->getContext(), *F, L->getStartLoc(), + "failed explicitly specified loop vectorization"); + else if (LH.getInterleave() != 1) + emitLoopInterleaveWarning( + F->getContext(), *F, L->getStartLoc(), + "failed explicitly specified loop interleaving"); + } + } }; } // end anonymous namespace Index: test/Transforms/LoopVectorize/diag-with-hotness-info.ll =================================================================== --- /dev/null +++ test/Transforms/LoopVectorize/diag-with-hotness-info.ll @@ -0,0 +1,191 @@ +; RUN: opt -S -loop-vectorize -pass-remarks-missed=loop-vectorize < %s 2>&1 | FileCheck %s + +; 1 void cold(char *A, char *B, char *C, char *D, char *E, int N) { +; 2 for(int i = 0; i < N; i++) { +; 3 A[i + 1] = A[i] + B[i]; +; 4 C[i] = D[i] * E[i]; +; 5 } +; 6 } +; 7 +; 8 void hot(char *A, char *B, char *C, char *D, char *E, int N) { +; 9 for(int i = 0; i < N; i++) { +; 10 A[i + 1] = A[i] + B[i]; +; 11 C[i] = D[i] * E[i]; +; 12 } +; 13 } +; 14 +; 15 void unknown(char *A, char *B, char *C, char *D, char *E, int N) { +; 16 for(int i = 0; i < N; i++) { +; 17 A[i + 1] = A[i] + B[i]; +; 18 C[i] = D[i] * E[i]; +; 19 } +; 20 } + +; CHECK: remark: /tmp/s.c:2:3: loop not vectorized: use -Rpass-analysis=loop-vectorize for more info (hotness: 600) +; CHECK: remark: /tmp/s.c:9:3: loop not vectorized: use -Rpass-analysis=loop-vectorize for more info (hotness: 600000000) +; CHECK: remark: /tmp/s.c:16:3: loop not vectorized: use -Rpass-analysis=loop-vectorize for more info{{$}} + +; ModuleID = '/tmp/s.c' +source_filename = "/tmp/s.c" +target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128" +target triple = "x86_64-apple-macosx10.11.0" + +; Function Attrs: norecurse nounwind ssp uwtable +define void @cold(i8* nocapture %A, i8* nocapture readonly %B, i8* nocapture %C, i8* nocapture readonly %D, i8* nocapture readonly %E, i32 %N) local_unnamed_addr #0 !dbg !7 !prof !56 { +entry: + %cmp28 = icmp sgt i32 %N, 0, !dbg !9 + br i1 %cmp28, label %for.body, label %for.cond.cleanup, !dbg !10 + +for.cond.cleanup: ; preds = %for.body, %entry + ret void, !dbg !11 + +for.body: ; preds = %entry, %for.body + %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ] + %arrayidx = getelementptr inbounds i8, i8* %A, i64 %indvars.iv, !dbg !12 + %0 = load i8, i8* %arrayidx, align 1, !dbg !12, !tbaa !13 + %arrayidx2 = getelementptr inbounds i8, i8* %B, i64 %indvars.iv, !dbg !16 + %1 = load i8, i8* %arrayidx2, align 1, !dbg !16, !tbaa !13 + %add = add i8 %1, %0, !dbg !17 + %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !10 + %arrayidx7 = getelementptr inbounds i8, i8* %A, i64 %indvars.iv.next, !dbg !18 + store i8 %add, i8* %arrayidx7, align 1, !dbg !19, !tbaa !13 + %arrayidx9 = getelementptr inbounds i8, i8* %D, i64 %indvars.iv, !dbg !20 + %2 = load i8, i8* %arrayidx9, align 1, !dbg !20, !tbaa !13 + %arrayidx12 = getelementptr inbounds i8, i8* %E, i64 %indvars.iv, !dbg !21 + %3 = load i8, i8* %arrayidx12, align 1, !dbg !21, !tbaa !13 + %mul = mul i8 %3, %2, !dbg !22 + %arrayidx16 = getelementptr inbounds i8, i8* %C, i64 %indvars.iv, !dbg !23 + store i8 %mul, i8* %arrayidx16, align 1, !dbg !24, !tbaa !13 + %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !10 + %exitcond = icmp eq i32 %lftr.wideiv, %N, !dbg !10 + br i1 %exitcond, label %for.cond.cleanup, label %for.body, !dbg !10, !llvm.loop !25, !prof !58 +} + +; Function Attrs: norecurse nounwind ssp uwtable +define void @hot(i8* nocapture %A, i8* nocapture readonly %B, i8* nocapture %C, i8* nocapture readonly %D, i8* nocapture readonly %E, i32 %N) local_unnamed_addr #0 !dbg !26 !prof !57 { +entry: + %cmp28 = icmp sgt i32 %N, 0, !dbg !27 + br i1 %cmp28, label %for.body, label %for.cond.cleanup, !dbg !28 + +for.cond.cleanup: ; preds = %for.body, %entry + ret void, !dbg !29 + +for.body: ; preds = %entry, %for.body + %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ] + %arrayidx = getelementptr inbounds i8, i8* %A, i64 %indvars.iv, !dbg !30 + %0 = load i8, i8* %arrayidx, align 1, !dbg !30, !tbaa !13 + %arrayidx2 = getelementptr inbounds i8, i8* %B, i64 %indvars.iv, !dbg !31 + %1 = load i8, i8* %arrayidx2, align 1, !dbg !31, !tbaa !13 + %add = add i8 %1, %0, !dbg !32 + %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !28 + %arrayidx7 = getelementptr inbounds i8, i8* %A, i64 %indvars.iv.next, !dbg !33 + store i8 %add, i8* %arrayidx7, align 1, !dbg !34, !tbaa !13 + %arrayidx9 = getelementptr inbounds i8, i8* %D, i64 %indvars.iv, !dbg !35 + %2 = load i8, i8* %arrayidx9, align 1, !dbg !35, !tbaa !13 + %arrayidx12 = getelementptr inbounds i8, i8* %E, i64 %indvars.iv, !dbg !36 + %3 = load i8, i8* %arrayidx12, align 1, !dbg !36, !tbaa !13 + %mul = mul i8 %3, %2, !dbg !37 + %arrayidx16 = getelementptr inbounds i8, i8* %C, i64 %indvars.iv, !dbg !38 + store i8 %mul, i8* %arrayidx16, align 1, !dbg !39, !tbaa !13 + %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !28 + %exitcond = icmp eq i32 %lftr.wideiv, %N, !dbg !28 + br i1 %exitcond, label %for.cond.cleanup, label %for.body, !dbg !28, !llvm.loop !40, !prof !59 +} + +; Function Attrs: norecurse nounwind ssp uwtable +define void @unknown(i8* nocapture %A, i8* nocapture readonly %B, i8* nocapture %C, i8* nocapture readonly %D, i8* nocapture readonly %E, i32 %N) local_unnamed_addr #0 !dbg !41 { +entry: + %cmp28 = icmp sgt i32 %N, 0, !dbg !42 + br i1 %cmp28, label %for.body, label %for.cond.cleanup, !dbg !43 + +for.cond.cleanup: ; preds = %for.body, %entry + ret void, !dbg !44 + +for.body: ; preds = %entry, %for.body + %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ] + %arrayidx = getelementptr inbounds i8, i8* %A, i64 %indvars.iv, !dbg !45 + %0 = load i8, i8* %arrayidx, align 1, !dbg !45, !tbaa !13 + %arrayidx2 = getelementptr inbounds i8, i8* %B, i64 %indvars.iv, !dbg !46 + %1 = load i8, i8* %arrayidx2, align 1, !dbg !46, !tbaa !13 + %add = add i8 %1, %0, !dbg !47 + %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !43 + %arrayidx7 = getelementptr inbounds i8, i8* %A, i64 %indvars.iv.next, !dbg !48 + store i8 %add, i8* %arrayidx7, align 1, !dbg !49, !tbaa !13 + %arrayidx9 = getelementptr inbounds i8, i8* %D, i64 %indvars.iv, !dbg !50 + %2 = load i8, i8* %arrayidx9, align 1, !dbg !50, !tbaa !13 + %arrayidx12 = getelementptr inbounds i8, i8* %E, i64 %indvars.iv, !dbg !51 + %3 = load i8, i8* %arrayidx12, align 1, !dbg !51, !tbaa !13 + %mul = mul i8 %3, %2, !dbg !52 + %arrayidx16 = getelementptr inbounds i8, i8* %C, i64 %indvars.iv, !dbg !53 + store i8 %mul, i8* %arrayidx16, align 1, !dbg !54, !tbaa !13 + %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !43 + %exitcond = icmp eq i32 %lftr.wideiv, %N, !dbg !43 + br i1 %exitcond, label %for.cond.cleanup, label %for.body, !dbg !43, !llvm.loop !55 +} + +attributes #0 = { norecurse nounwind ssp uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="core2" "target-features"="+cx16,+fxsr,+mmx,+sse,+sse2,+sse3,+ssse3,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" } + +!llvm.dbg.cu = !