1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx,+xop | FileCheck %s --check-prefix=X32
3; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx2,+xop | FileCheck %s --check-prefix=X32
4; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+xop | FileCheck %s --check-prefix=X64
5; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+xop | FileCheck %s --check-prefix=X64
6
7declare <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double>, <2 x double>, <2 x i64>, i8) nounwind readnone
8declare <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double>, <4 x double>, <4 x i64>, i8) nounwind readnone
9
10declare <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float>, <4 x float>, <4 x i32>, i8) nounwind readnone
11declare <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float>, <8 x float>, <8 x i32>, i8) nounwind readnone
12
13declare <16 x i8> @llvm.x86.xop.vpperm(<16 x i8>, <16 x i8>, <16 x i8>) nounwind readnone
14
15define <2 x double> @combine_vpermil2pd_identity(<2 x double> %a0, <2 x double> %a1) {
16; X32-LABEL: combine_vpermil2pd_identity:
17; X32:       # BB#0:
18; X32-NEXT:    vmovaps %xmm1, %xmm0
19; X32-NEXT:    retl
20;
21; X64-LABEL: combine_vpermil2pd_identity:
22; X64:       # BB#0:
23; X64-NEXT:    vmovaps %xmm1, %xmm0
24; X64-NEXT:    retq
25  %res0 = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> %a1, <2 x double> %a0, <2 x i64> <i64 2, i64 0>, i8 0)
26  %res1 = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> %res0, <2 x double> undef, <2 x i64> <i64 2, i64 0>, i8 0)
27  ret <2 x double> %res1
28}
29
30define <4 x double> @combine_vpermil2pd256_identity(<4 x double> %a0, <4 x double> %a1) {
31; X32-LABEL: combine_vpermil2pd256_identity:
32; X32:       # BB#0:
33; X32-NEXT:    vmovaps %ymm1, %ymm0
34; X32-NEXT:    retl
35;
36; X64-LABEL: combine_vpermil2pd256_identity:
37; X64:       # BB#0:
38; X64-NEXT:    vmovaps %ymm1, %ymm0
39; X64-NEXT:    retq
40  %res0 = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %a1, <4 x double> %a0, <4 x i64> <i64 2, i64 0, i64 2, i64 0>, i8 0)
41  %res1 = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %res0, <4 x double> undef, <4 x i64> <i64 2, i64 0, i64 2, i64 0>, i8 0)
42  ret <4 x double> %res1
43}
44
45define <4 x double> @combine_vpermil2pd256_0z73(<4 x double> %a0, <4 x double> %a1) {
46; X32-LABEL: combine_vpermil2pd256_0z73:
47; X32:       # BB#0:
48; X32-NEXT:    vpermil2pd {{.*#+}} ymm0 = ymm0[0],zero,ymm1[3],ymm0[3]
49; X32-NEXT:    retl
50;
51; X64-LABEL: combine_vpermil2pd256_0z73:
52; X64:       # BB#0:
53; X64-NEXT:    vpermil2pd {{.*#+}} ymm0 = ymm0[0],zero,ymm1[3],ymm0[3]
54; X64-NEXT:    retq
55  %res0 = shufflevector <4 x double> %a0, <4 x double> %a1, <4 x i32> <i32 0, i32 undef, i32 7, i32 3>
56  %res1 = shufflevector <4 x double> %res0, <4 x double> zeroinitializer, <4 x i32> <i32 0, i32 7, i32 2, i32 3>
57  ret <4 x double> %res1
58}
59
60define <4 x float> @combine_vpermil2ps_identity(<4 x float> %a0, <4 x float> %a1) {
61; X32-LABEL: combine_vpermil2ps_identity:
62; X32:       # BB#0:
63; X32-NEXT:    vmovaps %xmm1, %xmm0
64; X32-NEXT:    retl
65;
66; X64-LABEL: combine_vpermil2ps_identity:
67; X64:       # BB#0:
68; X64-NEXT:    vmovaps %xmm1, %xmm0
69; X64-NEXT:    retq
