1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt -passes=slp-vectorizer -S -mcpu=cascadelake -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s 3 4define void @foo(i1 %arg) { 5; CHECK-LABEL: @foo( 6; CHECK-NEXT: entry: 7; CHECK-NEXT: [[CONV:%.*]] = uitofp i16 0 to float 8; CHECK-NEXT: [[SUB:%.*]] = fsub float 6.553500e+04, 0.000000e+00 9; CHECK-NEXT: br label [[BB1:%.*]] 10; CHECK: bb1: 11; CHECK-NEXT: [[TMP0:%.*]] = insertelement <4 x float> <float poison, float poison, float 0.000000e+00, float 0.000000e+00>, float [[SUB]], i32 0 12; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x float> [[TMP0]], float [[CONV]], i32 1 13; CHECK-NEXT: br label [[BB2:%.*]] 14; CHECK: bb2: 15; CHECK-NEXT: [[TMP2:%.*]] = phi <4 x float> [ [[TMP1]], [[BB1]] ], [ [[TMP14:%.*]], [[BB3:%.*]] ] 16; CHECK-NEXT: [[TMP3:%.*]] = load double, ptr null, align 8 17; CHECK-NEXT: br i1 [[ARG:%.*]], label [[BB3]], label [[BB4:%.*]] 18; CHECK: bb4: 19; CHECK-NEXT: [[TMP4:%.*]] = fpext <4 x float> [[TMP2]] to <4 x double> 20; CHECK-NEXT: [[CONV2:%.*]] = uitofp i16 0 to double 21; CHECK-NEXT: [[TMP5:%.*]] = insertelement <2 x double> <double 0.000000e+00, double poison>, double [[TMP3]], i32 1 22; CHECK-NEXT: [[TMP6:%.*]] = insertelement <2 x double> <double 0.000000e+00, double poison>, double [[CONV2]], i32 1 23; CHECK-NEXT: [[TMP7:%.*]] = fsub <2 x double> [[TMP5]], [[TMP6]] 24; CHECK-NEXT: [[TMP8:%.*]] = fadd <2 x double> [[TMP5]], [[TMP6]] 25; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <2 x double> [[TMP7]], <2 x double> [[TMP8]], <2 x i32> <i32 0, i32 3> 26; CHECK-NEXT: [[TMP15:%.*]] = shufflevector <2 x double> [[TMP9]], <2 x double> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison> 27; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <4 x double> <double poison, double poison, double 0.000000e+00, double 0.000000e+00>, <4 x double> [[TMP15]], <4 x i32> <i32 4, i32 5, i32 2, i32 3> 28; CHECK-NEXT: [[TMP11:%.*]] = fcmp ogt <4 x double> [[TMP10]], [[TMP4]] 29; CHECK-NEXT: [[TMP12:%.*]] = fptrunc <4 x double> [[TMP10]] to <4 x float> 30; CHECK-NEXT: [[TMP13:%.*]] = select <4 x i1> [[TMP11]], <4 x float> [[TMP2]], <4 x float> [[TMP12]] 31; CHECK-NEXT: br label [[BB3]] 32; CHECK: bb3: 33; CHECK-NEXT: [[TMP14]] = phi <4 x float> [ [[TMP13]], [[BB4]] ], [ [[TMP2]], [[BB2]] ] 34; CHECK-NEXT: br label [[BB2]] 35; 36entry: 37 %conv = uitofp i16 zeroinitializer to float 38 %sub = fsub float 6.553500e+04, zeroinitializer 39 br label %bb1 40 41bb1: 42 br label %bb2 43 44bb2: 45 %0 = phi float [ %sub, %bb1 ], [ %9, %bb3 ] 46 %1 = phi float [ %conv, %bb1 ], [ %10, %bb3 ] 47 %2 = phi float [ zeroinitializer, %bb1 ], [ %11, %bb3 ] 48 %3 = phi float [ zeroinitializer, %bb1 ], [ %12, %bb3 ] 49 %4 = load double, ptr zeroinitializer, align 8 50 br i1 %arg, label %bb3, label %bb4 51 52bb4: 53 %ext = fpext float %3 to double 54 %cmp1 = fcmp ogt double zeroinitializer, %ext 55 %5 = fptrunc double zeroinitializer to float 56 %sel1 = select i1 %cmp1, float %3, float %5 57 %ext2 = fpext float %2 to double 58 %cmp2 = fcmp ogt double zeroinitializer, %ext2 59 %6 = fptrunc double zeroinitializer to float 60 %sel2 = select i1 %cmp2, float %2, float %6 61 %ext3 = fpext float %1 to double 62 %conv2 = uitofp i16 zeroinitializer to double 63 %add1 = fadd double %4, %conv2 64 %cmp3 = fcmp ogt double %add1, %ext3 65 %7 = fptrunc double %add1 to float 66 %sel3 = select i1 %cmp3, float %1, float %7 67 %ext4 = fpext float %0 to double 68 %sub1 = fsub double zeroinitializer, zeroinitializer 69 %cmp4 = fcmp ogt double %sub1, %ext4 70 %8 = fptrunc double %sub1 to float 71 %sel4 = select i1 %cmp4, float %0, float %8 72 br label %bb3 73 74bb3: 75 %9 = phi float [ %sel4, %bb4 ], [ %0, %bb2 ] 76 %10 = phi float [ %sel3, %bb4 ], [ %1, %bb2 ] 77 %11 = phi float [ %sel2, %bb4 ], [ %2, %bb2 ] 78 %12 = phi float [ %sel1, %bb4 ], [ %3, %bb2 ] 79 br label %bb2 80} 81