[llvm] 393cc6a - [LoopVectorize] Regenerate runtime-check.ll
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Mon Sep 19 02:38:58 PDT 2022
Author: Simon Pilgrim
Date: 2022-09-19T10:25:48+01:00
New Revision: 393cc6a354c625abadff1d45c53b847ba36b45f2
URL: https://github.com/llvm/llvm-project/commit/393cc6a354c625abadff1d45c53b847ba36b45f2
DIFF: https://github.com/llvm/llvm-project/commit/393cc6a354c625abadff1d45c53b847ba36b45f2.diff
LOG: [LoopVectorize] Regenerate runtime-check.ll
Added:
Modified:
llvm/test/Transforms/LoopVectorize/runtime-check.ll
Removed:
################################################################################
diff --git a/llvm/test/Transforms/LoopVectorize/runtime-check.ll b/llvm/test/Transforms/LoopVectorize/runtime-check.ll
index fe0f9248f9e0..480c32a7194f 100644
--- a/llvm/test/Transforms/LoopVectorize/runtime-check.ll
+++ b/llvm/test/Transforms/LoopVectorize/runtime-check.ll
@@ -13,56 +13,78 @@ target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f3
define i32 @foo(float* nocapture %a, float* nocapture %b, i32 %n) nounwind uwtable ssp {
; CHECK-LABEL: @foo(
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[B2:%.*]] = ptrtoint float* [[B:%.*]] to i64, [[DBG4:!dbg !.*]]
-; CHECK-NEXT: [[A1:%.*]] = ptrtoint float* [[A:%.*]] to i64, [[DBG4]]
-; CHECK-NEXT: [[CMP6:%.*]] = icmp sgt i32 [[N:%.*]], 0, [[DBG4]]
-; CHECK-NEXT: br i1 [[CMP6]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]], [[DBG4]]
+; CHECK-NEXT: [[B2:%.*]] = ptrtoint float* [[B:%.*]] to i64, !dbg [[DBG4:![0-9]+]]
+; CHECK-NEXT: [[A1:%.*]] = ptrtoint float* [[A:%.*]] to i64, !dbg [[DBG4]]
+; CHECK-NEXT: [[CMP6:%.*]] = icmp sgt i32 [[N:%.*]], 0, !dbg [[DBG4]]
+; CHECK-NEXT: br i1 [[CMP6]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]], !dbg [[DBG4]]
; CHECK: for.body.preheader:
-; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1, [[DBG9:!dbg !.*]]
-; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64, [[DBG9]]
-; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1, [[DBG9]]
-; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3, [[DBG9]]
-; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]], [[DBG9]]
+; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1, !dbg [[DBG9:![0-9]+]]
+; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64, !dbg [[DBG9]]
+; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1, !dbg [[DBG9]]
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3, !dbg [[DBG9]]
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]], !dbg [[DBG9]]
; CHECK: vector.memcheck:
-; CHECK-NEXT: [[TMP3:%.*]] = sub i64 [[A1]], [[B2]], [[DBG9]]
-; CHECK-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP3]], 16, [[DBG9]]
-; CHECK-NEXT: br i1 [[DIFF_CHECK]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]], [[DBG9]]
+; CHECK-NEXT: [[TMP3:%.*]] = sub i64 [[A1]], [[B2]], !dbg [[DBG9]]
+; CHECK-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP3]], 16, !dbg [[DBG9]]
+; CHECK-NEXT: br i1 [[DIFF_CHECK]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]], !dbg [[DBG9]]
; CHECK: vector.ph:
-; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], -4, [[DBG9]]
-; CHECK-NEXT: br label [[VECTOR_BODY:%.*]], [[DBG9]]
+; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], -4, !dbg [[DBG9]]
+; CHECK-NEXT: br label [[VECTOR_BODY:%.*]], !dbg [[DBG9]]
; CHECK: vector.body:
-; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ], [[DBG9]]
-; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[INDEX]], [[DBG9]]
-; CHECK-NEXT: [[TMP7:%.*]] = bitcast float* [[TMP6]] to <4 x float>*, [[DBG9]]
-; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP7]], align 4, [[DBG9]]
-; CHECK-NEXT: [[TMP8:%.*]] = fmul <4 x float> [[WIDE_LOAD]], <float 3.000000e+00, float 3.000000e+00, float 3.000000e+00, float 3.000000e+00>, [[DBG9]]
-; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDEX]], [[DBG9]]
