1 ; RUN: llc < %s -mcpu=x86-64 -mattr=+avx -x86-experimental-vector-shuffle-lowering | FileCheck %s --check-prefix=ALL --check-prefix=AVX1
2 ; RUN: llc < %s -mcpu=x86-64 -mattr=+avx2 -x86-experimental-vector-shuffle-lowering | FileCheck %s --check-prefix=ALL --check-prefix=AVX2
4 target triple = "x86_64-unknown-unknown"
6 define <8 x float> @shuffle_v8f32_00000000(<8 x float> %a, <8 x float> %b) {
7 ; ALL-LABEL: @shuffle_v8f32_00000000
9 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,0,0]
10 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
12 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
13 ret <8 x float> %shuffle
16 define <8 x float> @shuffle_v8f32_00000010(<8 x float> %a, <8 x float> %b) {
17 ; ALL-LABEL: @shuffle_v8f32_00000010
19 ; ALL-NEXT: vpermilps {{.*}} # xmm1 = xmm0[0,0,0,0]
20 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,1,0]
21 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
23 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0>
24 ret <8 x float> %shuffle
27 define <8 x float> @shuffle_v8f32_00000200(<8 x float> %a, <8 x float> %b) {
28 ; ALL-LABEL: @shuffle_v8f32_00000200
30 ; ALL-NEXT: vpermilps {{.*}} # xmm1 = xmm0[0,0,0,0]
31 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,2,0,0]
32 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
34 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0>
35 ret <8 x float> %shuffle
38 define <8 x float> @shuffle_v8f32_00003000(<8 x float> %a, <8 x float> %b) {
39 ; ALL-LABEL: @shuffle_v8f32_00003000
41 ; ALL-NEXT: vpermilps {{.*}} # xmm1 = xmm0[0,0,0,0]
42 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[3,0,0,0]
43 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
45 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 3, i32 0, i32 0, i32 0>
46 ret <8 x float> %shuffle
49 define <8 x float> @shuffle_v8f32_00040000(<8 x float> %a, <8 x float> %b) {
50 ; ALL-LABEL: @shuffle_v8f32_00040000
52 ; ALL-NEXT: vextractf128 $1, %ymm0, %xmm1
53 ; ALL-NEXT: vshufps {{.*}} # xmm1 = xmm1[0,0],xmm0[0,0]
54 ; ALL-NEXT: vshufps {{.*}} # xmm1 = xmm0[0,0],xmm1[2,0]
55 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,0,0]
56 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
58 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 4, i32 0, i32 0, i32 0, i32 0>
59 ret <8 x float> %shuffle
62 define <8 x float> @shuffle_v8f32_00500000(<8 x float> %a, <8 x float> %b) {
63 ; ALL-LABEL: @shuffle_v8f32_00500000
65 ; ALL-NEXT: vextractf128 $1, %ymm0, %xmm1
66 ; ALL-NEXT: vshufps {{.*}} # xmm1 = xmm1[1,0],xmm0[0,0]
67 ; ALL-NEXT: vshufps {{.*}} # xmm1 = xmm0[0,0],xmm1[0,2]
68 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,0,0]
69 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
71 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 5, i32 0, i32 0, i32 0, i32 0, i32 0>
72 ret <8 x float> %shuffle
75 define <8 x float> @shuffle_v8f32_06000000(<8 x float> %a, <8 x float> %b) {
76 ; ALL-LABEL: @shuffle_v8f32_06000000
78 ; ALL-NEXT: vextractf128 $1, %ymm0, %xmm1
79 ; ALL-NEXT: vshufps {{.*}} # xmm1 = xmm1[2,0],xmm0[0,0]
80 ; ALL-NEXT: vshufps {{.*}} # xmm1 = xmm1[2,0],xmm0[0,0]
81 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,0,0]
82 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
84 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
85 ret <8 x float> %shuffle
88 define <8 x float> @shuffle_v8f32_70000000(<8 x float> %a, <8 x float> %b) {
89 ; ALL-LABEL: @shuffle_v8f32_70000000
91 ; ALL-NEXT: vextractf128 $1, %ymm0, %xmm1
92 ; ALL-NEXT: vshufps {{.*}} # xmm1 = xmm1[3,0],xmm0[0,0]
93 ; ALL-NEXT: vshufps {{.*}} # xmm1 = xmm1[0,2],xmm0[0,0]
94 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,0,0]
95 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
97 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
98 ret <8 x float> %shuffle
101 define <8 x float> @shuffle_v8f32_01014545(<8 x float> %a, <8 x float> %b) {
102 ; ALL-LABEL: @shuffle_v8f32_01014545
104 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,1,0,1,4,5,4,5]
106 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5>
107 ret <8 x float> %shuffle
110 define <8 x float> @shuffle_v8f32_00112233(<8 x float> %a, <8 x float> %b) {
111 ; ALL-LABEL: @shuffle_v8f32_00112233
113 ; ALL-NEXT: vunpcklps {{.*}} # xmm1 = xmm0[0,0,1,1]
114 ; ALL-NEXT: vunpckhps {{.*}} # xmm0 = xmm0[2,2,3,3]
115 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
117 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 1, i32 1, i32 2, i32 2, i32 3, i32 3>
118 ret <8 x float> %shuffle
121 define <8 x float> @shuffle_v8f32_00001111(<8 x float> %a, <8 x float> %b) {
122 ; ALL-LABEL: @shuffle_v8f32_00001111
124 ; ALL-NEXT: vpermilps {{.*}} # xmm1 = xmm0[0,0,0,0]
125 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[1,1,1,1]
126 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
128 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1>
129 ret <8 x float> %shuffle
132 define <8 x float> @shuffle_v8f32_81a3c5e7(<8 x float> %a, <8 x float> %b) {
133 ; ALL-LABEL: @shuffle_v8f32_81a3c5e7
135 ; ALL-NEXT: vblendps {{.*}} # ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
137 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 8, i32 1, i32 10, i32 3, i32 12, i32 5, i32 14, i32 7>
138 ret <8 x float> %shuffle
141 define <8 x float> @shuffle_v8f32_08080808(<8 x float> %a, <8 x float> %b) {
142 ; ALL-LABEL: @shuffle_v8f32_08080808
144 ; ALL-NEXT: vpermilps {{.*}} # xmm1 = xmm1[0,0,2,0]
145 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,1,0,3]
146 ; ALL-NEXT: vblendps {{.*}} # xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
147 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
149 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 8, i32 0, i32 8, i32 0, i32 8, i32 0, i32 8>
150 ret <8 x float> %shuffle
153 define <8 x float> @shuffle_v8f32_08084c4c(<8 x float> %a, <8 x float> %b) {
154 ; ALL-LABEL: @shuffle_v8f32_08084c4c
156 ; ALL-NEXT: vshufps {{.*}} # ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
157 ; ALL-NEXT: vshufps {{.*}} # ymm0 = ymm0[0,2,1,3,4,6,5,7]
