1; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512f,+avx512bw | FileCheck %s
2
3define i32 @f(<16 x float> %A, <16 x float> %AA, i8* %B, <8 x double> %C, <8 x double> %CC, <8 x i64> %E, <8 x i64> %EE, <16 x i32> %F, <16 x i32> %FF, <32 x i16> %G, <32 x i16> %GG, <64 x i8> %H, <64 x i8> %HH, i32 * %loadptr) {
4; CHECK: vmovntps %z
5  %v0 = load i32, i32* %loadptr, align 1
6  %cast = bitcast i8* %B to <16 x float>*
7  %A2 = fadd <16 x float> %A, %AA
8  store <16 x float> %A2, <16 x float>* %cast, align 64, !nontemporal !0
9  %v1 = load i32, i32* %loadptr, align 1
10; CHECK: vmovntdq %z
11  %cast1 = bitcast i8* %B to <8 x i64>*
12  %E2 = add <8 x i64> %E, %EE
13  store <8 x i64> %E2, <8 x i64>* %cast1, align 64, !nontemporal !0
14  %v2 = load i32, i32* %loadptr, align 1
15; CHECK: vmovntpd %z
16  %cast2 = bitcast i8* %B to <8 x double>*
17  %C2 = fadd <8 x double> %C, %CC
18  store <8 x double> %C2, <8 x double>* %cast2, align 64, !nontemporal !0
19  %v3 = load i32, i32* %loadptr, align 1
20; CHECK: vmovntdq %z
21  %cast3 = bitcast i8* %B to <16 x i32>*
22  %F2 = add <16 x i32> %F, %FF
23  store <16 x i32> %F2, <16 x i32>* %cast3, align 64, !nontemporal !0
24  %v4 = load i32, i32* %loadptr, align 1
25; CHECK: vmovntdq %z
26  %cast4 = bitcast i8* %B to <32 x i16>*
27  %G2 = add <32 x i16> %G, %GG
28  store <32 x i16> %G2, <32 x i16>* %cast4, align 64, !nontemporal !0
29  %v5 = load i32, i32* %loadptr, align 1
30; CHECK: vmovntdq %z
31  %cast5 = bitcast i8* %B to <64 x i8>*
32  %H2 = add <64 x i8> %H, %HH
33  store <64 x i8> %H2, <64 x i8>* %cast5, align 64, !nontemporal !0
34  %v6 = load i32, i32* %loadptr, align 1
35  %sum1 = add i32 %v0, %v1
36  %sum2 = add i32 %sum1, %v2
37  %sum3 = add i32 %sum2, %v3
38  %sum4 = add i32 %sum3, %v4
39  %sum5 = add i32 %sum4, %v5
40  %sum6 = add i32 %sum5, %v6
41  ret i32 %sum6
42}
43
44!0 = !{i32 1}
45