Index: lib/Transforms/Vectorize/LoopVectorize.cpp =================================================================== --- lib/Transforms/Vectorize/LoopVectorize.cpp (revision 210586) +++ lib/Transforms/Vectorize/LoopVectorize.cpp (working copy) @@ -209,6 +209,24 @@ class LoopVectorizationLegality; class LoopVectorizationCostModel; +/// Optimization analysis message produced during vectorization. Messages inform +/// the user why vectorization did not occur including reporting any +/// unvectorizable variables, types, and function calls. +class Report { + std::string Message; + raw_string_ostream Out; + +public: + Report() : Out(Message) { Out << "loop not vectorized: "; } + + template Report &operator<<(A &Value) { + Out << Value; + return *this; + } + + std::string &str() { return Out.str(); } +}; + /// InnerLoopVectorizer vectorizes loops which contain only one basic /// block to a specified vectorization factor (VF). /// This class performs the widening of scalars into vectors, or multiple @@ -515,10 +533,12 @@ unsigned NumPredStores; LoopVectorizationLegality(Loop *L, ScalarEvolution *SE, const DataLayout *DL, - DominatorTree *DT, TargetLibraryInfo *TLI) + DominatorTree *DT, TargetLibraryInfo *TLI, + Function *F) : NumLoads(0), NumStores(0), NumPredStores(0), TheLoop(L), SE(SE), DL(DL), - DT(DT), TLI(TLI), Induction(nullptr), WidestIndTy(nullptr), - HasFunNoNaNAttr(false), MaxSafeDepDistBytes(-1U) {} + DT(DT), TLI(TLI), TheFunction(F), Induction(nullptr), + WidestIndTy(nullptr), HasFunNoNaNAttr(false), MaxSafeDepDistBytes(-1U) { + } /// This enum represents the kinds of reductions that we support. enum ReductionKind { @@ -747,6 +767,14 @@ /// invariant. void collectStridedAcccess(Value *LoadOrStoreInst); + /// Report an analysis message to assist the user in diagnosing loops that are + /// not vectorized. + void emitAnalysis(Report &Message) { + emitOptimizationRemarkAnalysis(TheFunction->getContext(), DEBUG_TYPE, + *TheFunction, TheLoop->getStartLoc(), + Message.str()); + } + /// The loop that we evaluate. Loop *TheLoop; /// Scev analysis. @@ -757,6 +785,8 @@ DominatorTree *DT; /// Target Library Info. TargetLibraryInfo *TLI; + /// Parent function + Function *TheFunction; // --- vectorization state --- // @@ -1155,10 +1185,16 @@ } } + // Function containing loop + Function *F = L->getHeader()->getParent(); + // Check if it is legal to vectorize the loop. - LoopVectorizationLegality LVL(L, SE, DL, DT, TLI); + LoopVectorizationLegality LVL(L, SE, DL, DT, TLI, F); if (!LVL.canVectorize()) { DEBUG(dbgs() << "LV: Not vectorizing: Cannot prove legality.\n"); + emitOptimizationRemarkMissed( + F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(), + "loop not vectorized"); return false; } @@ -1167,7 +1203,6 @@ // Check the function attributes to find out if this function should be // optimized for size. - Function *F = L->getHeader()->getParent(); bool OptForSize = Hints.getForce() != LoopVectorizeHints::FK_Enabled && F->hasFnAttribute(Attribute::OptimizeForSize); @@ -1190,6 +1225,12 @@ if (F->hasFnAttribute(Attribute::NoImplicitFloat)) { DEBUG(dbgs() << "LV: Can't vectorize when the NoImplicitFloat" "attribute is used.\n"); + emitOptimizationRemarkAnalysis( + F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(), + "implicit floating point instructions are disabled due to " + "NoImplicitFloat attribute"); + emitOptimizationRemarkMissed(F->getContext(), DEBUG_TYPE, *F, + L->getStartLoc(), "loop not vectorized"); return false; } @@ -1209,8 +1250,12 @@ if (VF.Width == 1) { DEBUG(dbgs() << "LV: Vectorization is possible but not beneficial.