1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx | FileCheck %s
4 @v2_0 = dso_local global <2 x i32> zeroinitializer, align 8
6 define dso_local void @test() {
8 ; CHECK: # %bb.0: # %entry
9 ; CHECK-NEXT: pushq %rbp
10 ; CHECK-NEXT: .cfi_def_cfa_offset 16
11 ; CHECK-NEXT: .cfi_offset %rbp, -16
12 ; CHECK-NEXT: movq %rsp, %rbp
13 ; CHECK-NEXT: .cfi_def_cfa_register %rbp
14 ; CHECK-NEXT: andq $-32, %rsp
15 ; CHECK-NEXT: subq $64, %rsp
16 ; CHECK-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
17 ; CHECK-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1
18 ; CHECK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
19 ; CHECK-NEXT: vxorps %xmm2, %xmm2, %xmm2
20 ; CHECK-NEXT: vcmptrueps %ymm2, %ymm2, %ymm2
21 ; CHECK-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
22 ; CHECK-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
23 ; CHECK-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[0,0],ymm1[6,4],ymm0[4,4]
24 ; CHECK-NEXT: vmovaps %ymm0, (%rsp)
25 ; CHECK-NEXT: movq %rbp, %rsp
26 ; CHECK-NEXT: popq %rbp
27 ; CHECK-NEXT: .cfi_def_cfa %rsp, 8
28 ; CHECK-NEXT: vzeroupper
31 %v8_0 = alloca <8 x i32>, align 32
32 %v8_0.0.v8_0.0..sroa_cast = bitcast <8 x i32>* %v8_0 to i8*
33 %0 = load <2 x i32>, <2 x i32>* @v2_0, align 8
34 %shuffle = shufflevector <2 x i32> %0, <2 x i32> <i32 -1, i32 -1>, <8 x i32> <i32 1, i32 3, i32 0, i32 0, i32 3, i32 3, i32 2, i32 2>
35 store volatile <8 x i32> %shuffle, <8 x i32>* %v8_0, align 32