70  %res0 = call <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float> %a1, <4 x float> %a0, <4 x i32> <i32 3, i32 2, i32 1, i32 0>, i8 0)
71  %res1 = call <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float> %res0, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>, i8 0)
72  ret <4 x float> %res1
73}
74
75define <4 x float> @combine_vpermil2ps_1z74(<4 x float> %a0, <4 x float> %a1) {
76; X32-LABEL: combine_vpermil2ps_1z74:
77; X32:       # BB#0:
78; X32-NEXT:    vpermil2ps {{.*#+}} xmm0 = xmm0[1],zero,xmm1[3,0]
79; X32-NEXT:    retl
80;
81; X64-LABEL: combine_vpermil2ps_1z74:
82; X64:       # BB#0:
83; X64-NEXT:    vpermil2ps {{.*#+}} xmm0 = xmm0[1],zero,xmm1[3,0]
84; X64-NEXT:    retq
85  %res0 = call <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float> %a0, <4 x float> %a1, <4 x i32> <i32 1, i32 1, i32 7, i32 4>, i8 0)
86  %res1 = shufflevector <4 x float> %res0, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 7, i32 2, i32 3>
87  ret <4 x float> %res1
88}
89
90define <8 x float> @combine_vpermil2ps256_identity(<8 x float> %a0, <8 x float> %a1) {
91; X32-LABEL: combine_vpermil2ps256_identity:
92; X32:       # BB#0:
93; X32-NEXT:    vmovaps %ymm1, %ymm0
94; X32-NEXT:    retl
95;
96; X64-LABEL: combine_vpermil2ps256_identity:
97; X64:       # BB#0:
98; X64-NEXT:    vmovaps %ymm1, %ymm0
99; X64-NEXT:    retq
100  %res0 = call <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float> %a1, <8 x float> %a0, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 1, i32 0, i32 3, i32 2>, i8 0)
101  %res1 = call <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float> %res0, <8 x float> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 1, i32 0, i32 3, i32 2>, i8 0)
102  ret <8 x float> %res1
103}
104
105define <8 x float> @combine_vpermil2ps256_08z945Az(<8 x float> %a0, <8 x float> %a1) {
106; X32-LABEL: combine_vpermil2ps256_08z945Az:
107; X32:       # BB#0:
108; X32-NEXT:    vpermil2ps {{.*#+}} ymm0 = ymm0[0],ymm1[0],zero,ymm1[1],ymm0[4,5],ymm1[6],zero
109; X32-NEXT:    retl
110;
111; X64-LABEL: combine_vpermil2ps256_08z945Az:
112; X64:       # BB#0:
113; X64-NEXT:    vpermil2ps {{.*#+}} ymm0 = ymm0[0],ymm1[0],zero,ymm1[1],ymm0[4,5],ymm1[6],zero
114; X64-NEXT:    retq
115  %res0 = call <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 0, i32 1, i32 6, i32 7>, i8 0)
116  %res1 = shufflevector <8 x float> %res0, <8 x float> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 8, i32 3, i32 4, i32 5, i32 6, i32 8>
117  ret <8 x float> %res1
118}
119
120define <8 x float> @combine_vpermil2ps256_zero(<8 x float> %a0, <8 x float> %a1) {
121; X32-LABEL: combine_vpermil2ps256_zero:
122; X32:       # BB#0:
123; X32-NEXT:    vxorps %ymm0, %ymm0, %ymm0
124; X32-NEXT:    retl
125;
126; X64-LABEL: combine_vpermil2ps256_zero:
127; X64:       # BB#0:
128; X64-NEXT:    vxorps %ymm0, %ymm0, %ymm0
129; X64-NEXT:    retq
130  %res0 = call <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float> %a1, <8 x float> %a0, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 8, i32 9, i32 10, i32 11>, i8 2)
131  ret <8 x float> %res0
132}
133
134define <4 x float> @combine_vpermil2ps_blend_with_zero(<4 x float> %a0, <4 x float> %a1) {
135; X32-LABEL: combine_vpermil2ps_blend_with_zero:
136; X32:       # BB#0:
137; X32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