-; CHECK-NEXT: [[TMP10:%.*]] = bitcast float* [[TMP9]] to <4 x float>*, [[DBG9]]
-; CHECK-NEXT: store <4 x float> [[TMP8]], <4 x float>* [[TMP10]], align 4, [[DBG9]]
-; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4, [[DBG9]]
-; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]], [[DBG9]]
-; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], [[DBG9]], [[LOOP15:!llvm.loop !.*]]
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ], !dbg [[DBG9]]
+; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[INDEX]], !dbg [[DBG9]]
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast float* [[TMP4]] to <4 x float>*, !dbg [[DBG9]]
+; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP5]], align 4, !dbg [[DBG9]]
+; CHECK-NEXT: [[TMP6:%.*]] = fmul <4 x float> [[WIDE_LOAD]], <float 3.000000e+00, float 3.000000e+00, float 3.000000e+00, float 3.000000e+00>, !dbg [[DBG9]]
+; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDEX]], !dbg [[DBG9]]
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast float* [[TMP7]] to <4 x float>*, !dbg [[DBG9]]
+; CHECK-NEXT: store <4 x float> [[TMP6]], <4 x float>* [[TMP8]], align 4, !dbg [[DBG9]]
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4, !dbg [[DBG9]]
+; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]], !dbg [[DBG9]]
+; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !dbg [[DBG9]], !llvm.loop [[LOOP10:![0-9]+]]
; CHECK: middle.block:
-; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]], [[DBG9]]
-; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]], [[DBG9]]
+; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]], !dbg [[DBG9]]
+; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]], !dbg [[DBG9]]
; CHECK: scalar.ph:
-; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ], [[DBG9]]
-; CHECK-NEXT: br label [[FOR_BODY:%.*]], [[DBG9]]
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ], !dbg [[DBG9]]
+; CHECK-NEXT: br label [[FOR_BODY:%.*]], !dbg [[DBG9]]
; CHECK: for.body:
-; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [[DBG9]]
-; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[INDVARS_IV]], [[DBG9]]
-; CHECK-NEXT: [[TMP12:%.*]] = load float, float* [[ARRAYIDX]], align 4, [[DBG9]]
-; CHECK-NEXT: [[MUL:%.*]] = fmul float [[TMP12]], 3.000000e+00, [[DBG9]]
-; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]], [[DBG9]]
-; CHECK-NEXT: store float [[MUL]], float* [[ARRAYIDX2]], align 4, [[DBG9]]
-; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1, [[DBG9]]
-; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32, [[DBG9]]
-; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]], [[DBG9]]
-; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], [[DBG9]], [[LOOP17:!llvm.loop !.*]]
+; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], !dbg [[DBG9]]
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[INDVARS_IV]], !dbg [[DBG9]]
+; CHECK-NEXT: [[TMP10:%.*]] = load float, float* [[ARRAYIDX]], align 4, !dbg [[DBG9]]
+; CHECK-NEXT: [[MUL:%.*]] = fmul float [[TMP10]], 3.000000e+00, !dbg [[DBG9]]
+; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]], !dbg [[DBG9]]
+; CHECK-NEXT: store float [[MUL]], float* [[ARRAYIDX2]], align 4, !dbg [[DBG9]]
+; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1, !dbg [[DBG9]]
+; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32, !dbg [[DBG9]]
+; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]], !dbg [[DBG9]]
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !dbg [[DBG9]], !llvm.loop [[LOOP12:![0-9]+]]
; CHECK: for.end.loopexit:
-; CHECK-NEXT: br label [[FOR_END]], [[DBG18:!dbg !.*]]
+; CHECK-NEXT: br label [[FOR_END]], !dbg [[DBG13:![0-9]+]]
; CHECK: for.end:
-; CHECK-NEXT: ret i32 undef, [[DBG18]]
+; CHECK-NEXT: ret i32 undef, !dbg [[DBG13]]
+;
+; FORCED_OPTSIZE-LABEL: @foo(
+; FORCED_OPTSIZE-NEXT: entry:
+; FORCED_OPTSIZE-NEXT: [[CMP6:%.*]] = icmp sgt i32 [[N:%.*]], 0, !dbg [[DBG4:![0-9]+]]
+; FORCED_OPTSIZE-NEXT: br i1 [[CMP6]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]], !dbg [[DBG4]]
+; FORCED_OPTSIZE: for.body.preheader:
+; FORCED_OPTSIZE-NEXT: br label [[FOR_BODY:%.*]], !dbg [[DBG9:![0-9]+]]
+; FORCED_OPTSIZE: for.body:
+; FORCED_OPTSIZE-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[FOR_BODY_PREHEADER]] ], !dbg [[DBG9]]
+; FORCED_OPTSIZE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[INDVARS_IV]], !dbg [[DBG9]]
+; FORCED_OPTSIZE-NEXT: [[TMP0:%.*]] = load float, float* [[ARRAYIDX]], align 4, !dbg [[DBG9]]
+; FORCED_OPTSIZE-NEXT: [[MUL:%.*]] = fmul float [[TMP0]], 3.000000e+00, !dbg [[DBG9]]
+; FORCED_OPTSIZE-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDVARS_IV]], !dbg [[DBG9]]
+; FORCED_OPTSIZE-NEXT: store float [[MUL]], float* [[ARRAYIDX2]], align 4, !dbg [[DBG9]]
+; FORCED_OPTSIZE-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1, !dbg [[DBG9]]
+; FORCED_OPTSIZE-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32, !dbg [[DBG9]]
+; FORCED_OPTSIZE-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]], !dbg [[DBG9]]
+; FORCED_OPTSIZE-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT:%.*]], label [[FOR_BODY]], !dbg [[DBG9]]
+; FORCED_OPTSIZE: for.end.loopexit:
+; FORCED_OPTSIZE-NEXT: br label [[FOR_END]], !dbg [[DBG10:![0-9]+]]
+; FORCED_OPTSIZE: for.end:
+; FORCED_OPTSIZE-NEXT: ret i32 undef, !dbg [[DBG10]]
;
entry:
%cmp6 = icmp sgt i32 %n, 0, !dbg !6
@@ -87,9 +109,87 @@ for.end: ; preds = %for.body, %entry
; Make sure that we try to vectorize loops with a runtime check if the
; dependency check fails.
-; CHECK-LABEL: test_runtime_check
-; CHECK: <4 x float>
define void @test_runtime_check(float* %a, float %b, i64 %offset, i64 %offset2, i64 %n) {
+; CHECK-LABEL: @test_runtime_check(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; CHECK: vector.memcheck:
+; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr float, float* [[A:%.*]], i64 [[OFFSET:%.*]]
+; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[N]], [[OFFSET]]
+; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr float, float* [[A]], i64 [[TMP0]]
+; CHECK-NEXT: [[SCEVGEP4:%.*]] = getelementptr float, float* [[A]], i64 [[OFFSET2:%.*]]
+; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[N]], [[OFFSET2]]
+; CHECK-NEXT: [[SCEVGEP6:%.*]] = getelementptr float, float* [[A]], i64 [[TMP1]]
+; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult float* [[SCEVGEP]], [[SCEVGEP6]]
+; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult float* [[SCEVGEP4]], [[SCEVGEP2]]
+; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK: vector.ph:
+; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -4
+; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[B:%.*]], i64 0
+; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
+; CHECK: vector.body:
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], [[OFFSET]]
+; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP2]]
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast float* [[TMP3]] to <4 x float>*
+; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 4, !alias.scope !14, !noalias !17
+; CHECK-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], [[OFFSET2]]
+; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP5]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast float* [[TMP6]] to <4 x float>*
+; CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP7]], align 4, !alias.scope !17
+; CHECK-NEXT: [[TMP8:%.*]] = fmul fast <4 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD8]]
+; CHECK-NEXT: [[TMP9:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[TMP8]]
+; CHECK-NEXT: [[TMP10:%.*]] = bitcast float* [[TMP3]] to <4 x float>*
+; CHECK-NEXT: store <4 x float> [[TMP9]], <4 x float>* [[TMP10]], align 4, !alias.scope !14, !noalias !17
