1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2 -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=CHECK,NON-POW2 %s 3; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2=false -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=CHECK,POW2-ONLY %s 4 5define void @vec3_vectorize_call(ptr %Colour, float %0) { 6; NON-POW2-LABEL: @vec3_vectorize_call( 7; NON-POW2-NEXT: entry: 8; NON-POW2-NEXT: [[TMP1:%.*]] = load <2 x float>, ptr [[COLOUR:%.*]], align 4 9; NON-POW2-NEXT: [[TMP2:%.*]] = insertelement <3 x float> poison, float [[TMP0:%.*]], i32 2 10; NON-POW2-NEXT: [[TMP4:%.*]] = call <3 x float> @llvm.vector.insert.v3f32.v2f32(<3 x float> [[TMP2]], <2 x float> [[TMP1]], i64 0) 11; NON-POW2-NEXT: [[TMP5:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP4]], <3 x float> zeroinitializer, <3 x float> zeroinitializer) 12; NON-POW2-NEXT: store <3 x float> [[TMP5]], ptr [[COLOUR]], align 4 13; NON-POW2-NEXT: ret void 14; 15; POW2-ONLY-LABEL: @vec3_vectorize_call( 16; POW2-ONLY-NEXT: entry: 17; POW2-ONLY-NEXT: [[TMP1:%.*]] = load <2 x float>, ptr [[COLOUR:%.*]], align 4 18; POW2-ONLY-NEXT: [[TMP2:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP1]], <2 x float> zeroinitializer, <2 x float> zeroinitializer) 19; POW2-ONLY-NEXT: store <2 x float> [[TMP2]], ptr [[COLOUR]], align 4 20; POW2-ONLY-NEXT: [[ARRAYIDX99_I1:%.*]] = getelementptr float, ptr [[COLOUR]], i64 2 21; POW2-ONLY-NEXT: [[TMP3:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0:%.*]], float 0.000000e+00, float 0.000000e+00) 22; POW2-ONLY-NEXT: store float [[TMP3]], ptr [[ARRAYIDX99_I1]], align 4 23; POW2-ONLY-NEXT: ret void 24; 25entry: 26 %1 = load float, ptr %Colour, align 4 27 %2 = call float @llvm.fmuladd.f32(float %1, float 0.000000e+00, float 0.000000e+00) 28 store float %2, ptr %Colour, align 4 29 %arrayidx91.i = getelementptr float, ptr %Colour, i64 1 30 %3 = load float, ptr %arrayidx91.i, align 4 31 %4 = call float @llvm.fmuladd.f32(float %3, float 0.000000e+00, float 0.000000e+00) 32 store float %4, ptr %arrayidx91.i, align 4 33 %arrayidx99.i1 = getelementptr float, ptr %Colour, i64 2 34 %5 = call float @llvm.fmuladd.f32(float %0, float 0.000000e+00, float 0.000000e+00) 35 store float %5, ptr %arrayidx99.i1, align 4 36 ret void 37} 38 39define void @vec3_fmuladd_64(ptr %Colour, double %0) { 40; CHECK-LABEL: @vec3_fmuladd_64( 41; CHECK-NEXT: entry: 42; CHECK-NEXT: [[ARRAYIDX80:%.*]] = getelementptr float, ptr [[COLOUR:%.*]], i64 2 43; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x double> poison, double [[TMP0:%.*]], i32 0 44; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> poison, <2 x i32> zeroinitializer 45; CHECK-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[TMP2]], <2 x double> zeroinitializer, <2 x double> zeroinitializer) 46; CHECK-NEXT: [[TMP4:%.*]] = fptrunc <2 x double> [[TMP3]] to <2 x float> 47; CHECK-NEXT: store <2 x float> [[TMP4]], ptr [[COLOUR]], align 4 48; CHECK-NEXT: [[TMP5:%.*]] = call double @llvm.fmuladd.f64(double [[TMP0]], double 0.000000e+00, double 0.000000e+00) 49; CHECK-NEXT: [[CONV82:%.*]] = fptrunc double [[TMP5]] to float 50; CHECK-NEXT: store float [[CONV82]], ptr [[ARRAYIDX80]], align 4 51; CHECK-NEXT: ret void 52; 53entry: 54 %arrayidx72 = getelementptr float, ptr %Colour, i64 1 55 %arrayidx80 = getelementptr float, ptr %Colour, i64 2 56 %1 = call double @llvm.fmuladd.f64(double %0, double 0.000000e+00, double 0.000000e+00) 57 %conv66 = fptrunc double %1 to float 58 store float %conv66, ptr %Colour, align 4 59 %2 = call double @llvm.fmuladd.f64(double %0, double 0.000000e+00, double 0.000000e+00) 60 %conv74 = fptrunc double %2 to float 61 store float %conv74, ptr %arrayidx72, align 4 62 %3 = call double @llvm.fmuladd.f64(double %0, double 0.000000e+00, double 0.000000e+00) 63 %conv82 = fptrunc double %3 to float 64 store float %conv82, ptr %arrayidx80, align 4 65 ret void 66} 67 68declare float @llvm.fmuladd.f32(float, float, float) 69 70declare double @llvm.fmuladd.f64(double, double, double) 71