138; X32-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
139; X32-NEXT:    retl
140;
141; X64-LABEL: combine_vpermil2ps_blend_with_zero:
142; X64:       # BB#0:
143; X64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
144; X64-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
145; X64-NEXT:    retq
146  %res0 = call <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float> %a0, <4 x float> %a1, <4 x i32> <i32 8, i32 1, i32 2, i32 3>, i8 2)
147  ret <4 x float> %res0
148}
149
150define <2 x double> @combine_vpermil2pd_as_shufpd(<2 x double> %a0, <2 x double> %a1) {
151; X32-LABEL: combine_vpermil2pd_as_shufpd:
152; X32:       # BB#0:
153; X32-NEXT:    vshufpd {{.*#+}} xmm0 = xmm0[1],xmm1[0]
154; X32-NEXT:    retl
155;
156; X64-LABEL: combine_vpermil2pd_as_shufpd:
157; X64:       # BB#0:
158; X64-NEXT:    vshufpd {{.*#+}} xmm0 = xmm0[1],xmm1[0]
159; X64-NEXT:    retq
160  %res0 = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> %a0, <2 x double> %a1, <2 x i64> <i64 2, i64 4>, i8 0)
161  ret <2 x double> %res0
162}
163
164define <4 x double> @combine_vpermil2pd256_as_shufpd(<4 x double> %a0, <4 x double> %a1) {
165; X32-LABEL: combine_vpermil2pd256_as_shufpd:
166; X32:       # BB#0:
167; X32-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[3]
168; X32-NEXT:    retl
169;
170; X64-LABEL: combine_vpermil2pd256_as_shufpd:
171; X64:       # BB#0:
172; X64-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[3]
173; X64-NEXT:    retq
174  %res0 = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %a0, <4 x double> %a1, <4 x i64> <i64 0, i64 4, i64 2, i64 7>, i8 0)
175  ret <4 x double> %res0
176}
177
178define <16 x i8> @combine_vpperm_identity(<16 x i8> %a0, <16 x i8> %a1) {
179; X32-LABEL: combine_vpperm_identity:
180; X32:       # BB#0:
181; X32-NEXT:    vmovaps %xmm1, %xmm0
182; X32-NEXT:    retl
183;
184; X64-LABEL: combine_vpperm_identity:
185; X64:       # BB#0:
186; X64-NEXT:    vmovaps %xmm1, %xmm0
187; X64-NEXT:    retq
188  %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> <i8 31, i8 30, i8 29, i8 28, i8 27, i8 26, i8 25, i8 24, i8 23, i8 22, i8 21, i8 20, i8 19, i8 18, i8 17, i8 16>)
189  %res1 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %res0, <16 x i8> undef, <16 x i8> <i8 15, i8 14, i8 13, i8 12, i8 11, i8 10, i8 9, i8 8, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
190  ret <16 x i8> %res1
191}
192
193define <16 x i8> @combine_vpperm_zero(<16 x i8> %a0, <16 x i8> %a1) {
194; X32-LABEL: combine_vpperm_zero:
195; X32:       # BB#0:
196; X32-NEXT:    vxorps %xmm0, %xmm0, %xmm0
197; X32-NEXT:    retl
198;
199; X64-LABEL: combine_vpperm_zero:
200; X64:       # BB#0:
201; X64-NEXT:    vxorps %xmm0, %xmm0, %xmm0
202; X64-NEXT:    retq
203  %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> <i8 128, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>)
204  %res1 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %res0, <16 x i8> undef, <16 x i8> <i8 0, i8 128, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>)
205  %res2 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %res1, <16 x i8> undef, <16 x i8> <i8 0, i8 1, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128>)
206  ret <16 x i8> %res2
207}
208
209define <16 x i8> @combine_vpperm_identity_bitcast(<16 x i8> %a0, <16 x i8> %a1) {
