From 12c1f7675f0fe14a3071eebf656f3805ea94a2a2 Mon Sep 17 00:00:00 2001 From: Tim Northover Date: Mon, 10 Sep 2018 14:26:44 +0000 Subject: [PATCH] InstCombine: move hasOneUse check to the top of foldICmpAddConstant There were two combines not covered by the check before now, neither of which actually differed from normal in the benefit analysis. The most recent seems to be because it was just added at the top of the function (naturally). The older is from way back in 2008 (r46687) when we just didn't put those checks in so routinely, and has been diligently maintained since. llvm-svn: 341831 --- .../InstCombine/InstCombineCompares.cpp | 6 +++--- .../ValueTracking/non-negative-phi-bits.ll | 2 +- llvm/test/Transforms/InstCombine/icmp-add.ll | 17 +++++++++++++++++ .../LoopVectorize/X86/masked_load_store.ll | 6 +++--- .../LoopVectorize/if-conversion-nest.ll | 2 +- .../Transforms/LoopVectorize/runtime-check.ll | 2 +- 6 files changed, 26 insertions(+), 9 deletions(-) diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp index e0a1083b9279..4b5891461bc3 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp @@ -2337,6 +2337,9 @@ Instruction *InstCombiner::foldICmpAddConstant(ICmpInst &Cmp, Type *Ty = Add->getType(); CmpInst::Predicate Pred = Cmp.getPredicate(); + if (!Add->hasOneUse()) + return nullptr; + // If the add does not wrap, we can always adjust the compare by subtracting // the constants. Equality comparisons are handled elsewhere. SGE/SLE/UGE/ULE // are canonicalized to SGT/SLT/UGT/ULT. @@ -2370,9 +2373,6 @@ Instruction *InstCombiner::foldICmpAddConstant(ICmpInst &Cmp, return new ICmpInst(ICmpInst::ICMP_UGE, X, ConstantInt::get(Ty, Lower)); } - if (!Add->hasOneUse()) - return nullptr; - // X+C (X & -C2) == C // iff C & (C2-1) == 0 // C2 is a power of 2 diff --git a/llvm/test/Analysis/ValueTracking/non-negative-phi-bits.ll b/llvm/test/Analysis/ValueTracking/non-negative-phi-bits.ll index 3b1c43df5a70..059bbaa3c4e7 100755 --- a/llvm/test/Analysis/ValueTracking/non-negative-phi-bits.ll +++ b/llvm/test/Analysis/ValueTracking/non-negative-phi-bits.ll @@ -8,7 +8,7 @@ define void @test() #0 { ; CHECK: for.body: ; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] ; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 -; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ult i64 [[INDVARS_IV]], 39 +; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ult i64 [[INDVARS_IV_NEXT]], 40 ; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]] ; CHECK: for.end: ; CHECK-NEXT: ret void diff --git a/llvm/test/Transforms/InstCombine/icmp-add.ll b/llvm/test/Transforms/InstCombine/icmp-add.ll index 901f98c82952..092b69485b07 100644 --- a/llvm/test/Transforms/InstCombine/icmp-add.ll +++ b/llvm/test/Transforms/InstCombine/icmp-add.ll @@ -73,6 +73,23 @@ define i1 @test4(i32 %a) { ret i1 %c } +define { i32, i1 } @test4multiuse(i32 %a) { +; CHECK-LABEL: @test4multiuse( +; CHECK: [[B:%.*]] = add i32 %a, -2147483644 +; CHECK: [[C:%.*]] = icmp slt i32 %b, -4 +; CHECK: [[TMP:%.*]] = insertvalue { i32, i1 } undef, i32 [[B]], 0 +; CHECK: [[RES:%.*]] = insertvalue { i32, i1 } [[TMP]], i1 [[C]], 1 +; CHECK: ret { i32, i1 } [[RES]] + + %b = add i32 %a, -2147483644 + %c = icmp slt i32 %b, -4 + + %tmp = insertvalue { i32, i1 } undef, i32 %b, 0 + %res = insertvalue { i32, i1 } %tmp, i1 %c, 1 + + ret { i32, i1 } %res +} + define <2 x i1> @test4vec(<2 x i32> %a) { ; CHECK-LABEL: @test4vec( ; CHECK-NEXT: [[C:%.