\n"); - if (UF == 1) + if (UF == 1) { + emitOptimizationRemarkMissed( + F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(), + "not beneficial to vectorize and user disabled interleaving."); return false; + } DEBUG(dbgs() << "LV: Trying to at least unroll the loops.\n"); // Report the unrolling decision. @@ -1220,6 +1265,7 @@ " (vectorization not beneficial)")); // We decided not to vectorize, but we may want to unroll. + InnerLoopUnroller Unroller(L, SE, LI, DT, DL, TLI, UF); Unroller.vectorize(&LVL); } else { @@ -3254,20 +3300,29 @@ bool LoopVectorizationLegality::canVectorize() { // We must have a loop in canonical form. Loops with indirectbr in them cannot // be canonicalized. - if (!TheLoop->getLoopPreheader()) + if (!TheLoop->getLoopPreheader()) { + emitAnalysis( + Report() << "loop is not in canonical form due to indirect branch"); return false; + } // We can only vectorize innermost loops. - if (TheLoop->getSubLoopsVector().size()) + if (TheLoop->getSubLoopsVector().size()) { + emitAnalysis(Report() << "loop is not the innermost loop"); return false; + } // We must have a single backedge. - if (TheLoop->getNumBackEdges() != 1) + if (TheLoop->getNumBackEdges() != 1) { + emitAnalysis(Report() << "loop does not have a single backedge"); return false; + } // We must have a single exiting block. - if (!TheLoop->getExitingBlock()) + if (!TheLoop->getExitingBlock()) { + emitAnalysis(Report() << "loop does not have a single exit point"); return false; + } // We need to have a loop header. DEBUG(dbgs() << "LV: Found a loop: " << @@ -3276,6 +3331,8 @@ // Check if we can if-convert non-single-bb loops. unsigned NumBlocks = TheLoop->getNumBlocks(); if (NumBlocks != 1 && !canVectorizeWithIfConvert()) { + emitAnalysis(Report() << "conditional branches could not be substituted " + "for a select instruction"); DEBUG(dbgs() << "LV: Can't if-convert the loop.\n"); return false; } @@ -3283,6 +3340,9 @@ // ScalarEvolution needs to be able to find the exit count. const SCEV *ExitCount = SE->getBackedgeTakenCount(TheLoop); if (ExitCount == SE->getCouldNotCompute()) { + emitAnalysis( + Report() + << "scalar evolution could not determine number of loop iterations"); DEBUG(dbgs() << "LV: SCEV could not compute the loop exit count.\n"); return false; } @@ -3335,7 +3395,8 @@ /// \brief Check that the instruction has outside loop users and is not an /// identified reduction variable. static bool hasOutsideLoopUser(const Loop *TheLoop, Instruction *Inst, - SmallPtrSet &Reductions) { + SmallPtrSet &Reductions, + Report *Message = nullptr) { // Reduction instructions are allowed to have exit users. All other // instructions must not have external users. if (!Reductions.count(Inst)) @@ -3344,6 +3405,8 @@ Instruction *UI = cast(U); // This user may be a reduction exit value. if (!TheLoop->contains(UI)) { + if (Message) + *Message << "found an outside user for: " << *UI; DEBUG(dbgs() << "LV: Found an outside user for : " << *UI << '\n'); return true; } @@ -3376,6 +3439,7 @@ if (!PhiTy->isIntegerTy() && !PhiTy->isFloatingPointTy() && !PhiTy->isPointerTy()) { + emitAnalysis(Report() << "non-int non-pointer PHI: " << PhiTy); DEBUG(dbgs() << "LV: Found an non-int non-pointer PHI.