159 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 8, i32 0, i32 8, i32 4, i32 12, i32 4, i32 12>
160 ret <8 x float> %shuffle
163 define <8 x float> @shuffle_v8f32_8823cc67(<8 x float> %a, <8 x float> %b) {
164 ; ALL-LABEL: @shuffle_v8f32_8823cc67
166 ; ALL-NEXT: vshufps {{.*}} # ymm0 = ymm1[0,0],ymm0[2,3],ymm1[4,4],ymm0[6,7]
168 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 8, i32 8, i32 2, i32 3, i32 12, i32 12, i32 6, i32 7>
169 ret <8 x float> %shuffle
172 define <8 x float> @shuffle_v8f32_9832dc76(<8 x float> %a, <8 x float> %b) {
173 ; ALL-LABEL: @shuffle_v8f32_9832dc76
175 ; ALL-NEXT: vshufps {{.*}} # ymm0 = ymm1[1,0],ymm0[3,2],ymm1[5,4],ymm0[7,6]
177 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 9, i32 8, i32 3, i32 2, i32 13, i32 12, i32 7, i32 6>
178 ret <8 x float> %shuffle
181 define <8 x float> @shuffle_v8f32_9810dc54(<8 x float> %a, <8 x float> %b) {
182 ; ALL-LABEL: @shuffle_v8f32_9810dc54
184 ; ALL-NEXT: vshufps {{.*}} # ymm0 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
186 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 9, i32 8, i32 1, i32 0, i32 13, i32 12, i32 5, i32 4>
187 ret <8 x float> %shuffle
190 define <8 x float> @shuffle_v8f32_08194c5d(<8 x float> %a, <8 x float> %b) {
191 ; ALL-LABEL: @shuffle_v8f32_08194c5d
193 ; ALL-NEXT: vunpcklps {{.*}} # ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
195 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
196 ret <8 x float> %shuffle
199 define <8 x float> @shuffle_v8f32_2a3b6e7f(<8 x float> %a, <8 x float> %b) {
200 ; ALL-LABEL: @shuffle_v8f32_2a3b6e7f
202 ; ALL-NEXT: vunpckhps {{.*}} # ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
204 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
205 ret <8 x float> %shuffle
208 define <8 x float> @shuffle_v8f32_08192a3b(<8 x float> %a, <8 x float> %b) {
209 ; ALL-LABEL: @shuffle_v8f32_08192a3b
211 ; ALL-NEXT: vpermilps {{.*}} # xmm2 = xmm1[0,2,2,3]
212 ; ALL-NEXT: vpermilps {{.*}} # xmm3 = xmm0[2,1,3,3]
213 ; ALL-NEXT: vblendps {{.*}} # xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3]
214 ; ALL-NEXT: vpermilps {{.*}} # xmm1 = xmm1[0,0,2,1]
215 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,1,1,3]
216 ; ALL-NEXT: vblendps {{.*}} # xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
217 ; ALL-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
219 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
220 ret <8 x float> %shuffle
223 define <8 x float> @shuffle_v8f32_08991abb(<8 x float> %a, <8 x float> %b) {
224 ; ALL-LABEL: @shuffle_v8f32_08991abb
226 ; ALL-NEXT: vpermilps {{.*}} # xmm2 = xmm0[1,1,2,3]
227 ; ALL-NEXT: vpermilps {{.*}} # xmm3 = xmm1[0,2,3,3]
228 ; ALL-NEXT: vblendps {{.*}} # xmm2 = xmm2[0],xmm3[1,2,3]
229 ; ALL-NEXT: vunpcklps {{.*}} # xmm1 = xmm1[0,0,1,1]
230 ; ALL-NEXT: vblendps {{.*}} # xmm0 = xmm0[0],xmm1[1,2,3]
231 ; ALL-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
233 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 8, i32 9, i32 9, i32 1, i32 10, i32 11, i32 11>
234 ret <8 x float> %shuffle
237 define <8 x float> @shuffle_v8f32_091b2d3f(<8 x float> %a, <8 x float> %b) {
238 ; ALL-LABEL: @shuffle_v8f32_091b2d3f
240 ; ALL-NEXT: vextractf128 $1, %ymm1, %xmm2
241 ; ALL-NEXT: vpermilps {{.*}} # xmm3 = xmm0[2,1,3,3]
242 ; ALL-NEXT: vblendps {{.*}} # xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3]
243 ; ALL-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,1,1,3]
244 ; ALL-NEXT: vblendps {{.*}} # xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
245 ; ALL-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
247 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 9, i32 1, i32 11, i32 2, i32 13, i32 3, i32 15>
248 ret <8 x float> %shuffle
251 define <8 x float> @shuffle_v8f32_09ab1def(<8 x float> %a, <8 x float> %b) {
252 ; ALL-LABEL: @shuffle_v8f32_09ab1def
254 ; ALL-NEXT: vextractf128 $1, %ymm1, %xmm2
255 ; ALL-NEXT: vpermilps {{.*}} # xmm3 = xmm0[1,1,2,3]
256 ; ALL-NEXT: vblendps {{.*}} # xmm2 = xmm3[0],xmm2[1,2,3]
257 ; ALL-NEXT: vblendps {{.*}} # xmm0 = xmm0[0],xmm1[1,2,3]
258 ; ALL-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
260 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 1, i32 13, i32 14, i32 15>
261 ret <8 x float> %shuffle
264 define <8 x float> @shuffle_v8f32_00014445(<8 x float> %a, <8 x float> %b) {
265 ; ALL-LABEL: @shuffle_v8f32_00014445
267 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,0,1,4,4,4,5]
269 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 1, i32 4, i32 4, i32 4, i32 5>
270 ret <8 x float> %shuffle
273 define <8 x float> @shuffle_v8f32_00204464(<8 x float> %a, <8 x float> %b) {
274 ; ALL-LABEL: @shuffle_v8f32_00204464
276 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,0,4,4,6,4]
278 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 0, i32 4, i32 4, i32 6, i32 4>
279 ret <8 x float> %shuffle
282 define <8 x float> @shuffle_v8f32_03004744(<8 x float> %a, <8 x float> %b) {
283 ; ALL-LABEL: @shuffle_v8f32_03004744
285 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,3,0,0,4,7,4,4]
287 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 3, i32 0, i32 0, i32 4, i32 7, i32 4, i32 4>
288 ret <8 x float> %shuffle
291 define <8 x float> @shuffle_v8f32_10005444(<8 x float> %a, <8 x float> %b) {
292 ; ALL-LABEL: @shuffle_v8f32_10005444
294 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,0,0,5,4,4,4]
296 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 5, i32 4, i32 4, i32 4>
297 ret <8 x float> %shuffle
300 define <8 x float> @shuffle_v8f32_22006644(<8 x float> %a, <8 x float> %b) {
301 ; ALL-LABEL: @shuffle_v8f32_22006644
303 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[2,2,0,0,6,6,4,4]
305 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 2, i32 2, i32 0, i32 0, i32 6, i32 6, i32 4, i32 4>
306 ret <8 x float> %shuffle
309 define <8 x float> @shuffle_v8f32_33307774(<8 x float> %a, <8 x float> %b) {
310 ; ALL-LABEL: @shuffle_v8f32_33307774
312 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[3,3,3,0,7,7,7,4]
314 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 3, i32 3, i32 0, i32 7, i32 7, i32 7, i32 4>