210; X32-LABEL: combine_vpperm_identity_bitcast:
211; X32:       # BB#0:
212; X32-NEXT:    vpaddq {{\.LCPI.*}}, %xmm0, %xmm0
213; X32-NEXT:    retl
214;
215; X64-LABEL: combine_vpperm_identity_bitcast:
216; X64:       # BB#0:
217; X64-NEXT:    vpaddq {{.*}}(%rip), %xmm0, %xmm0
218; X64-NEXT:    retq
219  %mask = bitcast <2 x i64> <i64 1084818905618843912, i64 506097522914230528> to <16 x i8>
220  %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> %mask)
221  %res1 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %res0, <16 x i8> undef, <16 x i8> %mask)
222  %res2 = bitcast <16 x i8> %res1 to <2 x i64>
223  %res3 = add <2 x i64> %res2, <i64 1084818905618843912, i64 506097522914230528>
224  %res4 = bitcast <2 x i64> %res3 to <16 x i8>
225  ret <16 x i8> %res4
226}
227
228define <16 x i8> @combine_vpperm_as_blend_with_zero(<16 x i8> %a0, <16 x i8> %a1) {
229; X32-LABEL: combine_vpperm_as_blend_with_zero:
230; X32:       # BB#0:
231; X32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
232; X32-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4,5,6,7]
233; X32-NEXT:    retl
234;
235; X64-LABEL: combine_vpperm_as_blend_with_zero:
236; X64:       # BB#0:
237; X64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
238; X64-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4,5,6,7]
239; X64-NEXT:    retq
240  %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> <i8 0, i8 1, i8 128, i8 129, i8 4, i8 5, i8 6, i8 7, i8 130, i8 131, i8 132, i8 133, i8 134, i8 135, i8 136, i8 137>)
241  ret <16 x i8> %res0
242}
243
244define <16 x i8> @combine_vpperm_as_unary_unpckhbw(<16 x i8> %a0, <16 x i8> %a1) {
245; X32-LABEL: combine_vpperm_as_unary_unpckhbw:
246; X32:       # BB#0:
247; X32-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
248; X32-NEXT:    retl
249;
250; X64-LABEL: combine_vpperm_as_unary_unpckhbw:
251; X64:       # BB#0:
252; X64-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
253; X64-NEXT:    retq
254  %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a0, <16 x i8> <i8 8, i8 undef, i8 9, i8 25, i8 10, i8 26, i8 11, i8 27, i8 12, i8 28, i8 13, i8 29, i8 14, i8 30, i8 15, i8 31>)
255  ret <16 x i8> %res0
256}
257
258define <16 x i8> @combine_vpperm_as_unpckhbw(<16 x i8> %a0, <16 x i8> %a1) {
259; X32-LABEL: combine_vpperm_as_unpckhbw:
260; X32:       # BB#0:
261; X32-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
262; X32-NEXT:    retl
263;
264; X64-LABEL: combine_vpperm_as_unpckhbw:
265; X64:       # BB#0:
266; X64-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
267; X64-NEXT:    retq
268  %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> <i8 8, i8 24, i8 9, i8 25, i8 10, i8 26, i8 11, i8 27, i8 12, i8 28, i8 13, i8 29, i8 14, i8 30, i8 15, i8 31>)
269  ret <16 x i8> %res0
270}
271
272define <16 x i8> @combine_vpperm_as_unpcklbw(<16 x i8> %a0, <16 x i8> %a1) {
273; X32-LABEL: combine_vpperm_as_unpcklbw:
274; X32:       # BB#0:
275; X32-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
276; X32-NEXT:    retl
277;
278; X64-LABEL: combine_vpperm_as_unpcklbw:
279; X64:       # BB#0:
280; X64-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
281; X64-NEXT:    retq
282  %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> <i8 16, i8 0, i8 17, i8 1, i8 18, i8 2, i8 19, i8 3, i8 20, i8 4, i8 21, i8 5, i8 22, i8 6, i8 23, i8 7>)