\n"); return false; } @@ -3386,13 +3450,18 @@ if (*bb != Header) { // Check that this instruction has no outside users or is an // identified reduction value with an outside user. - if(!hasOutsideLoopUser(TheLoop, it, AllowedExit)) + Report Message; + if (!hasOutsideLoopUser(TheLoop, it, AllowedExit, &Message)) continue; + emitAnalysis(Message); + DEBUG(dbgs() << "LV: Found an non-int non-pointer PHI.\n"); return false; } // We only allow if-converted PHIs with more than two incoming values. if (Phi->getNumIncomingValues() != 2) { + emitAnalysis(Report() << "conditional branch could not be " + "substituded for a select instruction"); DEBUG(dbgs() << "LV: Found an invalid PHI.\n"); return false; } @@ -3467,6 +3536,7 @@ continue; } + emitAnalysis(Report() << "unidentified PHI: " << *Phi); DEBUG(dbgs() << "LV: Found an unidentified PHI."<< *Phi <<"\n"); return false; }// end of PHI handling @@ -3475,6 +3545,7 @@ // calls and we do handle certain intrinsic and libm functions. CallInst *CI = dyn_cast(it); if (CI && !getIntrinsicIDForCall(CI, TLI) && !isa(CI)) { + emitAnalysis(Report() << "call cannot be vectorized: " << *CI); DEBUG(dbgs() << "LV: Found a call site.\n"); return false; } @@ -3484,6 +3555,7 @@ if (CI && hasVectorInstrinsicScalarOpd(getIntrinsicIDForCall(CI, TLI), 1)) { if (!SE->isLoopInvariant(SE->getSCEV(CI->getOperand(1)), TheLoop)) { + emitAnalysis(Report() << "intrinsic cannot be vectorized: " << *CI); DEBUG(dbgs() << "LV: Found unvectorizable intrinsic " << *CI << "\n"); return false; } @@ -3493,6 +3565,8 @@ // Also, we can't vectorize extractelement instructions. if ((!VectorType::isValidElementType(it->getType()) && !it->getType()->isVoidTy()) || isa(it)) { + emitAnalysis(Report() + << "type cannot be vectorized: " << *it->getType()); DEBUG(dbgs() << "LV: Found unvectorizable type.\n"); return false; } @@ -3521,8 +3595,11 @@ if (!Induction) { DEBUG(dbgs() << "LV: Did not find one integer induction var.\n"); - if (Inductions.empty()) + if (Inductions.empty()) { + emitAnalysis(Report() + << "integer-type loop induction variable was not found"); return false; + } } return true; @@ -4429,8 +4506,8 @@ continue; LoadInst *Ld = dyn_cast(it); - if (!Ld) return false; - if (!Ld->isSimple() && !IsAnnotatedParallel) { + if (!Ld || (!Ld->isSimple() && !IsAnnotatedParallel)) { + emitAnalysis(Report() << "non-simple load"); DEBUG(dbgs() << "LV: Found a non-simple load.\n"); return false; } @@ -4445,6 +4522,7 @@ StoreInst *St = dyn_cast(it); if (!St) return false; if (!St->isSimple() && !IsAnnotatedParallel) { + emitAnalysis(Report() << "non-simple store"); DEBUG(dbgs() << "LV: Found a non-simple store.\n"); return false; } @@ -4481,6 +4559,7 @@ Value* Ptr = ST->getPointerOperand(); if (isUniform(Ptr)) { + emitAnalysis(Report() << "store into uniform address"); DEBUG(dbgs() << "LV: We don't allow storing to uniform addresses\n"); return false; } @@ -4559,6 +4638,7 @@ } if (NeedRTCheck && !CanDoRT) { + emitAnalysis(Report() << "cannot identify array bounds"); DEBUG(dbgs() << "LV: We can't vectorize because we can't find " << "the array bounds.