315 ret <8 x float> %shuffle
318 define <8 x float> @shuffle_v8f32_32107654(<8 x float> %a, <8 x float> %b) {
319 ; ALL-LABEL: @shuffle_v8f32_32107654
321 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[3,2,1,0,7,6,5,4]
323 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
324 ret <8 x float> %shuffle
327 define <8 x float> @shuffle_v8f32_00234467(<8 x float> %a, <8 x float> %b) {
328 ; ALL-LABEL: @shuffle_v8f32_00234467
330 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,3,4,4,6,7]
332 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 4, i32 4, i32 6, i32 7>
333 ret <8 x float> %shuffle
336 define <8 x float> @shuffle_v8f32_00224466(<8 x float> %a, <8 x float> %b) {
337 ; ALL-LABEL: @shuffle_v8f32_00224466
339 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,2,4,4,6,6]
341 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
342 ret <8 x float> %shuffle
345 define <8 x float> @shuffle_v8f32_10325476(<8 x float> %a, <8 x float> %b) {
346 ; ALL-LABEL: @shuffle_v8f32_10325476
348 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,3,2,5,4,7,6]
350 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
351 ret <8 x float> %shuffle
354 define <8 x float> @shuffle_v8f32_11335577(<8 x float> %a, <8 x float> %b) {
355 ; ALL-LABEL: @shuffle_v8f32_11335577
357 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,1,3,3,5,5,7,7]
359 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7>
360 ret <8 x float> %shuffle
363 define <8 x float> @shuffle_v8f32_10235467(<8 x float> %a, <8 x float> %b) {
364 ; ALL-LABEL: @shuffle_v8f32_10235467
366 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,2,3,5,4,6,7]
368 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
369 ret <8 x float> %shuffle
372 define <8 x float> @shuffle_v8f32_10225466(<8 x float> %a, <8 x float> %b) {
373 ; ALL-LABEL: @shuffle_v8f32_10225466
375 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,2,2,5,4,6,6]
377 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 2, i32 5, i32 4, i32 6, i32 6>
378 ret <8 x float> %shuffle
381 define <8 x float> @shuffle_v8f32_00015444(<8 x float> %a, <8 x float> %b) {
382 ; ALL-LABEL: @shuffle_v8f32_00015444
384 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,0,1,5,4,4,4]
386 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 1, i32 5, i32 4, i32 4, i32 4>
387 ret <8 x float> %shuffle
390 define <8 x float> @shuffle_v8f32_00204644(<8 x float> %a, <8 x float> %b) {
391 ; ALL-LABEL: @shuffle_v8f32_00204644
393 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,0,4,6,4,4]
395 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 0, i32 4, i32 6, i32 4, i32 4>
396 ret <8 x float> %shuffle
399 define <8 x float> @shuffle_v8f32_03004474(<8 x float> %a, <8 x float> %b) {
400 ; ALL-LABEL: @shuffle_v8f32_03004474
402 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,3,0,0,4,4,7,4]
404 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 3, i32 0, i32 0, i32 4, i32 4, i32 7, i32 4>
405 ret <8 x float> %shuffle
408 define <8 x float> @shuffle_v8f32_10004444(<8 x float> %a, <8 x float> %b) {
409 ; ALL-LABEL: @shuffle_v8f32_10004444
411 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,0,0,4,4,4,4]
413 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4>
414 ret <8 x float> %shuffle
417 define <8 x float> @shuffle_v8f32_22006446(<8 x float> %a, <8 x float> %b) {
418 ; ALL-LABEL: @shuffle_v8f32_22006446
420 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[2,2,0,0,6,4,4,6]
422 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 2, i32 2, i32 0, i32 0, i32 6, i32 4, i32 4, i32 6>
423 ret <8 x float> %shuffle
426 define <8 x float> @shuffle_v8f32_33307474(<8 x float> %a, <8 x float> %b) {
427 ; ALL-LABEL: @shuffle_v8f32_33307474
429 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[3,3,3,0,7,4,7,4]
431 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 3, i32 3, i32 0, i32 7, i32 4, i32 7, i32 4>
432 ret <8 x float> %shuffle
435 define <8 x float> @shuffle_v8f32_32104567(<8 x float> %a, <8 x float> %b) {
436 ; ALL-LABEL: @shuffle_v8f32_32104567
438 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[3,2,1,0,4,5,6,7]
440 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7>
441 ret <8 x float> %shuffle
444 define <8 x float> @shuffle_v8f32_00236744(<8 x float> %a, <8 x float> %b) {
445 ; ALL-LABEL: @shuffle_v8f32_00236744
447 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,3,6,7,4,4]
449 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 6, i32 7, i32 4, i32 4>
450 ret <8 x float> %shuffle
453 define <8 x float> @shuffle_v8f32_00226644(<8 x float> %a, <8 x float> %b) {
454 ; ALL-LABEL: @shuffle_v8f32_00226644
456 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,2,6,6,4,4]
458 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 6, i32 6, i32 4, i32 4>
459 ret <8 x float> %shuffle
462 define <8 x float> @shuffle_v8f32_10324567(<8 x float> %a, <8 x float> %b) {
463 ; ALL-LABEL: @shuffle_v8f32_10324567
465 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,3,2,4,5,6,7]
467 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 6, i32 7>
468 ret <8 x float> %shuffle
471 define <8 x float> @shuffle_v8f32_11334567(<8 x float> %a, <8 x float> %b) {
472 ; ALL-LABEL: @shuffle_v8f32_11334567
474 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,1,3,3,4,5,6,7]
476 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 4, i32 5, i32 6, i32 7>
477 ret <8 x float> %shuffle
480 define <8 x float> @shuffle_v8f32_01235467(<8 x float> %a, <8 x float> %b) {
481 ; ALL-LABEL: @shuffle_v8f32_01235467
483 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,1,2,3,5,4,6,7]
485 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
486 ret <8 x float> %shuffle
489 define <8 x float> @shuffle_v8f32_01235466(<8 x float> %a, <8 x float> %b) {
490 ; ALL-LABEL: @shuffle_v8f32_01235466
492 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,1,2,3,5,4,6,6]
494 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 6, i32 6>
495 ret <8 x float> %shuffle
498 define <8 x float> @shuffle_v8f32_002u6u44(<8 x float> %a, <8 x float> %b) {
