1; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx -mattr=+avx | FileCheck %s
2
3@x = common global <8 x float> zeroinitializer, align 32
4@y = common global <4 x double> zeroinitializer, align 32
5@z = common global <4 x float> zeroinitializer, align 16
6
7define void @zero128() nounwind ssp {
8entry:
9  ; CHECK: vpxor
10  ; CHECK: vmovaps
11  store <4 x float> zeroinitializer, <4 x float>* @z, align 16
12  ret void
13}
14
15define void @zero256() nounwind ssp {
16entry:
17  ; CHECK: vxorps
18  ; CHECK: vmovaps
19  ; CHECK: vmovaps
20  store <8 x float> zeroinitializer, <8 x float>* @x, align 32
21  store <4 x double> zeroinitializer, <4 x double>* @y, align 32
22  ret void
23}
24
25; CHECK: vpcmpeqd
26; CHECK: vinsertf128 $1
27define void @ones([0 x float]* nocapture %RET, [0 x float]* nocapture %aFOO) nounwind {
28allocas:
29  %ptr2vec615 = bitcast [0 x float]* %RET to <8 x float>*
30  store <8 x float> <float 0xFFFFFFFFE0000000, float 0xFFFFFFFFE0000000, float
310xFFFFFFFFE0000000, float 0xFFFFFFFFE0000000, float 0xFFFFFFFFE0000000, float
320xFFFFFFFFE0000000, float 0xFFFFFFFFE0000000, float 0xFFFFFFFFE0000000>, <8 x
33float>* %ptr2vec615, align 32
34  ret void
35}
36
37; CHECK: vpcmpeqd
38; CHECK: vinsertf128 $1
39define void @ones2([0 x i32]* nocapture %RET, [0 x i32]* nocapture %aFOO) nounwind {
40allocas:
41  %ptr2vec615 = bitcast [0 x i32]* %RET to <8 x i32>*
42  store <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, <8 x i32>* %ptr2vec615, align 32
43  ret void
44}
45
46;;; Just make sure this doesn't crash
47; CHECK: _ISelCrash
48define <4 x i64> @ISelCrash(<4 x i64> %a) nounwind uwtable readnone ssp {
49entry:
50  %shuffle = shufflevector <4 x i64> %a, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 4>
51  ret <4 x i64> %shuffle
52}
53
54;;;
55;;; Check that some 256-bit vectors are xformed into 128 ops
56; CHECK: _A
57; CHECK: vshufpd $1
58; CHECK-NEXT: vextractf128 $1
59; CHECK-NEXT: vshufpd $1
60; CHECK-NEXT: vinsertf128 $1
61define <4 x i64> @A(<4 x i64> %a, <4 x i64> %b) nounwind uwtable readnone ssp {
62entry:
63  %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 0, i32 7, i32 6>
64  ret <4 x i64> %shuffle
65}
66
67; CHECK: _B
68; CHECK: vshufpd $1, %ymm
69define <4 x i64> @B(<4 x i64> %a, <4 x i64> %b) nounwind uwtable readnone ssp {
70entry:
71  %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 undef, i32 undef, i32 6>
72  ret <4 x i64> %shuffle
73}
74
75; CHECK: movlhps
76; CHECK-NEXT: vextractf128  $1
77; CHECK-NEXT: movlhps
78; CHECK-NEXT: vinsertf128 $1
79define <4 x i64> @C(<4 x i64> %a, <4 x i64> %b) nounwind uwtable readnone ssp {
80entry:
81  %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 undef, i32 0, i32 undef, i32 6>
82  ret <4 x i64> %shuffle
83}
84
85; CHECK: vpshufd $-96
86; CHECK: vpshufd $-6
87; CHECK: vinsertf128 $1
88define <8 x i32> @D(<8 x i32> %a, <8 x i32> %b) nounwind uwtable readnone ssp {
89entry:
90  %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 10, i32 10, i32 11, i32 11>
91  ret <8 x i32> %shuffle
92}
93
94;;; Don't crash on movd
95; CHECK: _VMOVZQI2PQI
96; CHECK: vmovd (%
97define <8 x i32> @VMOVZQI2PQI([0 x float]* nocapture %aFOO) nounwind {
98allocas:
99  %ptrcast.i33.i = bitcast [0 x float]* %aFOO to i32*
100  %val.i34.i = load i32* %ptrcast.i33.i, align 4
101  %ptroffset.i22.i992 = getelementptr [0 x float]* %aFOO, i64 0, i64 1
102  %ptrcast.i23.i = bitcast float* %ptroffset.i22.i992 to i32*
103  %val.i24.i = load i32* %ptrcast.i23.i, align 4
104  %updatedret.i30.i = insertelement <8 x i32> undef, i32 %val.i34.i, i32 1
105  ret <8 x i32> %updatedret.i30.i
106}
107
108