1 ; RUN: llc < %s -mtriple=x86_64-unknown -mcpu=corei7 | FileCheck %s
4 define <2 x double> @test_x86_sse41_blend_pd(<2 x double> %a0, <2 x double> %a1) {
5 %1 = call <2 x double> @llvm.x86.sse41.blendpd(<2 x double> %a0, <2 x double> %a1, i32 0)
8 ; CHECK-LABEL: test_x86_sse41_blend_pd
13 define <4 x float> @test_x86_sse41_blend_ps(<4 x float> %a0, <4 x float> %a1) {
14 %1 = call <4 x float> @llvm.x86.sse41.blendps(<4 x float> %a0, <4 x float> %a1, i32 0)
17 ; CHECK-LABEL: test_x86_sse41_blend_ps
22 define <8 x i16> @test_x86_sse41_pblend_w(<8 x i16> %a0, <8 x i16> %a1) {
23 %1 = call <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16> %a0, <8 x i16> %a1, i32 0)
26 ; CHECK-LABEL: test_x86_sse41_pblend_w
31 define <2 x double> @test2_x86_sse41_blend_pd(<2 x double> %a0, <2 x double> %a1) {
32 %1 = call <2 x double> @llvm.x86.sse41.blendpd(<2 x double> %a0, <2 x double> %a1, i32 -1)
35 ; CHECK-LABEL: test2_x86_sse41_blend_pd
37 ; CHECK: movaps %xmm1, %xmm0
41 define <4 x float> @test2_x86_sse41_blend_ps(<4 x float> %a0, <4 x float> %a1) {
42 %1 = call <4 x float> @llvm.x86.sse41.blendps(<4 x float> %a0, <4 x float> %a1, i32 -1)
45 ; CHECK-LABEL: test2_x86_sse41_blend_ps
47 ; CHECK: movaps %xmm1, %xmm0
51 define <8 x i16> @test2_x86_sse41_pblend_w(<8 x i16> %a0, <8 x i16> %a1) {
52 %1 = call <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16> %a0, <8 x i16> %a1, i32 -1)
55 ; CHECK-LABEL: test2_x86_sse41_pblend_w
57 ; CHECK: movaps %xmm1, %xmm0
61 define <2 x double> @test3_x86_sse41_blend_pd(<2 x double> %a0) {
62 %1 = call <2 x double> @llvm.x86.sse41.blendpd(<2 x double> %a0, <2 x double> %a0, i32 7)
65 ; CHECK-LABEL: test3_x86_sse41_blend_pd
70 define <4 x float> @test3_x86_sse41_blend_ps(<4 x float> %a0) {
71 %1 = call <4 x float> @llvm.x86.sse41.blendps(<4 x float> %a0, <4 x float> %a0, i32 7)
74 ; CHECK-LABEL: test3_x86_sse41_blend_ps
79 define <8 x i16> @test3_x86_sse41_pblend_w(<8 x i16> %a0) {
80 %1 = call <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16> %a0, <8 x i16> %a0, i32 7)
83 ; CHECK-LABEL: test3_x86_sse41_pblend_w
88 declare <2 x double> @llvm.x86.sse41.blendpd(<2 x double>, <2 x double>, i32)
89 declare <4 x float> @llvm.x86.sse41.blendps(<4 x float>, <4 x float>, i32)
90 declare <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16>, <8 x i16>, i32)