1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -passes=slp-vectorizer -S -mtriple=x86_64-unknown -mcpu=corei7-avx | FileCheck %s 3 4%struct.complex = type { float, float } 5 6define void @foo (ptr %A, ptr %B, ptr %Result) { 7; CHECK-LABEL: @foo( 8; CHECK-NEXT: entry: 9; CHECK-NEXT: [[TMP0:%.*]] = add i64 256, 0 10; CHECK-NEXT: br label [[LOOP:%.*]] 11; CHECK: loop: 12; CHECK-NEXT: [[TMP1:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[TMP18:%.*]], [[LOOP]] ] 13; CHECK-NEXT: [[TMP2:%.*]] = phi <2 x float> [ zeroinitializer, [[ENTRY]] ], [ [[TMP20:%.*]], [[LOOP]] ] 14; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX:%.*]], ptr [[A:%.*]], i64 [[TMP1]], i32 0 15; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX]], ptr [[B:%.*]], i64 [[TMP1]], i32 0 16; CHECK-NEXT: [[TMP5:%.*]] = load float, ptr [[TMP4]], align 4 17; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX]], ptr [[B]], i64 [[TMP1]], i32 1 18; CHECK-NEXT: [[TMP7:%.*]] = load float, ptr [[TMP6]], align 4 19; CHECK-NEXT: [[TMP8:%.*]] = load <2 x float>, ptr [[TMP3]], align 4 20; CHECK-NEXT: [[TMP9:%.*]] = insertelement <2 x float> poison, float [[TMP5]], i32 0 21; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <2 x float> [[TMP9]], <2 x float> poison, <2 x i32> zeroinitializer 22; CHECK-NEXT: [[TMP11:%.*]] = fmul <2 x float> [[TMP8]], [[TMP10]] 23; CHECK-NEXT: [[TMP12:%.*]] = insertelement <2 x float> poison, float [[TMP7]], i32 0 24; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <2 x float> [[TMP12]], <2 x float> poison, <2 x i32> zeroinitializer 25; CHECK-NEXT: [[TMP14:%.*]] = fmul <2 x float> [[TMP8]], [[TMP13]] 26; CHECK-NEXT: [[TMP15:%.*]] = shufflevector <2 x float> [[TMP14]], <2 x float> poison, <2 x i32> <i32 1, i32 0> 27; CHECK-NEXT: [[TMP16:%.*]] = fsub <2 x float> [[TMP11]], [[TMP15]] 28; CHECK-NEXT: [[TMP17:%.*]] = fadd <2 x float> [[TMP11]], [[TMP15]] 29; CHECK-NEXT: [[TMP21:%.*]] = shufflevector <2 x float> [[TMP16]], <2 x float> [[TMP17]], <2 x i32> <i32 0, i32 3> 30; CHECK-NEXT: [[TMP20]] = fadd <2 x float> [[TMP2]], [[TMP21]] 31; CHECK-NEXT: [[TMP18]] = add nuw nsw i64 [[TMP1]], 1 32; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i64 [[TMP18]], [[TMP0]] 33; CHECK-NEXT: br i1 [[TMP19]], label [[EXIT:%.*]], label [[LOOP]] 34; CHECK: exit: 35; CHECK-NEXT: store <2 x float> [[TMP20]], ptr [[RESULT:%.*]], align 4 36; CHECK-NEXT: ret void 37; 38entry: 39 %0 = add i64 256, 0 40 br label %loop 41 42loop: 43 %1 = phi i64 [ 0, %entry ], [ %20, %loop ] 44 %2 = phi float [ 0.000000e+00, %entry ], [ %19, %loop ] 45 %3 = phi float [ 0.000000e+00, %entry ], [ %18, %loop ] 46 %4 = getelementptr inbounds %"struct.complex", ptr %A, i64 %1, i32 0 47 %5 = load float, ptr %4, align 4 48 %6 = getelementptr inbounds %"struct.complex", ptr %A, i64 %1, i32 1 49 %7 = load float, ptr %6, align 4 50 %8 = getelementptr inbounds %"struct.complex", ptr %B, i64 %1, i32 0 51 %9 = load float, ptr %8, align 4 52 %10 = getelementptr inbounds %"struct.complex", ptr %B, i64 %1, i32 1 53 %11 = load float, ptr %10, align 4 54 %12 = fmul float %5, %9 55 %13 = fmul float %7, %11 56 %14 = fsub float %12, %13 57 %15 = fmul float %7, %9 58 %16 = fmul float %5, %11 59 %17 = fadd float %15, %16 60 %18 = fadd float %3, %14 61 %19 = fadd float %2, %17 62 %20 = add nuw nsw i64 %1, 1 63 %21 = icmp eq i64 %20, %0 64 br i1 %21, label %exit, label %loop 65 66exit: 67 store float %18, ptr %Result, align 4 68 %22 = getelementptr inbounds %"struct.complex", ptr %Result, i32 0, i32 1 69 store float %19, ptr %22, align 4 70 ret void 71} 72 73