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; CHECK: middle.block:
+; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; CHECK-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK: scalar.ph:
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT: br label [[FOR_BODY:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]
+; CHECK-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM]]
+; CHECK-NEXT: [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
+; CHECK-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]
+; CHECK-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
+; CHECK-NEXT: [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
+; CHECK-NEXT: [[M:%.*]] = fmul fast float [[L2]], [[B]]
+; CHECK-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]
+; CHECK-NEXT: store float [[AD]], float* [[ARR_IDX]], align 4
+; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; CHECK: loopexit:
+; CHECK-NEXT: ret void
+;
+; FORCED_OPTSIZE-LABEL: @test_runtime_check(
+; FORCED_OPTSIZE-NEXT: entry:
+; FORCED_OPTSIZE-NEXT: br label [[FOR_BODY:%.*]]
+; FORCED_OPTSIZE: for.body:
+; FORCED_OPTSIZE-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; FORCED_OPTSIZE-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET:%.*]]
+; FORCED_OPTSIZE-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[IND_SUM]]
+; FORCED_OPTSIZE-NEXT: [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
+; FORCED_OPTSIZE-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2:%.*]]
+; FORCED_OPTSIZE-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
+; FORCED_OPTSIZE-NEXT: [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
+; FORCED_OPTSIZE-NEXT: [[M:%.*]] = fmul fast float [[B:%.*]], [[L2]]
+; FORCED_OPTSIZE-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]
+; FORCED_OPTSIZE-NEXT: store float [[AD]], float* [[ARR_IDX]], align 4
+; FORCED_OPTSIZE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; FORCED_OPTSIZE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N:%.*]]
+; FORCED_OPTSIZE-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT:%.*]], label [[FOR_BODY]]
+; FORCED_OPTSIZE: loopexit:
+; FORCED_OPTSIZE-NEXT: ret void
+;
entry:
br label %for.body
@@ -122,10 +222,60 @@ loopexit:
; c[i] = c[i-1] + b;
; }
; }
-;
-; CHECK-LABEL: test_runtime_check2
-; CHECK-NOT: <4 x float>
+
define void @test_runtime_check2(float* %a, float %b, i64 %offset, i64 %offset2, i64 %n, float* %c) {
+; CHECK-LABEL: @test_runtime_check2(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br label [[FOR_BODY:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET:%.*]]
+; CHECK-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[IND_SUM]]
+; CHECK-NEXT: [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
+; CHECK-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2:%.*]]
+; CHECK-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
+; CHECK-NEXT: [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
+; CHECK-NEXT: [[M:%.*]] = fmul fast float [[L2]], [[B:%.*]]
+; CHECK-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]
+; CHECK-NEXT: store float [[AD]], float* [[ARR_IDX]], align 4
+; CHECK-NEXT: [[C_IND:%.*]] = add nsw i64 [[IV]], -1
+; CHECK-NEXT: [[C_IDX:%.*]] = getelementptr inbounds float, float* [[C:%.*]], i64 [[C_IND]]
+; CHECK-NEXT: [[LC:%.*]] = load float, float* [[C_IDX]], align 4
+; CHECK-NEXT: [[VC:%.*]] = fadd float [[LC]], 1.000000e+00
+; CHECK-NEXT: [[C_IDX2:%.*]] = getelementptr inbounds float, float* [[C]], i64 [[IV]]
+; CHECK-NEXT: store float [[VC]], float* [[C_IDX2]], align 4
+; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N:%.*]]
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT:%.*]], label [[FOR_BODY]]
+; CHECK: loopexit:
+; CHECK-NEXT: ret void
+;
+; FORCED_OPTSIZE-LABEL: @test_runtime_check2(
+; FORCED_OPTSIZE-NEXT: entry:
+; FORCED_OPTSIZE-NEXT: br label [[FOR_BODY:%.*]]
