|
| 1 | +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4 |
| 2 | +; RUN: opt -S --passes=slp-vectorizer -mtriple=x86_64-linux-gnu < %s | FileCheck %s |
| 3 | + |
| 4 | +define void @test(ptr noalias %arg, ptr noalias %arg1, ptr %arg2) { |
| 5 | +; CHECK-LABEL: define void @test( |
| 6 | +; CHECK-SAME: ptr noalias [[ARG:%.*]], ptr noalias [[ARG1:%.*]], ptr [[ARG2:%.*]]) { |
| 7 | +; CHECK-NEXT: bb: |
| 8 | +; CHECK-NEXT: [[TMP_I_I:%.*]] = getelementptr i8, ptr [[ARG1]], i64 24 |
| 9 | +; CHECK-NEXT: [[TMP_I_I4:%.*]] = getelementptr i8, ptr [[ARG]], i64 24 |
| 10 | +; CHECK-NEXT: [[TMP0:%.*]] = load <4 x float>, ptr [[TMP_I_I]], align 8 |
| 11 | +; CHECK-NEXT: [[TMP1:%.*]] = extractelement <4 x float> [[TMP0]], i32 1 |
| 12 | +; CHECK-NEXT: store float [[TMP1]], ptr [[ARG2]], align 4 |
| 13 | +; CHECK-NEXT: [[TMP2:%.*]] = fcmp olt <4 x float> [[TMP0]], zeroinitializer |
| 14 | +; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> poison, <4 x i32> <i32 2, i32 3, i32 2, i32 3> |
| 15 | +; CHECK-NEXT: [[TMP4:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> poison, <4 x i32> <i32 0, i32 1, i32 0, i32 1> |
| 16 | +; CHECK-NEXT: [[TMP5:%.*]] = select <4 x i1> [[TMP2]], <4 x float> [[TMP3]], <4 x float> [[TMP4]] |
| 17 | +; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <4 x float> [[TMP5]], <4 x float> poison, <4 x i32> <i32 1, i32 2, i32 3, i32 0> |
| 18 | +; CHECK-NEXT: store <4 x float> [[TMP6]], ptr [[TMP_I_I4]], align 8 |
| 19 | +; CHECK-NEXT: ret void |
| 20 | +; |
| 21 | +bb: |
| 22 | + %tmp.i.i = getelementptr i8, ptr %arg1, i64 24 |
| 23 | + %tmp1.i.i = load float, ptr %tmp.i.i, align 8 |
| 24 | + %tmp.i.i2 = getelementptr i8, ptr %arg1, i64 32 |
| 25 | + %tmp1.i.i3 = load float, ptr %tmp.i.i2, align 8 |
| 26 | + %tmp1.i.i.i = fcmp olt float %tmp1.i.i3, 0.000000e+00 |
| 27 | + %tmp9 = select i1 %tmp1.i.i.i, float %tmp1.i.i3, float %tmp1.i.i |
| 28 | + %tmp.i.i4 = getelementptr i8, ptr %arg, i64 24 |
| 29 | + store float %tmp9, ptr %tmp.i.i4, align 8 |
| 30 | + %tmp1.i.i.i10 = fcmp olt float %tmp1.i.i, 0.000000e+00 |
| 31 | + %tmp13 = select i1 %tmp1.i.i.i10, float %tmp1.i.i3, float %tmp1.i.i |
| 32 | + %tmp.i.i12 = getelementptr i8, ptr %arg, i64 28 |
| 33 | + store float %tmp13, ptr %tmp.i.i12, align 4 |
| 34 | + %tmp.i.i13 = getelementptr i8, ptr %arg1, i64 28 |
| 35 | + %tmp1.i.i14 = load float, ptr %tmp.i.i13, align 4 |
| 36 | + %tmp.i.i15 = getelementptr i8, ptr %arg1, i64 36 |
| 37 | + %tmp1.i.i16 = load float, ptr %tmp.i.i15, align 4 |
| 38 | + %tmp1.i.i.i18 = fcmp olt float %tmp1.i.i16, 0.000000e+00 |
| 39 | + %tmp17 = select i1 %tmp1.i.i.i18, float %tmp1.i.i16, float %tmp1.i.i14 |
| 40 | + %tmp.i.i20 = getelementptr i8, ptr %arg, i64 32 |
| 41 | + store float %tmp17, ptr %tmp.i.i20, align 8 |
| 42 | + store float %tmp1.i.i14, ptr %arg2, align 4 |
| 43 | + %tmp1.i.i.i24 = fcmp olt float %tmp1.i.i14, 0.000000e+00 |
| 44 | + %tmp20 = select i1 %tmp1.i.i.i24, float %tmp1.i.i16, float %tmp1.i.i14 |
| 45 | + %tmp.i.i26 = getelementptr i8, ptr %arg, i64 36 |
| 46 | + store float %tmp20, ptr %tmp.i.i26, align 4 |
| 47 | + ret void |
| 48 | +} |
0 commit comments