\n"); PtrRtCheck.reset(); @@ -4589,6 +4669,7 @@ // Check that we did not collect too many pointers or found an unsizeable // pointer. if (!CanDoRT || NumComparisons > RuntimeMemoryCheckThreshold) { + emitAnalysis(Report() << "memory checks"); DEBUG(dbgs() << "LV: Can't vectorize with memory checks\n"); PtrRtCheck.reset(); return false; Index: test/Transforms/LoopVectorize/X86/vectorization-remarks-missed.ll =================================================================== --- test/Transforms/LoopVectorize/X86/vectorization-remarks-missed.ll (revision 0) +++ test/Transforms/LoopVectorize/X86/vectorization-remarks-missed.ll (working copy) @@ -0,0 +1,125 @@ +; RUN: opt < %s -O3 -mtriple=x86_64-unknown-linux -S -pass-remarks-missed='loop-vectorize' 2>&1 | FileCheck -check-prefix=MARK %s + +; MARK: remark: {{.*}}:5:5: loop not vectorized +; MARK: remark: {{.*}}:10:5: loop not vectorized + +target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128" + +; Function Attrs: nounwind ssp uwtable +define void @test_calltest(i32* nocapture readonly %List, i32 %Length) #0 { +entry: + %cmp3 = icmp sgt i32 %Length, 0, !dbg !11 + br i1 %cmp3, label %for.body.preheader, label %for.end, !dbg !11 + +for.body.preheader: ; preds = %entry + br label %for.body, !dbg !13 + +for.body: ; preds = %for.body.preheader, %for.body + %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ] + %arrayidx = getelementptr inbounds i32* %List, i64 %indvars.iv, !dbg !13 + %0 = load i32* %arrayidx, align 4, !dbg !13, !tbaa !14 + %call = tail call i32 @calltest(i32 %0) #2, !dbg !13 + %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !11 + %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !11 + %exitcond = icmp eq i32 %lftr.wideiv, %Length, !dbg !11 + br i1 %exitcond, label %for.end.loopexit, label %for.body, !dbg !11 + +for.end.loopexit: ; preds = %for.body + br label %for.end + +for.end: ; preds = %for.end.loopexit, %entry + ret void, !dbg !18 +} + +declare i32 @calltest(i32) #1 + +; Function Attrs: nounwind ssp uwtable +define void @test_array_bounds(i32* nocapture %List, i32 %Length) #0 { +entry: + %cmp10 = icmp sgt i32 %Length, 0, !dbg !19 + br i1 %cmp10, label %for.body.preheader, label %for.end, !dbg !19 + +for.body.preheader: ; preds = %entry + %xtraiter = urem i32 %Length, 2 + %lcmp = icmp ne i32 %xtraiter, 0 + br i1 %lcmp, label %for.body.unr, label %for.body.preheader.split + +for.body.unr: ; preds = %for.body.preheader + %arrayidx.unr = getelementptr inbounds i32* %List, i64 0, !dbg !21 + %0 = load i32* %arrayidx.unr, align 4, !dbg !21, !tbaa !14 + %idxprom3.unr = sext i32 %0 to i64, !dbg !21 + %arrayidx4.unr = getelementptr inbounds i32* %List, i64 %idxprom3.unr, !dbg !21 + store i32 %0, i32* %arrayidx4.unr, align 4, !dbg !21, !tbaa !14 + %indvars.iv.next.unr = add nuw nsw i64 0, 1, !dbg !19 + %lftr.wideiv.unr = trunc i64 %indvars.iv.next.unr to i32, !dbg !19 + %exitcond.unr = icmp eq i32 %lftr.wideiv.unr, %Length, !dbg !19 + br label %for.body.preheader.split + +for.body.preheader.split: ; preds = %for.body.unr, %for.body.preheader + %indvars.iv.unr = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next.unr, %for.body.unr ] + %1 = icmp ult i32 %Length, 2 + br i1 %1, label %for.end.loopexit, label %for.body.preheader.split.split + +for.body.preheader.split.split: ; preds = %for.body.preheader.split + br label %for.body, !dbg !21 + +for.body: ; preds = %for.body, %for.body.preheader.split.split + %indvars.iv = phi i64 [ %indvars.iv.unr, %for.body.preheader.split.split ], [ %indvars.iv.next.1, %for.body ] + %arrayidx = getelementptr inbounds i32* %List, i64 %indvars.iv, !dbg !21 + %2 = load i32* %arrayidx, align 4, !dbg !21, !tbaa !14 + %idxprom3 = sext i32 %2 to i64, !dbg !21 + %arrayidx4 = getelementptr inbounds i32* %List, i64 %idxprom3, !dbg !21 + store i32 %2, i32* %arrayidx4, align 4, !dbg !21, !tbaa !14 + %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !19 + %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !19 + %arrayidx.1 = getelementptr inbounds i32* %List, i64 %indvars.iv.next, !dbg !21 + %3 = load i32* %arrayidx.1, align 4, !dbg !21, !tbaa !14 + %idxprom3.1 = sext i32 %3 to i64, !dbg !21 + %arrayidx4.1 = getelementptr inbounds i32* %List, i64 %idxprom3.1, !dbg !21 + store i32 %3, i32* %arrayidx4.1, align 4, !dbg !21, !tbaa !14 + %indvars.iv.next.1 = add nuw nsw i64 %indvars.iv.next, 1, !dbg !19 + %lftr.wideiv.1 = trunc i64 %indvars.iv.next.1 to i32, !dbg !19 + %exitcond.1 = icmp eq i32 %lftr.wideiv.1, %Length, !dbg !19 + br i1 %exitcond.1, label %for.end.loopexit.unr-lcssa, label %for.body, !dbg !19 + +for.end.loopexit.unr-lcssa: ; preds = %for.body + br label %for.end.loopexit + +for.end.loopexit: ; preds = %for.body.preheader.split, %for.end.loopexit.unr-lcssa + br label %for.end + +for.end: ; preds = %for.end.loopexit, %entry + ret void, !dbg !22 +} + +attributes #0 = { nounwind ssp uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" } +attributes #1 = { "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" } +attributes #2 = { nounwind } + +!llvm.dbg.cu = !{!0} +!llvm.module.flags = !{!8, !9} +!llvm.ident = !{!10} + +!0 = metadata !{i32 786449, metadata !1, i32 12, metadata !"clang version 3.5.0 (http://llvm.org/git/clang.git 6cb9eb5cd6a98a2fb13ae499aa5cd3a8feaa14e0) (http://llvm.org/git/llvm.git 4887e4317ea978ee5ba25908c0f1cfac1abd095a)", i1 true, metadata !"", i32 0, metadata !2, metadata !2, metadata !3, metadata !2, metadata !2, metadata !"", i32 2} ; [ DW_TAG_compile_unit ] [/Users/tnowicki/Development/feature_pragma_vectorize/vectorize_report3.c] [DW_LANG_C99] +!1 = metadata !{metadata !"vectorize_report3.c", metadata !"/Users/tnowicki/Development/feature_pragma_vectorize"} +!2 = metadata !{} +!3 = metadata !{metadata !4, metadata !7} +!4 = metadata !{i32 786478, metadata !1, metadata !5, metadata !"test_calltest", metadata !"test_calltest", metadata !"", i32 3, metadata !6, i1 false, i1 true, i32 0, i32 0, null, i32 256, i1 true, void (i32*, i32)* @test_calltest, null, null, metadata !2, i32 3} ; [ DW_TAG_subprogram ] [line 3] [def] [test_calltest] +!5 = metadata !{i32 786473, metadata !1} ; [ DW_TAG_file_type ] [/Users/tnowicki/Development/feature_pragma_vectorize/vectorize_report3.c] +!6 = metadata !{i32 786453, i32 0, null, metadata !"", i32 0, i64 0, i64 0, i64 0, i32 0, null, metadata !2, i32 0, null, null, null} ; [ DW_TAG_subroutine_type ] [line 0, size 0, align 0, offset 0] [from ] +!7 = metadata !{i32 786478, metadata !1, metadata !5, metadata !"test_array_bounds", metadata !"test_array_bounds", metadata !"", i32 8, metadata !6, i1 false, i1 true, i32 0, i32 0, null, i32 256, i1 true, void (i32*, i32)* @test_array_bounds, null, null, metadata !2, i32 8} ; [ DW_TAG_subprogram ] [line 8] [def] [test_array_bounds] +!8 = metadata !{i32 2, metadata !"Dwarf Version", i32 2} +!9 = metadata !{i32 2, metadata !"Debug Info Version", i32 1} +!10 = metadata !{metadata !"clang version 3.5.0 (http://llvm.org/git/clang.git 6cb9eb5cd6a98a2fb13ae499aa5cd3a8feaa14e0) (http://llvm.org/git/llvm.git 4887e4317ea978ee5ba25908c0f1cfac1abd095a)"} +!11 = metadata !{i32 4, i32 8, metadata !12, null} +!12 = metadata !