499 ; ALL-LABEL: @shuffle_v8f32_002u6u44
501 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,u,6,u,4,4]
503 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 undef, i32 6, i32 undef, i32 4, i32 4>
504 ret <8 x float> %shuffle
507 define <8 x float> @shuffle_v8f32_00uu66uu(<8 x float> %a, <8 x float> %b) {
508 ; ALL-LABEL: @shuffle_v8f32_00uu66uu
510 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,u,u,6,6,u,u]
512 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 undef, i32 undef, i32 6, i32 6, i32 undef, i32 undef>
513 ret <8 x float> %shuffle
516 define <8 x float> @shuffle_v8f32_103245uu(<8 x float> %a, <8 x float> %b) {
517 ; ALL-LABEL: @shuffle_v8f32_103245uu
519 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,3,2,4,5,u,u]
521 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 undef, i32 undef>
522 ret <8 x float> %shuffle
525 define <8 x float> @shuffle_v8f32_1133uu67(<8 x float> %a, <8 x float> %b) {
526 ; ALL-LABEL: @shuffle_v8f32_1133uu67
528 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,1,3,3,u,u,6,7]
530 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 undef, i32 undef, i32 6, i32 7>
531 ret <8 x float> %shuffle
534 define <8 x float> @shuffle_v8f32_0uu354uu(<8 x float> %a, <8 x float> %b) {
535 ; ALL-LABEL: @shuffle_v8f32_0uu354uu
537 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,u,u,3,5,4,u,u]
539 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 undef, i32 undef, i32 3, i32 5, i32 4, i32 undef, i32 undef>
540 ret <8 x float> %shuffle
543 define <8 x float> @shuffle_v8f32_uuu3uu66(<8 x float> %a, <8 x float> %b) {
544 ; ALL-LABEL: @shuffle_v8f32_uuu3uu66
546 ; ALL-NEXT: vpermilps {{.*}} # ymm0 = ymm0[u,u,u,3,u,u,6,6]
548 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 3, i32 undef, i32 undef, i32 6, i32 6>
549 ret <8 x float> %shuffle
552 define <8 x i32> @shuffle_v8i32_00000000(<8 x i32> %a, <8 x i32> %b) {
553 ; AVX1-LABEL: @shuffle_v8i32_00000000
555 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,0,0]
556 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
559 ; AVX2-LABEL: @shuffle_v8i32_00000000
561 ; AVX2-NEXT: vpxor %ymm1, %ymm1, %ymm1
562 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
564 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
565 ret <8 x i32> %shuffle
568 define <8 x i32> @shuffle_v8i32_00000010(<8 x i32> %a, <8 x i32> %b) {
569 ; AVX1-LABEL: @shuffle_v8i32_00000010
571 ; AVX1-NEXT: vpermilps {{.*}} # xmm1 = xmm0[0,0,0,0]
572 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,1,0]
573 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
576 ; AVX2-LABEL: @shuffle_v8i32_00000010
578 ; AVX2-NEXT: vmovdqa .LCPI56_0(%rip), %ymm1
579 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
581 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0>
582 ret <8 x i32> %shuffle
585 define <8 x i32> @shuffle_v8i32_00000200(<8 x i32> %a, <8 x i32> %b) {
586 ; AVX1-LABEL: @shuffle_v8i32_00000200
588 ; AVX1-NEXT: vpermilps {{.*}} # xmm1 = xmm0[0,0,0,0]
589 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,2,0,0]
590 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
593 ; AVX2-LABEL: @shuffle_v8i32_00000200
595 ; AVX2-NEXT: vmovdqa .LCPI57_0(%rip), %ymm1
596 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
598 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0>
599 ret <8 x i32> %shuffle
602 define <8 x i32> @shuffle_v8i32_00003000(<8 x i32> %a, <8 x i32> %b) {
603 ; AVX1-LABEL: @shuffle_v8i32_00003000
605 ; AVX1-NEXT: vpermilps {{.*}} # xmm1 = xmm0[0,0,0,0]
606 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[3,0,0,0]
607 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
610 ; AVX2-LABEL: @shuffle_v8i32_00003000
612 ; AVX2-NEXT: vmovdqa .LCPI58_0(%rip), %ymm1
613 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
615 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 3, i32 0, i32 0, i32 0>
616 ret <8 x i32> %shuffle
619 define <8 x i32> @shuffle_v8i32_00040000(<8 x i32> %a, <8 x i32> %b) {
620 ; AVX1-LABEL: @shuffle_v8i32_00040000
622 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
623 ; AVX1-NEXT: vshufps {{.*}} # xmm1 = xmm1[0,0],xmm0[0,0]
624 ; AVX1-NEXT: vshufps {{.*}} # xmm1 = xmm0[0,0],xmm1[2,0]
625 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,0,0]
626 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
629 ; AVX2-LABEL: @shuffle_v8i32_00040000
631 ; AVX2-NEXT: vmovdqa .LCPI59_0(%rip), %ymm1
632 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
634 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 4, i32 0, i32 0, i32 0, i32 0>
635 ret <8 x i32> %shuffle
638 define <8 x i32> @shuffle_v8i32_00500000(<8 x i32> %a, <8 x i32> %b) {
639 ; AVX1-LABEL: @shuffle_v8i32_00500000
641 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
642 ; AVX1-NEXT: vshufps {{.*}} # xmm1 = xmm1[1,0],xmm0[0,0]
643 ; AVX1-NEXT: vshufps {{.*}} # xmm1 = xmm0[0,0],xmm1[0,2]
644 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,0,0]
645 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
648 ; AVX2-LABEL: @shuffle_v8i32_00500000
650 ; AVX2-NEXT: vmovdqa .LCPI60_0(%rip), %ymm1
651 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
653 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 5, i32 0, i32 0, i32 0, i32 0, i32 0>
654 ret <8 x i32> %shuffle
657 define <8 x i32> @shuffle_v8i32_06000000(<8 x i32> %a, <8 x i32> %b) {
658 ; AVX1-LABEL: @shuffle_v8i32_06000000
660 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
661 ; AVX1-NEXT: vshufps {{.*}} # xmm1 = xmm1[2,0],xmm0[0,0]
662 ; AVX1-NEXT: vshufps {{.*}} # xmm1 = xmm1[2,0],xmm0[0,0]
663 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,0,0]
664 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
667 ; AVX2-LABEL: @shuffle_v8i32_06000000
669 ; AVX2-NEXT: vmovdqa .LCPI61_0(%rip), %ymm1
670 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
672 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
673 ret <8 x i32> %shuffle
676 define <8 x i32> @shuffle_v8i32_70000000(<8 x i32> %a, <8 x i32> %b) {
677 ; AVX1-LABEL: @shuffle_v8i32_70000000
679 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