*]] = icmp slt <2 x i32> %a, diff --git a/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll b/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll index b3387435f2bc..8e948639ba1d 100644 --- a/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll +++ b/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll @@ -1878,7 +1878,7 @@ define void @foo4(double* %A, double* %B, i32* %trigger) { ; AVX1-NEXT: br label [[FOR_INC]] ; AVX1: for.inc: ; AVX1-NEXT: [[INDVARS_IV_NEXT:%.*]] = or i64 [[INDVARS_IV]], 16 -; AVX1-NEXT: [[CMP:%.*]] = icmp ult i64 [[INDVARS_IV]], 9984 +; AVX1-NEXT: [[CMP:%.*]] = icmp ult i64 [[INDVARS_IV_NEXT]], 10000 ; AVX1-NEXT: br i1 [[CMP]], label [[FOR_BODY_1:%.*]], label [[FOR_END:%.*]] ; AVX1: for.end: ; AVX1-NEXT: ret void @@ -1920,7 +1920,7 @@ define void @foo4(double* %A, double* %B, i32* %trigger) { ; AVX2-NEXT: br label [[FOR_INC]] ; AVX2: for.inc: ; AVX2-NEXT: [[INDVARS_IV_NEXT:%.*]] = or i64 [[INDVARS_IV]], 16 -; AVX2-NEXT: [[CMP:%.*]] = icmp ult i64 [[INDVARS_IV]], 9984 +; AVX2-NEXT: [[CMP:%.*]] = icmp ult i64 [[INDVARS_IV_NEXT]], 10000 ; AVX2-NEXT: br i1 [[CMP]], label [[FOR_BODY_1:%.*]], label [[FOR_END:%.*]] ; AVX2: for.end: ; AVX2-NEXT: ret void @@ -2119,7 +2119,7 @@ define void @foo4(double* %A, double* %B, i32* %trigger) { ; AVX512-NEXT: br label [[FOR_INC_3]] ; AVX512: for.inc.3: ; AVX512-NEXT: [[INDVARS_IV_NEXT_3]] = add nsw i64 [[INDVARS_IV]], 64 -; AVX512-NEXT: [[CMP_3:%.*]] = icmp ult i64 [[INDVARS_IV_NEXT_2]], 9984 +; AVX512-NEXT: [[CMP_3:%.*]] = icmp ult i64 [[INDVARS_IV_NEXT_3]], 10000 ; AVX512-NEXT: br i1 [[CMP_3]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop !52 ; entry: diff --git a/llvm/test/Transforms/LoopVectorize/if-conversion-nest.ll b/llvm/test/Transforms/LoopVectorize/if-conversion-nest.ll index 458e350a5e09..311d92b94af1 100644 --- a/llvm/test/Transforms/LoopVectorize/if-conversion-nest.ll +++ b/llvm/test/Transforms/LoopVectorize/if-conversion-nest.ll @@ -12,7 +12,7 @@ define i32 @foo(i32* nocapture %A, i32* nocapture %B, i32 %n) { ; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1 ; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64 ; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1 -; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3 +; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4 ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]] ; CHECK: vector.memcheck: ; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1 diff --git a/llvm/test/Transforms/LoopVectorize/runtime-check.ll b/llvm/test/Transforms/LoopVectorize/runtime-check.ll index de43a613c090..88489faa831a 100644 --- a/llvm/test/Transforms/LoopVectorize/runtime-check.ll +++ b/llvm/test/Transforms/LoopVectorize/runtime-check.ll @@ -18,7 +18,7 @@ define i32 @foo(float* nocapture %a, float* nocapture %b, i32 %n) nounwind uwtab ; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1, !dbg !9 ; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64, !dbg !9 ; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1, !dbg !9 -; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3, !dbg !9 +; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4, !dbg !9 ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]], !dbg !9 ; CHECK: vector.memcheck: ; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1, !dbg !9