; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -slp-vectorizer -S < %s | FileCheck %s
target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2"
target triple = "x86_64-unknown-linux-gnu"
define void @"foo"(i8 addrspace(1)* %0, i8 addrspace(1)* %1) #0 {
; CHECK-LABEL: @foo(
; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x i8 addrspace(1)*> poison, i8 addrspace(1)* [[TMP0:%.*]], i32 0
; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <4 x i8 addrspace(1)*> [[TMP3]], <4 x i8 addrspace(1)*> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr i8, <4 x i8 addrspace(1)*> [[SHUFFLE]], <4 x i64> <i64 8, i64 12, i64 28, i64 24>
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i8, i8 addrspace(1)* [[TMP1:%.*]], i64 8
; CHECK-NEXT: [[TMP6:%.*]] = bitcast i8 addrspace(1)* [[TMP5]] to float addrspace(1)*
; CHECK-NEXT: [[TMP7:%.*]] = bitcast <4 x i8 addrspace(1)*> [[TMP4]] to <4 x float addrspace(1)*>
; CHECK-NEXT: [[TMP8:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p1f32(<4 x float addrspace(1)*> [[TMP7]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
; CHECK-NEXT: [[SHUFFLE1:%.*]] = shufflevector <4 x float> [[TMP8]], <4 x float> poison, <8 x i32> <i32 0, i32 3, i32 0, i32 3, i32 2, i32 1, i32 2, i32 1>
; CHECK-NEXT: [[TMP9:%.*]] = bitcast float addrspace(1)* [[TMP6]] to <8 x float> addrspace(1)*
; CHECK-NEXT: [[TMP10:%.*]] = load <8 x float>, <8 x float> addrspace(1)* [[TMP9]], align 4
; CHECK-NEXT: [[TMP11:%.*]] = fmul <8 x float> [[SHUFFLE1]], [[TMP10]]
; CHECK-NEXT: [[TMP12:%.*]] = fadd <8 x float> [[TMP11]], zeroinitializer
; CHECK-NEXT: [[SHUFFLE2:%.*]] = shufflevector <8 x float> [[TMP12]], <8 x float> poison, <8 x i32> <i32 0, i32 5, i32 2, i32 7, i32 4, i32 1, i32 6, i32 3>
; CHECK-NEXT: [[TMP13:%.*]] = extractelement <4 x float addrspace(1)*> [[TMP7]], i32 0
; CHECK-NEXT: [[TMP14:%.*]] = bitcast float addrspace(1)* [[TMP13]] to <8 x float> addrspace(1)*
; CHECK-NEXT: store <8 x float> [[SHUFFLE2]], <8 x float> addrspace(1)* [[TMP14]], align 4
; CHECK-NEXT: ret void
;
%3 = getelementptr inbounds i8, i8 addrspace(1)* %0, i64 8
%4 = bitcast i8 addrspace(1)* %3 to float addrspace(1)*
%5 = load float, float addrspace(1)* %4 , align 4
%6 = getelementptr inbounds i8, i8 addrspace(1)* %1, i64 8
%7 = bitcast i8 addrspace(1)* %6 to float addrspace(1)*
%8 = load float, float addrspace(1)* %7 , align 4
%9 = fmul float %5, %8
%10 = fadd float %9, 0.000000e+00
%11 = getelementptr inbounds i8, i8 addrspace(1)* %0, i64 12
%12 = bitcast i8 addrspace(1)* %11 to float addrspace(1)*
%13 = load float, float addrspace(1)* %12 , align 4
%14 = getelementptr inbounds i8, i8 addrspace(1)* %1, i64 28
%15 = bitcast i8 addrspace(1)* %14 to float addrspace(1)*
%16 = load float, float addrspace(1)* %15 , align 4
%17 = fmul float %13, %16
%18 = fadd float %17, 0.000000e+00
%19 = getelementptr inbounds i8, i8 addrspace(1)* %1, i64 16
%20 = bitcast i8 addrspace(1)* %19 to float addrspace(1)*
%21 = load float, float addrspace(1)* %20 , align 4
%22 = fmul float %5, %21
%23 = fadd float %22, 0.000000e+00
%24 = getelementptr inbounds i8, i8 addrspace(1)* %1, i64 36
%25 = bitcast i8 addrspace(1)* %24 to float addrspace(1)*
%26 = load float, float addrspace(1)* %25 , align 4
%27 = fmul float %13, %26
%28 = fadd float %27, 0.000000e+00
%29 = getelementptr inbounds i8, i8 addrspace(1)* %0, i64 28
%30 = bitcast i8 addrspace(1)* %29 to float addrspace(1)*
%31 = load float, float addrspace(1)* %30 , align 4
%32 = getelementptr inbounds i8, i8 addrspace(1)* %1, i64 24
%33 = bitcast i8 addrspace(1)* %32 to float addrspace(1)*
%34 = load float, float addrspace(1)* %33 , align 4
%35 = fmul float %31, %34
%36 = fadd float %35, 0.000000e+00
%37 = getelementptr inbounds i8, i8 addrspace(1)* %0, i64 24
%38 = bitcast i8 addrspace(1)* %37 to float addrspace(1)*
%39 = load float, float addrspace(1)* %38 , align 4
%40 = getelementptr inbounds i8, i8 addrspace(1)* %1, i64 12
%41 = bitcast i8 addrspace(1)* %40 to float addrspace(1)*
%42 = load float, float addrspace(1)* %41 , align 4
%43 = fmul float %39, %42
%44 = fadd float %43, 0.000000e+00
%45 = getelementptr inbounds i8, i8 addrspace(1)* %1, i64 32
%46 = bitcast i8 addrspace(1)* %45 to float addrspace(1)*
%47 = load float, float addrspace(1)* %46 , align 4
%48 = fmul float %31, %47
%49 = fadd float %48, 0.000000e+00
%50 = getelementptr inbounds i8, i8 addrspace(1)* %1, i64 20
%51 = bitcast i8 addrspace(1)* %50 to float addrspace(1)*
%52 = load float, float addrspace(1)* %51 , align 4
%53 = fmul float %39, %52
%54 = fadd float %53, 0.000000e+00
store float %10, float addrspace(1)* %4 , align 4
store float %18, float addrspace(1)* %12 , align 4
%55 = getelementptr inbounds i8, i8 addrspace(1)* %0, i64 16
%56 = bitcast i8 addrspace(1)* %55 to float addrspace(1)*
store float %23, float addrspace(1)* %56 , align 4
%57 = getelementptr inbounds i8, i8 addrspace(1)* %0, i64 20
%58 = bitcast i8 addrspace(1)* %57 to float addrspace(1)*
store float %28, float addrspace(1)* %58 , align 4
store float %36, float addrspace(1)* %38 , align 4
store float %44, float addrspace(1)* %30 , align 4
%59 = getelementptr inbounds i8, i8 addrspace(1)* %0, i64 32
%60 = bitcast i8 addrspace(1)* %59 to float addrspace(1)*
store float %49, float addrspace(1)* %60 , align 4
%61 = getelementptr inbounds i8, i8 addrspace(1)* %0, i64 36
%62 = bitcast i8 addrspace(1)* %61 to float addrspace(1)*
store float %54, float addrspace(1)* %62 , align 4
ret void
}
attributes #0 = { "target-cpu"="skylake" }