680 ; AVX1-NEXT: vshufps {{.*}} # xmm1 = xmm1[3,0],xmm0[0,0]
681 ; AVX1-NEXT: vshufps {{.*}} # xmm1 = xmm1[0,2],xmm0[0,0]
682 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,0,0,0]
683 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
686 ; AVX2-LABEL: @shuffle_v8i32_70000000
688 ; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
689 ; AVX2-NEXT: movl $7, %eax
690 ; AVX2-NEXT: vpinsrd $0, %eax, %xmm1, %xmm1
691 ; AVX2-NEXT: vpxor %ymm2, %ymm2, %ymm2
692 ; AVX2-NEXT: vinserti128 $0, %xmm1, %ymm2, %ymm1
693 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
695 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
696 ret <8 x i32> %shuffle
699 define <8 x i32> @shuffle_v8i32_01014545(<8 x i32> %a, <8 x i32> %b) {
700 ; AVX1-LABEL: @shuffle_v8i32_01014545
702 ; AVX1-NEXT: vpermilpd {{.*}} # ymm0 = ymm0[0,0,2,2]
705 ; AVX2-LABEL: @shuffle_v8i32_01014545
707 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[0,1,0,1,4,5,4,5]
709 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5>
710 ret <8 x i32> %shuffle
713 define <8 x i32> @shuffle_v8i32_00112233(<8 x i32> %a, <8 x i32> %b) {
714 ; AVX1-LABEL: @shuffle_v8i32_00112233
716 ; AVX1-NEXT: vunpcklps {{.*}} # xmm1 = xmm0[0,0,1,1]
717 ; AVX1-NEXT: vunpckhps {{.*}} # xmm0 = xmm0[2,2,3,3]
718 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
721 ; AVX2-LABEL: @shuffle_v8i32_00112233
723 ; AVX2-NEXT: vmovdqa .LCPI64_0(%rip), %ymm1
724 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
726 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 1, i32 1, i32 2, i32 2, i32 3, i32 3>
727 ret <8 x i32> %shuffle
730 define <8 x i32> @shuffle_v8i32_00001111(<8 x i32> %a, <8 x i32> %b) {
731 ; AVX1-LABEL: @shuffle_v8i32_00001111
733 ; AVX1-NEXT: vpermilps {{.*}} # xmm1 = xmm0[0,0,0,0]
734 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[1,1,1,1]
735 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
738 ; AVX2-LABEL: @shuffle_v8i32_00001111
740 ; AVX2-NEXT: vmovdqa .LCPI65_0(%rip), %ymm1
741 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
743 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1>
744 ret <8 x i32> %shuffle
747 define <8 x i32> @shuffle_v8i32_81a3c5e7(<8 x i32> %a, <8 x i32> %b) {
748 ; AVX1-LABEL: @shuffle_v8i32_81a3c5e7
750 ; AVX1-NEXT: vblendps {{.*}} # ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
753 ; AVX2-LABEL: @shuffle_v8i32_81a3c5e7
755 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
757 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 8, i32 1, i32 10, i32 3, i32 12, i32 5, i32 14, i32 7>
758 ret <8 x i32> %shuffle
761 define <8 x i32> @shuffle_v8i32_08080808(<8 x i32> %a, <8 x i32> %b) {
762 ; AVX1-LABEL: @shuffle_v8i32_08080808
764 ; AVX1-NEXT: vpermilps {{.*}} # xmm1 = xmm1[0,0,2,0]
765 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,1,0,3]
766 ; AVX1-NEXT: vblendps {{.*}} # xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
767 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
770 ; AVX2-LABEL: @shuffle_v8i32_08080808
772 ; AVX2-NEXT: vpxor %ymm2, %ymm2, %ymm2
773 ; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1
774 ; AVX2-NEXT: vpermq {{.*}} # ymm0 = ymm0[0,0,0,0]
775 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
777 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 8, i32 0, i32 8, i32 0, i32 8, i32 0, i32 8>
778 ret <8 x i32> %shuffle
781 define <8 x i32> @shuffle_v8i32_08084c4c(<8 x i32> %a, <8 x i32> %b) {
782 ; AVX1-LABEL: @shuffle_v8i32_08084c4c
784 ; AVX1-NEXT: vshufps {{.*}} # ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
785 ; AVX1-NEXT: vshufps {{.*}} # ymm0 = ymm0[0,2,1,3,4,6,5,7]
788 ; AVX2-LABEL: @shuffle_v8i32_08084c4c
790 ; AVX2-NEXT: vpshufd {{.*}} # ymm1 = ymm1[0,0,2,0,4,4,6,4]
791 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[0,1,0,1,4,5,4,5]
792 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
794 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 8, i32 0, i32 8, i32 4, i32 12, i32 4, i32 12>
795 ret <8 x i32> %shuffle
798 define <8 x i32> @shuffle_v8i32_8823cc67(<8 x i32> %a, <8 x i32> %b) {
799 ; AVX1-LABEL: @shuffle_v8i32_8823cc67
801 ; AVX1-NEXT: vshufps {{.*}} # ymm0 = ymm1[0,0],ymm0[2,3],ymm1[4,4],ymm0[6,7]
804 ; AVX2-LABEL: @shuffle_v8i32_8823cc67
806 ; AVX2-NEXT: vpshufd {{.*}} # ymm1 = ymm1[0,0,2,3,4,4,6,7]
807 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
809 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 8, i32 8, i32 2, i32 3, i32 12, i32 12, i32 6, i32 7>
810 ret <8 x i32> %shuffle
813 define <8 x i32> @shuffle_v8i32_9832dc76(<8 x i32> %a, <8 x i32> %b) {
814 ; AVX1-LABEL: @shuffle_v8i32_9832dc76
816 ; AVX1-NEXT: vshufps {{.*}} # ymm0 = ymm1[1,0],ymm0[3,2],ymm1[5,4],ymm0[7,6]
819 ; AVX2-LABEL: @shuffle_v8i32_9832dc76
821 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[0,1,3,2,4,5,7,6]
822 ; AVX2-NEXT: vpshufd {{.*}} # ymm1 = ymm1[1,0,2,3,5,4,6,7]
823 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
825 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 9, i32 8, i32 3, i32 2, i32 13, i32 12, i32 7, i32 6>
826 ret <8 x i32> %shuffle
829 define <8 x i32> @shuffle_v8i32_9810dc54(<8 x i32> %a, <8 x i32> %b) {
830 ; AVX1-LABEL: @shuffle_v8i32_9810dc54
832 ; AVX1-NEXT: vshufps {{.*}} # ymm0 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
835 ; AVX2-LABEL: @shuffle_v8i32_9810dc54
837 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[0,1,1,0,4,5,5,4]
838 ; AVX2-NEXT: vpshufd {{.*}} # ymm1 = ymm1[1,0,2,3,5,4,6,7]
839 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
841 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 9, i32 8, i32 1, i32 0, i32 13, i32 12, i32 5, i32 4>
842 ret <8 x i32> %shuffle
845 define <8 x i32> @shuffle_v8i32_08194c5d(<8 x i32> %a, <8 x i32> %b) {
846 ; AVX1-LABEL: @shuffle_v8i32_08194c5d
848 ; AVX1-NEXT: vunpcklps {{.*}} # ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
851 ; AVX2-LABEL: @shuffle_v8i32_08194c5d
853 ; AVX2-NEXT: vpshufd {{.*}} # ymm1 = ymm1[0,0,2,1,4,4,6,5]
854 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[0,1,1,3,4,5,5,7]
855 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
