1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s
3
4define void @foo(<3 x float> %in, <4 x i8>* nocapture %out) nounwind {
5; CHECK-LABEL: foo:
6; CHECK:       # BB#0:
7; CHECK-NEXT:    cvttps2dq %xmm0, %xmm0
8; CHECK-NEXT:    movl $255, %eax
9; CHECK-NEXT:    pinsrd $3, %eax, %xmm0
10; CHECK-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
11; CHECK-NEXT:    movd %xmm0, (%rdi)
12; CHECK-NEXT:    retq
13  %t0 = fptoui <3 x float> %in to <3 x i8>
14  %t1 = shufflevector <3 x i8> %t0, <3 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 undef>
15  %t2 = insertelement <4 x i8> %t1, i8 -1, i32 3
16  store <4 x i8> %t2, <4 x i8>* %out, align 4
17  ret void
18}
19
20; Verify that the DAGCombiner doesn't wrongly fold a build_vector into a
21; blend with a zero vector if the build_vector contains negative zero.
22
23define <4 x float> @test_negative_zero_1(<4 x float> %A) {
24; CHECK-LABEL: test_negative_zero_1:
25; CHECK:       # BB#0: # %entry
26; CHECK-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2],zero
27; CHECK-NEXT:    retq
28entry:
29  %0 = extractelement <4 x float> %A, i32 0
30  %1 = insertelement <4 x float> undef, float %0, i32 0
31  %2 = insertelement <4 x float> %1, float -0.0, i32 1
32  %3 = extractelement <4 x float> %A, i32 2
33  %4 = insertelement <4 x float> %2, float %3, i32 2
34  %5 = insertelement <4 x float> %4, float 0.0, i32 3
35  ret <4 x float> %5
36}
37
38define <2 x double> @test_negative_zero_2(<2 x double> %A) {
39; CHECK-LABEL: test_negative_zero_2:
40; CHECK:       # BB#0: # %entry
41; CHECK-NEXT:    movhpd {{.*#+}} xmm0 = xmm0[0],mem[0]
42; CHECK-NEXT:    retq
43entry:
44  %0 = extractelement <2 x double> %A, i32 0
45  %1 = insertelement <2 x double> undef, double %0, i32 0
46  %2 = insertelement <2 x double> %1, double -0.0, i32 1
47  ret <2 x double> %2
48}
49
50define <4 x float> @test_buildvector_v4f32_register(float %f0, float %f1, float %f2, float %f3) {
51; CHECK-LABEL: test_buildvector_v4f32_register:
52; CHECK:       # BB#0:
53; CHECK-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
54; CHECK-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1],xmm2[0],xmm0[3]
55; CHECK-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[0]
56; CHECK-NEXT:    retq
57  %ins0 = insertelement <4 x float> undef, float %f0, i32 0
58  %ins1 = insertelement <4 x float> %ins0, float %f1, i32 1
59  %ins2 = insertelement <4 x float> %ins1, float %f2, i32 2
60  %ins3 = insertelement <4 x float> %ins2, float %f3, i32 3
61  ret <4 x float> %ins3
62}
63
64define <4 x float> @test_buildvector_v4f32_load(float* %p0, float* %p1, float* %p2, float* %p3) {
65; CHECK-LABEL: test_buildvector_v4f32_load:
66; CHECK:       # BB#0:
67; CHECK-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
68; CHECK-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
69; CHECK-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
70; CHECK-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
71; CHECK-NEXT:    retq
72  %f0 = load float, float* %p0, align 4
73  %f1 = load float, float* %p1, align 4
74  %f2 = load float, float* %p2, align 4
75  %f3 = load float, float* %p3, align 4
76  %ins0 = insertelement <4 x float> undef, float %f0, i32 0
77  %ins1 = insertelement <4 x float> %ins0, float %f1, i32 1
78  %ins2 = insertelement <4 x float> %ins1, float %f2, i32 2
79  %ins3 = insertelement <4 x float> %ins2, float %f3, i32 3
80  ret <4 x float> %ins3
81}
82
83define <4 x float> @test_buildvector_v4f32_partial_load(float %f0, float %f1, float %f2, float* %p3) {
84; CHECK-LABEL: test_buildvector_v4f32_partial_load:
85; CHECK:       # BB#0:
86; CHECK-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
87; CHECK-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1],xmm2[0],xmm0[3]
88; CHECK-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
89; CHECK-NEXT:    retq
90  %f3 = load float, float* %p3, align 4
91  %ins0 = insertelement <4 x float> undef, float %f0, i32 0
92  %ins1 = insertelement <4 x float> %ins0, float %f1, i32 1
93  %ins2 = insertelement <4 x float> %ins1, float %f2, i32 2
94  %ins3 = insertelement <4 x float> %ins2, float %f3, i32 3
95  ret <4 x float> %ins3
96}
97