283  ret <16 x i8> %res0
284}
285
286define <4 x i32> @combine_vpperm_10zz32BA(<4 x i32> %a0, <4 x i32> %a1) {
287; X32-LABEL: combine_vpperm_10zz32BA:
288; X32:       # BB#0:
289; X32-NEXT:    vpperm {{.*#+}} xmm0 = xmm0[2,3,0,1],zero,zero,zero,zero,xmm0[6,7,4,5],xmm1[6,7,4,5]
290; X32-NEXT:    retl
291;
292; X64-LABEL: combine_vpperm_10zz32BA:
293; X64:       # BB#0:
294; X64-NEXT:    vpperm {{.*#+}} xmm0 = xmm0[2,3,0,1],zero,zero,zero,zero,xmm0[6,7,4,5],xmm1[6,7,4,5]
295; X64-NEXT:    retq
296  %res0 = shufflevector <4 x i32> %a0, <4 x i32> %a1, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
297  %res1 = bitcast <4 x i32> %res0 to <16 x i8>
298  %res2 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %res1, <16 x i8> undef, <16 x i8> <i8 2, i8 3, i8 0, i8 1, i8 128, i8 128, i8 128, i8 128, i8 10, i8 11, i8 8, i8 9, i8 14, i8 15, i8 12, i8 13>)
299  %res3 = bitcast <16 x i8> %res2 to <4 x i32>
300  ret <4 x i32> %res3
301}
302
303define void @buildvector_v4f23_0404(float %a, float %b, <4 x float>* %ptr) {
304; X32-LABEL: buildvector_v4f23_0404:
305; X32:       # BB#0:
306; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
307; X32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
308; X32-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
309; X32-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
310; X32-NEXT:    vmovaps %xmm0, (%eax)
311; X32-NEXT:    retl
312;
313; X64-LABEL: buildvector_v4f23_0404:
314; X64:       # BB#0:
315; X64-NEXT:    vpermil2ps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[0],xmm1[0]
316; X64-NEXT:    vmovaps %xmm0, (%rdi)
317; X64-NEXT:    retq
318  %v0 = insertelement <4 x float> undef, float %a, i32 0
319  %v1 = insertelement <4 x float> %v0,   float %b, i32 1
320  %v2 = insertelement <4 x float> %v1,   float %a, i32 2
321  %v3 = insertelement <4 x float> %v2,   float %b, i32 3
322  store <4 x float> %v3, <4 x float>* %ptr
323  ret void
324}
325
326define <2 x double> @constant_fold_vpermil2pd() {
327; X32-LABEL: constant_fold_vpermil2pd:
328; X32:       # BB#0:
329; X32-NEXT:    vmovaps {{.*#+}} xmm0 = [-2.000000e+00,2.000000e+00]
330; X32-NEXT:    retl
331;
332; X64-LABEL: constant_fold_vpermil2pd:
333; X64:       # BB#0:
334; X64-NEXT:    vmovaps {{.*#+}} xmm0 = [-2.000000e+00,2.000000e+00]
335; X64-NEXT:    retq
336  %1 = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> <double 1.0, double 2.0>, <2 x double> <double -2.0, double -1.0>, <2 x i64> <i64 4, i64 2>, i8 2)
337  ret <2 x double> %1
338}
339
340define <4 x double> @constant_fold_vpermil2pd_256() {
341; X32-LABEL: constant_fold_vpermil2pd_256:
342; X32:       # BB#0:
343; X32-NEXT:    vmovaps {{.*#+}} ymm0 = [-4.000000e+00,0.000000e+00,4.000000e+00,3.000000e+00]
344; X32-NEXT:    retl
345;
346; X64-LABEL: constant_fold_vpermil2pd_256:
347; X64:       # BB#0:
348; X64-NEXT:    vmovaps {{.*#+}} ymm0 = [-4.000000e+00,0.000000e+00,4.000000e+00,3.000000e+00]
349; X64-NEXT:    retq
350  %1 = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> <double 1.0, double 2.0, double 3.0, double 4.0>, <4 x double> <double -4.0, double -3.0, double -2.0, double -1.0>, <4 x i64> <i64 4, i64 8, i64 2, i64 0>, i8 2)
351  ret <4 x double> %1
352}
353
354define <4 x float> @constant_fold_vpermil2ps() {