857 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
858 ret <8 x i32> %shuffle
861 define <8 x i32> @shuffle_v8i32_2a3b6e7f(<8 x i32> %a, <8 x i32> %b) {
862 ; AVX1-LABEL: @shuffle_v8i32_2a3b6e7f
864 ; AVX1-NEXT: vunpckhps {{.*}} # ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
867 ; AVX2-LABEL: @shuffle_v8i32_2a3b6e7f
869 ; AVX2-NEXT: vpshufd {{.*}} # ymm1 = ymm1[0,2,2,3,4,6,6,7]
870 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[2,1,3,3,6,5,7,7]
871 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
873 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
874 ret <8 x i32> %shuffle
877 define <8 x i32> @shuffle_v8i32_08192a3b(<8 x i32> %a, <8 x i32> %b) {
878 ; AVX1-LABEL: @shuffle_v8i32_08192a3b
880 ; AVX1-NEXT: vpermilps {{.*}} # xmm2 = xmm1[0,2,2,3]
881 ; AVX1-NEXT: vpermilps {{.*}} # xmm3 = xmm0[2,1,3,3]
882 ; AVX1-NEXT: vblendps {{.*}} # xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3]
883 ; AVX1-NEXT: vpermilps {{.*}} # xmm1 = xmm1[0,0,2,1]
884 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,1,1,3]
885 ; AVX1-NEXT: vblendps {{.*}} # xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
886 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
889 ; AVX2-LABEL: @shuffle_v8i32_08192a3b
891 ; AVX2-NEXT: vmovdqa .LCPI74_0(%rip), %ymm2
892 ; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1
893 ; AVX2-NEXT: vmovdqa .LCPI74_1(%rip), %ymm2
894 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
895 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
897 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
898 ret <8 x i32> %shuffle
901 define <8 x i32> @shuffle_v8i32_08991abb(<8 x i32> %a, <8 x i32> %b) {
902 ; AVX1-LABEL: @shuffle_v8i32_08991abb
904 ; AVX1-NEXT: vpermilps {{.*}} # xmm2 = xmm0[1,1,2,3]
905 ; AVX1-NEXT: vpermilps {{.*}} # xmm3 = xmm1[0,2,3,3]
906 ; AVX1-NEXT: vblendps {{.*}} # xmm2 = xmm2[0],xmm3[1,2,3]
907 ; AVX1-NEXT: vunpcklps {{.*}} # xmm1 = xmm1[0,0,1,1]
908 ; AVX1-NEXT: vblendps {{.*}} # xmm0 = xmm0[0],xmm1[1,2,3]
909 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
912 ; AVX2-LABEL: @shuffle_v8i32_08991abb
914 ; AVX2-NEXT: vmovdqa .LCPI75_0(%rip), %ymm2
915 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
916 ; AVX2-NEXT: vmovdqa .LCPI75_1(%rip), %ymm2
917 ; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1
918 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
920 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 8, i32 9, i32 9, i32 1, i32 10, i32 11, i32 11>
921 ret <8 x i32> %shuffle
924 define <8 x i32> @shuffle_v8i32_091b2d3f(<8 x i32> %a, <8 x i32> %b) {
925 ; AVX1-LABEL: @shuffle_v8i32_091b2d3f
927 ; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
928 ; AVX1-NEXT: vpermilps {{.*}} # xmm3 = xmm0[2,1,3,3]
929 ; AVX1-NEXT: vblendps {{.*}} # xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3]
930 ; AVX1-NEXT: vpermilps {{.*}} # xmm0 = xmm0[0,1,1,3]
931 ; AVX1-NEXT: vblendps {{.*}} # xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
932 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
935 ; AVX2-LABEL: @shuffle_v8i32_091b2d3f
937 ; AVX2-NEXT: vmovdqa .LCPI76_0(%rip), %ymm2
938 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
939 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
941 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 9, i32 1, i32 11, i32 2, i32 13, i32 3, i32 15>
942 ret <8 x i32> %shuffle
945 define <8 x i32> @shuffle_v8i32_09ab1def(<8 x i32> %a, <8 x i32> %b) {
946 ; AVX1-LABEL: @shuffle_v8i32_09ab1def
948 ; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
949 ; AVX1-NEXT: vpermilps {{.*}} # xmm3 = xmm0[1,1,2,3]
950 ; AVX1-NEXT: vblendps {{.*}} # xmm2 = xmm3[0],xmm2[1,2,3]
951 ; AVX1-NEXT: vblendps {{.*}} # xmm0 = xmm0[0],xmm1[1,2,3]
952 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
955 ; AVX2-LABEL: @shuffle_v8i32_09ab1def
957 ; AVX2-NEXT: vmovdqa .LCPI77_0(%rip), %ymm2
958 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
959 ; AVX2-NEXT: vpblendd {{.*}} # ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
961 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 1, i32 13, i32 14, i32 15>
962 ret <8 x i32> %shuffle
965 define <8 x i32> @shuffle_v8i32_00014445(<8 x i32> %a, <8 x i32> %b) {
966 ; AVX1-LABEL: @shuffle_v8i32_00014445
968 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,0,1,4,4,4,5]
971 ; AVX2-LABEL: @shuffle_v8i32_00014445
973 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[0,0,0,1,4,4,4,5]
975 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 1, i32 4, i32 4, i32 4, i32 5>
976 ret <8 x i32> %shuffle
979 define <8 x i32> @shuffle_v8i32_00204464(<8 x i32> %a, <8 x i32> %b) {
980 ; AVX1-LABEL: @shuffle_v8i32_00204464
982 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,0,4,4,6,4]
985 ; AVX2-LABEL: @shuffle_v8i32_00204464
987 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[0,0,2,0,4,4,6,4]
989 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 0, i32 4, i32 4, i32 6, i32 4>
990 ret <8 x i32> %shuffle
993 define <8 x i32> @shuffle_v8i32_03004744(<8 x i32> %a, <8 x i32> %b) {
994 ; AVX1-LABEL: @shuffle_v8i32_03004744
996 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,3,0,0,4,7,4,4]
999 ; AVX2-LABEL: @shuffle_v8i32_03004744
1001 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[0,3,0,0,4,7,4,4]
1003 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 3, i32 0, i32 0, i32 4, i32 7, i32 4, i32 4>
1004 ret <8 x i32> %shuffle
1007 define <8 x i32> @shuffle_v8i32_10005444(<8 x i32> %a, <8 x i32> %b) {
1008 ; AVX1-LABEL: @shuffle_v8i32_10005444
1010 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,0,0,5,4,4,4]
1013 ; AVX2-LABEL: @shuffle_v8i32_10005444
1015 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[1,0,0,0,5,4,4,4]
1017 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 5, i32 4, i32 4, i32 4>
1018 ret <8 x i32> %shuffle
1021 define <8 x i32> @shuffle_v8i32_22006644(<8 x i32> %a, <8 x i32> %b) {
1022 ; AVX1-LABEL: @shuffle_v8i32_22006644
1024 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[2,2,0,0,6,6,4,4]
1027 ; AVX2-LABEL: @shuffle_v8i32_22006644