{i32 786443, metadata !1, metadata !4, i32 4, i32 3, i32 0, i32 0} ; [ DW_TAG_lexical_block ] [/Users/tnowicki/Development/feature_pragma_vectorize/vectorize_report3.c] +!13 = metadata !{i32 5, i32 5, metadata !12, null} +!14 = metadata !{metadata !15, metadata !15, i64 0} +!15 = metadata !{metadata !"int", metadata !16, i64 0} +!16 = metadata !{metadata !"omnipotent char", metadata !17, i64 0} +!17 = metadata !{metadata !"Simple C/C++ TBAA"} +!18 = metadata !{i32 6, i32 1, metadata !4, null} +!19 = metadata !{i32 9, i32 8, metadata !20, null} +!20 = metadata !{i32 786443, metadata !1, metadata !7, i32 9, i32 3, i32 0, i32 1} ; [ DW_TAG_lexical_block ] [/Users/tnowicki/Development/feature_pragma_vectorize/vectorize_report3.c] +!21 = metadata !{i32 10, i32 5, metadata !20, null} +!22 = metadata !{i32 11, i32 1, metadata !7, null} Index: test/Transforms/LoopVectorize/ee-crash.ll =================================================================== --- test/Transforms/LoopVectorize/ee-crash.ll (revision 210586) +++ test/Transforms/LoopVectorize/ee-crash.ll (working copy) @@ -1,4 +1,5 @@ ; RUN: opt < %s -loop-vectorize -force-vector-unroll=1 -force-vector-width=4 -S | FileCheck %s +; RUN: opt < %s -loop-vectorize -force-vector-unroll=1 -force-vector-width=4 -S -pass-remarks-analysis='loop-vectorize' 2>&1 | FileCheck -check-prefix=MARK %s target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" target triple = "x86_64-apple-macosx10.8.0" @@ -8,6 +9,7 @@ ; CHECK-LABEL: @_Z4foo1Pii( ; CHECK-NOT: <4 x i32> ; CHECK: ret +; MARK: remark: :0:0: loop not vectorized: type cannot be vectorized: i32 define i32 @_Z4foo1Pii(i32* %A, i32 %n, <2 x i32> %q) #0 { entry: %idx.ext = sext i32 %n to i64 Index: test/Transforms/LoopVectorize/gcc-examples.ll =================================================================== --- test/Transforms/LoopVectorize/gcc-examples.ll (revision 210586) +++ test/Transforms/LoopVectorize/gcc-examples.ll (working copy) @@ -1,5 +1,6 @@ ; RUN: opt < %s -loop-vectorize -force-vector-width=4 -force-vector-unroll=1 -dce -instcombine -S | FileCheck %s ; RUN: opt < %s -loop-vectorize -force-vector-width=4 -force-vector-unroll=4 -dce -instcombine -S | FileCheck %s -check-prefix=UNROLL +; RUN: opt < %s -loop-vectorize -force-vector-width=4 -force-vector-unroll=1 -dce -instcombine -S -pass-remarks-analysis='loop-vectorize' 2>&1 | FileCheck -check-prefix=MARK %s target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" target triple = "x86_64-apple-macosx10.8.0" @@ -214,6 +215,7 @@ ;UNROLL: store <4 x i32> ;UNROLL: store <4 x i32> ;UNROLL: ret void +;MARK: remark: :0:0: loop not vectorized: unidentified PHI define void @example8(i32 %x) nounwind uwtable ssp { br label %.preheader Index: test/Transforms/LoopVectorize/hoist-loads.ll =================================================================== --- test/Transforms/LoopVectorize/hoist-loads.ll (revision 210586) +++ test/Transforms/LoopVectorize/hoist-loads.ll (working copy) @@ -1,4 +1,5 @@ ; RUN: opt -loop-vectorize -force-vector-width=2 -force-vector-unroll=1 -S < %s | FileCheck %s +; RUN: opt -loop-vectorize -force-vector-width=2 -force-vector-unroll=1 -S -pass-remarks-analysis='loop-vectorize' < %s 2>&1 | FileCheck -check-prefix=MARK %s target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" @@ -8,7 +9,6 @@ ; Make sure we can vectorize in the presence of hoistable conditional loads. ; CHECK-LABEL: @hoist_cond_load( ; CHECK: load <2 x