355; X32-LABEL: constant_fold_vpermil2ps:
356; X32:       # BB#0:
357; X32-NEXT:    vmovaps {{.*#+}} xmm0 = [-4.000000e+00,1.000000e+00,3.000000e+00,0.000000e+00]
358; X32-NEXT:    retl
359;
360; X64-LABEL: constant_fold_vpermil2ps:
361; X64:       # BB#0:
362; X64-NEXT:    vmovaps {{.*#+}} xmm0 = [-4.000000e+00,1.000000e+00,3.000000e+00,0.000000e+00]
363; X64-NEXT:    retq
364  %1 = call <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, <4 x float> <float -4.0, float -3.0, float -2.0, float -1.0>, <4 x i32> <i32 4, i32 0, i32 2, i32 8>, i8 2)
365  ret <4 x float> %1
366}
367
368define <8 x float> @constant_fold_vpermil2ps_256() {
369; X32-LABEL: constant_fold_vpermil2ps_256:
370; X32:       # BB#0:
371; X32-NEXT:    vmovaps {{.*#+}} ymm0 = [-8.000000e+00,1.000000e+00,3.000000e+00,0.000000e+00,5.000000e+00,0.000000e+00,5.000000e+00,7.000000e+00]
372; X32-NEXT:    retl
373;
374; X64-LABEL: constant_fold_vpermil2ps_256:
375; X64:       # BB#0:
376; X64-NEXT:    vmovaps {{.*#+}} ymm0 = [-8.000000e+00,1.000000e+00,3.000000e+00,0.000000e+00,5.000000e+00,0.000000e+00,5.000000e+00,7.000000e+00]
377; X64-NEXT:    retq
378  %1 = call <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, <8 x float> <float -8.0, float -7.0, float -6.0, float -5.0, float -4.0, float -3.0, float -2.0, float -1.0>, <8 x i32> <i32 4, i32 0, i32 2, i32 8, i32 0, i32 8, i32 0, i32 2>, i8 2)
379  ret <8 x float> %1
380}
381
382define <16 x i8> @constant_fold_vpperm() {
383; X32-LABEL: constant_fold_vpperm:
384; X32:       # BB#0:
385; X32-NEXT:    vmovaps {{.*#+}} xmm0 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
386; X32-NEXT:    retl
387;
388; X64-LABEL: constant_fold_vpperm:
389; X64:       # BB#0:
390; X64-NEXT:    vmovaps {{.*#+}} xmm0 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
391; X64-NEXT:    retq
392  %1 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> <i8 0, i8 -1, i8 -2, i8 -3, i8 -4, i8 -5, i8 -6, i8 -7, i8 -8, i8 -9, i8 -10, i8 -11, i8 -12, i8 -13, i8 -14, i8 -15>, <16 x i8> <i8 15, i8 14, i8 13, i8 12, i8 11, i8 10, i8 9, i8 8, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>, <16 x i8> <i8 31, i8 30, i8 29, i8 28, i8 27, i8 26, i8 25, i8 24, i8 23, i8 22, i8 21, i8 20, i8 19, i8 18, i8 17, i8 16>)
393  ret <16 x i8> %1
394}
395
396define <4 x float> @PR31296(i8* %in) {
397; X32-LABEL: PR31296:
398; X32:       # BB#0: # %entry
399; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
400; X32-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
401; X32-NEXT:    vmovaps {{.*#+}} xmm1 = <0,1,u,u>
402; X32-NEXT:    vpermil2ps {{.*#+}} xmm0 = xmm0[0],xmm1[0,0,1]
403; X32-NEXT:    retl
404;
405; X64-LABEL: PR31296:
406; X64:       # BB#0: # %entry
407; X64-NEXT:    movl (%rdi), %eax
408; X64-NEXT:    vmovq %rax, %xmm0
409; X64-NEXT:    vmovaps {{.*#+}} xmm1 = <0,1,u,u>
410; X64-NEXT:    vpermil2ps {{.*#+}} xmm0 = xmm0[0],xmm1[0,0,1]
411; X64-NEXT:    retq
412entry:
413  %0 = getelementptr i8, i8* %in, i32 0
414  %1 = bitcast i8* %0 to i32*
415  %2 = load i32, i32* %1
416  %3 = zext i32 %2 to i128
417  %4 = bitcast i128 %3 to <4 x float>
418  %5 = shufflevector <4 x float> %4, <4 x float> <float 0.000000e+00, float 1.000000e+00, float undef, float undef>, <4 x i32> <i32 0, i32 4, i32 4, i32 5>
419  ret <4 x float> %5
420}
421