1029 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[2,2,0,0,6,6,4,4]
1031 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 2, i32 2, i32 0, i32 0, i32 6, i32 6, i32 4, i32 4>
1032 ret <8 x i32> %shuffle
1035 define <8 x i32> @shuffle_v8i32_33307774(<8 x i32> %a, <8 x i32> %b) {
1036 ; AVX1-LABEL: @shuffle_v8i32_33307774
1038 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[3,3,3,0,7,7,7,4]
1041 ; AVX2-LABEL: @shuffle_v8i32_33307774
1043 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[3,3,3,0,7,7,7,4]
1045 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 3, i32 3, i32 0, i32 7, i32 7, i32 7, i32 4>
1046 ret <8 x i32> %shuffle
1049 define <8 x i32> @shuffle_v8i32_32107654(<8 x i32> %a, <8 x i32> %b) {
1050 ; AVX1-LABEL: @shuffle_v8i32_32107654
1052 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[3,2,1,0,7,6,5,4]
1055 ; AVX2-LABEL: @shuffle_v8i32_32107654
1057 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[3,2,1,0,7,6,5,4]
1059 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
1060 ret <8 x i32> %shuffle
1063 define <8 x i32> @shuffle_v8i32_00234467(<8 x i32> %a, <8 x i32> %b) {
1064 ; AVX1-LABEL: @shuffle_v8i32_00234467
1066 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,3,4,4,6,7]
1069 ; AVX2-LABEL: @shuffle_v8i32_00234467
1071 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[0,0,2,3,4,4,6,7]
1073 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 4, i32 4, i32 6, i32 7>
1074 ret <8 x i32> %shuffle
1077 define <8 x i32> @shuffle_v8i32_00224466(<8 x i32> %a, <8 x i32> %b) {
1078 ; AVX1-LABEL: @shuffle_v8i32_00224466
1080 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,2,4,4,6,6]
1083 ; AVX2-LABEL: @shuffle_v8i32_00224466
1085 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[0,0,2,2,4,4,6,6]
1087 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
1088 ret <8 x i32> %shuffle
1091 define <8 x i32> @shuffle_v8i32_10325476(<8 x i32> %a, <8 x i32> %b) {
1092 ; AVX1-LABEL: @shuffle_v8i32_10325476
1094 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,3,2,5,4,7,6]
1097 ; AVX2-LABEL: @shuffle_v8i32_10325476
1099 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[1,0,3,2,5,4,7,6]
1101 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
1102 ret <8 x i32> %shuffle
1105 define <8 x i32> @shuffle_v8i32_11335577(<8 x i32> %a, <8 x i32> %b) {
1106 ; AVX1-LABEL: @shuffle_v8i32_11335577
1108 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,1,3,3,5,5,7,7]
1111 ; AVX2-LABEL: @shuffle_v8i32_11335577
1113 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[1,1,3,3,5,5,7,7]
1115 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7>
1116 ret <8 x i32> %shuffle
1119 define <8 x i32> @shuffle_v8i32_10235467(<8 x i32> %a, <8 x i32> %b) {
1120 ; AVX1-LABEL: @shuffle_v8i32_10235467
1122 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,2,3,5,4,6,7]
1125 ; AVX2-LABEL: @shuffle_v8i32_10235467
1127 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[1,0,2,3,5,4,6,7]
1129 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
1130 ret <8 x i32> %shuffle
1133 define <8 x i32> @shuffle_v8i32_10225466(<8 x i32> %a, <8 x i32> %b) {
1134 ; AVX1-LABEL: @shuffle_v8i32_10225466
1136 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,2,2,5,4,6,6]
1139 ; AVX2-LABEL: @shuffle_v8i32_10225466
1141 ; AVX2-NEXT: vpshufd {{.*}} # ymm0 = ymm0[1,0,2,2,5,4,6,6]
1143 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 2, i32 5, i32 4, i32 6, i32 6>
1144 ret <8 x i32> %shuffle
1147 define <8 x i32> @shuffle_v8i32_00015444(<8 x i32> %a, <8 x i32> %b) {
1148 ; AVX1-LABEL: @shuffle_v8i32_00015444
1150 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,0,1,5,4,4,4]
1153 ; AVX2-LABEL: @shuffle_v8i32_00015444
1155 ; AVX2-NEXT: vmovdqa .LCPI91_0(%rip), %ymm1
1156 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1158 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 1, i32 5, i32 4, i32 4, i32 4>
1159 ret <8 x i32> %shuffle
1162 define <8 x i32> @shuffle_v8i32_00204644(<8 x i32> %a, <8 x i32> %b) {
1163 ; AVX1-LABEL: @shuffle_v8i32_00204644
1165 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,0,4,6,4,4]
1168 ; AVX2-LABEL: @shuffle_v8i32_00204644
1170 ; AVX2-NEXT: vmovdqa .LCPI92_0(%rip), %ymm1
1171 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1173 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 0, i32 4, i32 6, i32 4, i32 4>
1174 ret <8 x i32> %shuffle
1177 define <8 x i32> @shuffle_v8i32_03004474(<8 x i32> %a, <8 x i32> %b) {
1178 ; AVX1-LABEL: @shuffle_v8i32_03004474
1180 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,3,0,0,4,4,7,4]
1183 ; AVX2-LABEL: @shuffle_v8i32_03004474
1185 ; AVX2-NEXT: vmovdqa .LCPI93_0(%rip), %ymm1
1186 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1188 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 3, i32 0, i32 0, i32 4, i32 4, i32 7, i32 4>
1189 ret <8 x i32> %shuffle
1192 define <8 x i32> @shuffle_v8i32_10004444(<8 x i32> %a, <8 x i32> %b) {
1193 ; AVX1-LABEL: @shuffle_v8i32_10004444
1195 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,0,0,4,4,4,4]
1198 ; AVX2-LABEL: @shuffle_v8i32_10004444
1200 ; AVX2-NEXT: vmovdqa .LCPI94_0(%rip), %ymm1
1201 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1203 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4>
1204 ret <8 x i32> %shuffle
1207 define <8 x i32> @shuffle_v8i32_22006446(<8 x i32> %a, <8 x i32> %b) {
1208 ; AVX1-LABEL: @shuffle_v8i32_22006446
1210 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[2,2,0,0,6,4,4,6]
1213 ; AVX2-LABEL: @shuffle_v8i32_22006446
1215 ; AVX2-NEXT: vmovdqa .LCPI95_0(%rip), %ymm1
1216 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1218 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 2, i32 2, i32 0, i32 0, i32 6, i32 4, i32 4, i32 6>
1219 ret <8 x i32> %shuffle
1222 define <8 x i32> @shuffle_v8i32_33307474(<8 x i32> %a, <8 x i32> %b) {
1223 ; AVX1-LABEL: @shuffle_v8i32_33307474
1225 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[3,3,3,0,7,4,7,4]
1228 ; AVX2-LABEL: @shuffle_v8i32_33307474
1230 ; AVX2-NEXT: vmovdqa .LCPI96_0(%rip), %ymm1