+; FORCED_OPTSIZE: for.body:
+; FORCED_OPTSIZE-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; FORCED_OPTSIZE-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET:%.*]]
+; FORCED_OPTSIZE-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[IND_SUM]]
+; FORCED_OPTSIZE-NEXT: [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
+; FORCED_OPTSIZE-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2:%.*]]
+; FORCED_OPTSIZE-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
+; FORCED_OPTSIZE-NEXT: [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
+; FORCED_OPTSIZE-NEXT: [[M:%.*]] = fmul fast float [[B:%.*]], [[L2]]
+; FORCED_OPTSIZE-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]
+; FORCED_OPTSIZE-NEXT: store float [[AD]], float* [[ARR_IDX]], align 4
+; FORCED_OPTSIZE-NEXT: [[C_IND:%.*]] = add i64 [[IV]], -1
+; FORCED_OPTSIZE-NEXT: [[C_IDX:%.*]] = getelementptr inbounds float, float* [[C:%.*]], i64 [[C_IND]]
+; FORCED_OPTSIZE-NEXT: [[LC:%.*]] = load float, float* [[C_IDX]], align 4
+; FORCED_OPTSIZE-NEXT: [[VC:%.*]] = fadd float [[LC]], 1.000000e+00
+; FORCED_OPTSIZE-NEXT: [[C_IDX2:%.*]] = getelementptr inbounds float, float* [[C]], i64 [[IV]]
+; FORCED_OPTSIZE-NEXT: store float [[VC]], float* [[C_IDX2]], align 4
+; FORCED_OPTSIZE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; FORCED_OPTSIZE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N:%.*]]
+; FORCED_OPTSIZE-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT:%.*]], label [[FOR_BODY]]
+; FORCED_OPTSIZE: loopexit:
+; FORCED_OPTSIZE-NEXT: ret void
+;
entry:
br label %for.body
@@ -157,10 +307,23 @@ loopexit:
; CHECK: !9 = !DILocation(line: 101, column: 1, scope: !{{.*}})
define dso_local void @forced_optsize(i64* noalias nocapture readonly %x_p, i64* noalias nocapture readonly %y_p, i64* noalias nocapture %z_p) minsize optsize {
-;
-; FORCED_OPTSIZE: remark: <unknown>:0:0: loop not vectorized: runtime pointer checks needed. Enable vectorization of this loop with '#pragma clang loop vectorize(enable)' when compiling with -Os/-Oz
; FORCED_OPTSIZE-LABEL: @forced_optsize(
-; FORCED_OPTSIZE-NOT: vector.body:
+; FORCED_OPTSIZE-NEXT: entry:
+; FORCED_OPTSIZE-NEXT: br label [[FOR_BODY:%.*]]
+; FORCED_OPTSIZE: for.cond.cleanup:
+; FORCED_OPTSIZE-NEXT: ret void
+; FORCED_OPTSIZE: for.body:
+; FORCED_OPTSIZE-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; FORCED_OPTSIZE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i64, i64* [[X_P:%.*]], i64 [[INDVARS_IV]]
+; FORCED_OPTSIZE-NEXT: [[TMP0:%.*]] = load i64, i64* [[ARRAYIDX]], align 8
+; FORCED_OPTSIZE-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i64, i64* [[Y_P:%.*]], i64 [[INDVARS_IV]]
+; FORCED_OPTSIZE-NEXT: [[TMP1:%.*]] = load i64, i64* [[ARRAYIDX2]], align 8
+; FORCED_OPTSIZE-NEXT: [[ADD:%.*]] = add nsw i64 [[TMP1]], [[TMP0]]
+; FORCED_OPTSIZE-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i64, i64* [[Z_P:%.*]], i64 [[INDVARS_IV]]
+; FORCED_OPTSIZE-NEXT: store i64 [[ADD]], i64* [[ARRAYIDX4]], align 8
+; FORCED_OPTSIZE-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; FORCED_OPTSIZE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 128
+; FORCED_OPTSIZE-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
;
entry:
br label %for.body
@@ -195,9 +358,9 @@ for.body:
!7 = !DILocation(line: 101, column: 1, scope: !5)
!8 = !DILocation(line: 102, column: 1, scope: !5)
!9 = distinct !DICompileUnit(language: DW_LANG_C99, producer: "clang",
- file: !10,
- isOptimized: true, flags: "-O2",
- splitDebugFilename: "abc.debug", emissionKind: 2)
+ file: !10,
+ isOptimized: true, flags: "-O2",
+ splitDebugFilename: "abc.debug", emissionKind: 2)
!10 = !DIFile(filename: "path/to/file", directory: "/path/to/dir")
!11 = !{i32 2, !"Debug Info Version", i32 3}
!12 = distinct !{!12, !13, !14}
More information about the llvm-commits
mailing list