float> - define void @hoist_cond_load() { entry: br label %for.body @@ -40,7 +40,7 @@ ; accessed. ; CHECK-LABEL: @dont_hoist_cond_load( ; CHECK-NOT: load <2 x float> - +; MARK: remark: :0:0: loop not vectorized: conditional branches could not be substituted for a select instruction define void @dont_hoist_cond_load() { entry: br label %for.body Index: test/Transforms/LoopVectorize/if-conversion-reduction.ll =================================================================== --- test/Transforms/LoopVectorize/if-conversion-reduction.ll (revision 210586) +++ test/Transforms/LoopVectorize/if-conversion-reduction.ll (working copy) @@ -1,4 +1,5 @@ ; RUN: opt < %s -loop-vectorize -force-vector-unroll=1 -force-vector-width=4 -enable-if-conversion -dce -instcombine -S | FileCheck %s +; RUN: opt < %s -loop-vectorize -force-vector-unroll=1 -force-vector-width=4 -enable-if-conversion -dce -instcombine -S -pass-remarks-analysis='loop-vectorize' 2>&1 | FileCheck -check-prefix=MARK %s target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" target triple = "x86_64-apple-macosx10.9.0" @@ -6,6 +7,7 @@ ;CHECK-LABEL: @reduction_func( ;CHECK-NOT: load <4 x i32> ;CHECK: ret i32 +;MARK: remark: :0:0: loop not vectorized: unidentified PHI define i32 @reduction_func(i32* nocapture %A, i32 %n) nounwind uwtable readonly ssp { entry: %cmp10 = icmp sgt i32 %n, 0 Index: test/Transforms/LoopVectorize/if-conversion.ll =================================================================== --- test/Transforms/LoopVectorize/if-conversion.ll (revision 210586) +++ test/Transforms/LoopVectorize/if-conversion.ll (working copy) @@ -1,4 +1,5 @@ ; RUN: opt < %s -loop-vectorize -force-vector-unroll=1 -force-vector-width=4 -enable-if-conversion -dce -instcombine -S | FileCheck %s +; RUN: opt < %s -loop-vectorize -force-vector-unroll=1 -force-vector-width=2 -enable-if-conversion -dce -instcombine -S -pass-remarks-analysis='loop-vectorize' 2>&1 | FileCheck -check-prefix=MARK %s target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" target triple = "x86_64-apple-macosx10.9.0" @@ -70,6 +71,8 @@ ; return sum; ; } +;MARK: remark: :0:0: loop not vectorized: conditional branches could not be substituted for a select instruction + ;CHECK-LABEL: @reduction_func( ;CHECK: load <4 x i32> ;CHECK: icmp sgt <4 x i32> @@ -116,6 +119,8 @@ ; CHECK-LABEL: trapping_constant_expression ; CHECK-NOT: or <4 x i32> +; MARK: remark: :0:0: loop not vectorized: conditional branches could not be substituted for a select instruction + define i32 @trapping_constant_expression() { entry: br label %for.body Index: test/Transforms/LoopVectorize/no_outside_user.ll =================================================================== --- test/Transforms/LoopVectorize/no_outside_user.ll (revision 210586) +++ test/Transforms/LoopVectorize/no_outside_user.ll (working copy) @@ -1,4 +1,5 @@ ; RUN: opt -S -loop-vectorize -force-vector-unroll=1 -force-vector-width=2 < %s | FileCheck %s +; RUN: opt -S -loop-vectorize -force-vector-unroll=1 -force-vector-width=2 -pass-remarks-analysis='loop-vectorize' < %s 2>&1 | FileCheck -check-prefix=MARK %s target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:32:32-n8:16:32-S128" @@ -15,6 +16,8 @@ ; CHECK-LABEL: @main( ; CHECK-NOT: <2 x i32> +; MARK: remark: :0:0: loop not vectorized: found an outside user for: %.lcssa = phi i32 [ %tmp17, %bb16 ] + define i32 @main() { bb: %b.promoted = load i32* @b, align 4