1231 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1233 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 3, i32 3, i32 0, i32 7, i32 4, i32 7, i32 4>
1234 ret <8 x i32> %shuffle
1237 define <8 x i32> @shuffle_v8i32_32104567(<8 x i32> %a, <8 x i32> %b) {
1238 ; AVX1-LABEL: @shuffle_v8i32_32104567
1240 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[3,2,1,0,4,5,6,7]
1243 ; AVX2-LABEL: @shuffle_v8i32_32104567
1245 ; AVX2-NEXT: vmovdqa .LCPI97_0(%rip), %ymm1
1246 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1248 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7>
1249 ret <8 x i32> %shuffle
1252 define <8 x i32> @shuffle_v8i32_00236744(<8 x i32> %a, <8 x i32> %b) {
1253 ; AVX1-LABEL: @shuffle_v8i32_00236744
1255 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,3,6,7,4,4]
1258 ; AVX2-LABEL: @shuffle_v8i32_00236744
1260 ; AVX2-NEXT: vmovdqa .LCPI98_0(%rip), %ymm1
1261 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1263 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 6, i32 7, i32 4, i32 4>
1264 ret <8 x i32> %shuffle
1267 define <8 x i32> @shuffle_v8i32_00226644(<8 x i32> %a, <8 x i32> %b) {
1268 ; AVX1-LABEL: @shuffle_v8i32_00226644
1270 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,2,6,6,4,4]
1273 ; AVX2-LABEL: @shuffle_v8i32_00226644
1275 ; AVX2-NEXT: vmovdqa .LCPI99_0(%rip), %ymm1
1276 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1278 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 6, i32 6, i32 4, i32 4>
1279 ret <8 x i32> %shuffle
1282 define <8 x i32> @shuffle_v8i32_10324567(<8 x i32> %a, <8 x i32> %b) {
1283 ; AVX1-LABEL: @shuffle_v8i32_10324567
1285 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,3,2,4,5,6,7]
1288 ; AVX2-LABEL: @shuffle_v8i32_10324567
1290 ; AVX2-NEXT: vmovdqa .LCPI100_0(%rip), %ymm1
1291 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1293 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 6, i32 7>
1294 ret <8 x i32> %shuffle
1297 define <8 x i32> @shuffle_v8i32_11334567(<8 x i32> %a, <8 x i32> %b) {
1298 ; AVX1-LABEL: @shuffle_v8i32_11334567
1300 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,1,3,3,4,5,6,7]
1303 ; AVX2-LABEL: @shuffle_v8i32_11334567
1305 ; AVX2-NEXT: vmovdqa .LCPI101_0(%rip), %ymm1
1306 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1308 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 4, i32 5, i32 6, i32 7>
1309 ret <8 x i32> %shuffle
1312 define <8 x i32> @shuffle_v8i32_01235467(<8 x i32> %a, <8 x i32> %b) {
1313 ; AVX1-LABEL: @shuffle_v8i32_01235467
1315 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,1,2,3,5,4,6,7]
1318 ; AVX2-LABEL: @shuffle_v8i32_01235467
1320 ; AVX2-NEXT: vmovdqa .LCPI102_0(%rip), %ymm1
1321 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1323 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
1324 ret <8 x i32> %shuffle
1327 define <8 x i32> @shuffle_v8i32_01235466(<8 x i32> %a, <8 x i32> %b) {
1328 ; AVX1-LABEL: @shuffle_v8i32_01235466
1330 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,1,2,3,5,4,6,6]
1333 ; AVX2-LABEL: @shuffle_v8i32_01235466
1335 ; AVX2-NEXT: vmovdqa .LCPI103_0(%rip), %ymm1
1336 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1338 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 6, i32 6>
1339 ret <8 x i32> %shuffle
1342 define <8 x i32> @shuffle_v8i32_002u6u44(<8 x i32> %a, <8 x i32> %b) {
1343 ; AVX1-LABEL: @shuffle_v8i32_002u6u44
1345 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,2,u,6,u,4,4]
1348 ; AVX2-LABEL: @shuffle_v8i32_002u6u44
1350 ; AVX2-NEXT: vmovdqa .LCPI104_0(%rip), %ymm1
1351 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1353 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 undef, i32 6, i32 undef, i32 4, i32 4>
1354 ret <8 x i32> %shuffle
1357 define <8 x i32> @shuffle_v8i32_00uu66uu(<8 x i32> %a, <8 x i32> %b) {
1358 ; AVX1-LABEL: @shuffle_v8i32_00uu66uu
1360 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,0,u,u,6,6,u,u]
1363 ; AVX2-LABEL: @shuffle_v8i32_00uu66uu
1365 ; AVX2-NEXT: vmovdqa .LCPI105_0(%rip), %ymm1
1366 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1368 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 undef, i32 undef, i32 6, i32 6, i32 undef, i32 undef>
1369 ret <8 x i32> %shuffle
1372 define <8 x i32> @shuffle_v8i32_103245uu(<8 x i32> %a, <8 x i32> %b) {
1373 ; AVX1-LABEL: @shuffle_v8i32_103245uu
1375 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,0,3,2,4,5,u,u]
1378 ; AVX2-LABEL: @shuffle_v8i32_103245uu
1380 ; AVX2-NEXT: vmovdqa .LCPI106_0(%rip), %ymm1
1381 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1383 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 undef, i32 undef>
1384 ret <8 x i32> %shuffle
1387 define <8 x i32> @shuffle_v8i32_1133uu67(<8 x i32> %a, <8 x i32> %b) {
1388 ; AVX1-LABEL: @shuffle_v8i32_1133uu67
1390 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[1,1,3,3,u,u,6,7]
1393 ; AVX2-LABEL: @shuffle_v8i32_1133uu67
1395 ; AVX2-NEXT: vmovdqa .LCPI107_0(%rip), %ymm1
1396 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1398 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 undef, i32 undef, i32 6, i32 7>
1399 ret <8 x i32> %shuffle
1402 define <8 x i32> @shuffle_v8i32_0uu354uu(<8 x i32> %a, <8 x i32> %b) {
1403 ; AVX1-LABEL: @shuffle_v8i32_0uu354uu
1405 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[0,u,u,3,5,4,u,u]
1408 ; AVX2-LABEL: @shuffle_v8i32_0uu354uu
1410 ; AVX2-NEXT: vmovdqa .LCPI108_0(%rip), %ymm1
1411 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1413 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 undef, i32 undef, i32 3, i32 5, i32 4, i32 undef, i32 undef>
1414 ret <8 x i32> %shuffle
1417 define <8 x i32> @shuffle_v8i32_uuu3uu66(<8 x i32> %a, <8 x i32> %b) {
1418 ; AVX1-LABEL: @shuffle_v8i32_uuu3uu66
1420 ; AVX1-NEXT: vpermilps {{.*}} # ymm0 = ymm0[u,u,u,3,u,u,6,6]
1423 ; AVX2-LABEL: @shuffle_v8i32_uuu3uu66
1425 ; AVX2-NEXT: vmovdqa .LCPI109_0(%rip), %ymm1
1426 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1428 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 3, i32 undef, i32 undef, i32 6, i32 6>
1429 ret <8 x i32> %shuffle