{!0} +!llvm.module.flags = !{!3, !4, !5} +!llvm.ident = !{!6} + +!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 3.9.0 (trunk 273572) (llvm/trunk 273585)", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !2) +!1 = !DIFile(filename: "/tmp/s.c", directory: "/tmp") +!2 = !{} +!3 = !{i32 2, !"Dwarf Version", i32 2} +!4 = !{i32 2, !"Debug Info Version", i32 3} +!5 = !{i32 1, !"PIC Level", i32 2} +!6 = !{!"clang version 3.9.0 (trunk 273572) (llvm/trunk 273585)"} +!7 = distinct !DISubprogram(name: "cold", scope: !1, file: !1, line: 1, type: !8, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, unit: !0, variables: !2) +!8 = !DISubroutineType(types: !2) +!9 = !DILocation(line: 2, column: 20, scope: !7) +!10 = !DILocation(line: 2, column: 3, scope: !7) +!11 = !DILocation(line: 6, column: 1, scope: !7) +!12 = !DILocation(line: 3, column: 16, scope: !7) +!13 = !{!14, !14, i64 0} +!14 = !{!"omnipotent char", !15, i64 0} +!15 = !{!"Simple C/C++ TBAA"} +!16 = !DILocation(line: 3, column: 23, scope: !7) +!17 = !DILocation(line: 3, column: 21, scope: !7) +!18 = !DILocation(line: 3, column: 5, scope: !7) +!19 = !DILocation(line: 3, column: 14, scope: !7) +!20 = !DILocation(line: 4, column: 12, scope: !7) +!21 = !DILocation(line: 4, column: 19, scope: !7) +!22 = !DILocation(line: 4, column: 17, scope: !7) +!23 = !DILocation(line: 4, column: 5, scope: !7) +!24 = !DILocation(line: 4, column: 10, scope: !7) +!25 = distinct !{!25, !10} +!26 = distinct !DISubprogram(name: "hot", scope: !1, file: !1, line: 8, type: !8, isLocal: false, isDefinition: true, scopeLine: 8, flags: DIFlagPrototyped, isOptimized: true, unit: !0, variables: !2) +!27 = !DILocation(line: 9, column: 20, scope: !26) +!28 = !DILocation(line: 9, column: 3, scope: !26) +!29 = !DILocation(line: 13, column: 1, scope: !26) +!30 = !DILocation(line: 10, column: 16, scope: !26) +!31 = !DILocation(line: 10, column: 23, scope: !26) +!32 = !DILocation(line: 10, column: 21, scope: !26) +!33 = !DILocation(line: 10, column: 5, scope: !26) +!34 = !DILocation(line: 10, column: 14, scope: !26) +!35 = !DILocation(line: 11, column: 12, scope: !26) +!36 = !DILocation(line: 11, column: 19, scope: !26) +!37 = !DILocation(line: 11, column: 17, scope: !26) +!38 = !DILocation(line: 11, column: 5, scope: !26) +!39 = !DILocation(line: 11, column: 10, scope: !26) +!40 = distinct !{!40, !28} +!41 = distinct !DISubprogram(name: "unknown", scope: !1, file: !1, line: 15, type: !8, isLocal: false, isDefinition: true, scopeLine: 15, flags: DIFlagPrototyped, isOptimized: true, unit: !0, variables: !2) +!42 = !DILocation(line: 16, column: 20, scope: !41) +!43 = !DILocation(line: 16, column: 3, scope: !41) +!44 = !DILocation(line: 20, column: 1, scope: !41) +!45 = !DILocation(line: 17, column: 16, scope: !41) +!46 = !DILocation(line: 17, column: 23, scope: !41) +!47 = !DILocation(line: 17, column: 21, scope: !41) +!48 = !DILocation(line: 17, column: 5, scope: !41) +!49 = !DILocation(line: 17, column: 14, scope: !41) +!50 = !DILocation(line: 18, column: 12, scope: !41) +!51 = !DILocation(line: 18, column: 19, scope: !41) +!52 = !DILocation(line: 18, column: 17, scope: !41) +!53 = !DILocation(line: 18, column: 5, scope: !41) +!54 = !DILocation(line: 18, column: 10, scope: !41) +!55 = distinct !{!55, !43} +!56 = !{!"function_entry_count", i64 3} +!57 = !{!"function_entry_count", i64 3000} +!58 = !{!"branch_weights", i32 1, i32 99} +!59 = !{!"branch_weights", i32 1, i32 99}