1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX1
3 ; RUN: llc < %s -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX2
5 target triple = "x86_64-unknown-unknown"
7 define <8 x float> @shuffle_v8f32_00000000(<8 x float> %a, <8 x float> %b) {
8 ; AVX1-LABEL: shuffle_v8f32_00000000:
10 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
11 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
14 ; AVX2-LABEL: shuffle_v8f32_00000000:
16 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
18 %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>
19 ret <8 x float> %shuffle
22 define <8 x float> @shuffle_v8f32_00000010(<8 x float> %a, <8 x float> %b) {
23 ; AVX1-LABEL: shuffle_v8f32_00000010:
25 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
26 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
27 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
30 ; AVX2-LABEL: shuffle_v8f32_00000010:
32 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,0,1,0]
33 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
35 %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>
36 ret <8 x float> %shuffle
39 define <8 x float> @shuffle_v8f32_00000200(<8 x float> %a, <8 x float> %b) {
40 ; AVX1-LABEL: shuffle_v8f32_00000200:
42 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
43 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,2,0,0]
44 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
47 ; AVX2-LABEL: shuffle_v8f32_00000200:
49 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,2,0,0]
50 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
52 %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>
53 ret <8 x float> %shuffle
56 define <8 x float> @shuffle_v8f32_00003000(<8 x float> %a, <8 x float> %b) {
57 ; AVX1-LABEL: shuffle_v8f32_00003000:
59 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
60 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,0,0,0]
61 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
64 ; AVX2-LABEL: shuffle_v8f32_00003000:
66 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,0,0,3,0,0,0]
67 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
69 %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>
70 ret <8 x float> %shuffle
73 define <8 x float> @shuffle_v8f32_00040000(<8 x float> %a, <8 x float> %b) {
74 ; AVX1-LABEL: shuffle_v8f32_00040000:
76 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,3]
77 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
78 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
79 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
82 ; AVX2-LABEL: shuffle_v8f32_00040000:
84 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,0,4,0,0,0,0]
85 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
87 %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>
88 ret <8 x float> %shuffle
91 define <8 x float> @shuffle_v8f32_00500000(<8 x float> %a, <8 x float> %b) {
92 ; AVX1-LABEL: shuffle_v8f32_00500000:
94 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
95 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
96 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,4,4]
99 ; AVX2-LABEL: shuffle_v8f32_00500000:
101 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,5,0,0,0,0,0]
102 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
104 %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>
105 ret <8 x float> %shuffle
108 define <8 x float> @shuffle_v8f32_06000000(<8 x float> %a, <8 x float> %b) {
109 ; AVX1-LABEL: shuffle_v8f32_06000000:
111 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
112 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
113 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,2,0,0,4,4,4,4]
116 ; AVX2-LABEL: shuffle_v8f32_06000000:
118 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,6,0,0,0,0,0,0]
119 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
121 %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>
122 ret <8 x float> %shuffle
125 define <8 x float> @shuffle_v8f32_70000000(<8 x float> %a, <8 x float> %b) {
126 ; AVX1-LABEL: shuffle_v8f32_70000000:
128 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
129 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
130 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,0,0,0,4,4,4,4]
133 ; AVX2-LABEL: shuffle_v8f32_70000000:
135 ; AVX2-NEXT: movl $7, %eax
136 ; AVX2-NEXT: vmovd %eax, %xmm1
137 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
139 %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>
140 ret <8 x float> %shuffle
143 define <8 x float> @shuffle_v8f32_01014545(<8 x float> %a, <8 x float> %b) {
144 ; ALL-LABEL: shuffle_v8f32_01014545:
146 ; ALL-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
148 %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>
149 ret <8 x float> %shuffle
152 define <8 x float> @shuffle_v8f32_00112233(<8 x float> %a, <8 x float> %b) {
153 ; AVX1-LABEL: shuffle_v8f32_00112233:
155 ; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm0[0,0,1,1]
156 ; AVX1-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2,2,3,3]
157 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
160 ; AVX2-LABEL: shuffle_v8f32_00112233:
162 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,1,1,2,2,3,3]
163 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
165 %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>
166 ret <8 x float> %shuffle
169 define <8 x float> @shuffle_v8f32_00001111(<8 x float> %a, <8 x float> %b) {
170 ; AVX1-LABEL: shuffle_v8f32_00001111:
172 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
173 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
174 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
177 ; AVX2-LABEL: shuffle_v8f32_00001111:
179 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,0,0,1,1,1,1]
180 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
182 %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>
183 ret <8 x float> %shuffle
186 define <8 x float> @shuffle_v8f32_81a3c5e7(<8 x float> %a, <8 x float> %b) {
187 ; ALL-LABEL: shuffle_v8f32_81a3c5e7:
189 ; ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
191 %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>
192 ret <8 x float> %shuffle
195 define <8 x float> @shuffle_v8f32_08080808(<8 x float> %a, <8 x float> %b) {
196 ; AVX1-LABEL: shuffle_v8f32_08080808:
198 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,0,2,0]
199 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
200 ; AVX1-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
201 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
202 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
205 ; AVX2-LABEL: shuffle_v8f32_08080808:
207 ; AVX2-NEXT: vbroadcastss %xmm1, %ymm1
208 ; AVX2-NEXT: vbroadcastsd %xmm0, %ymm0
209 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
211 %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>
212 ret <8 x float> %shuffle
215 define <8 x float> @shuffle_v8f32_08084c4c(<8 x float> %a, <8 x float> %b) {
216 ; ALL-LABEL: shuffle_v8f32_08084c4c:
218 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
219 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
221 %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>
222 ret <8 x float> %shuffle
225 define <8 x float> @shuffle_v8f32_8823cc67(<8 x float> %a, <8 x float> %b) {
226 ; ALL-LABEL: shuffle_v8f32_8823cc67:
228 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,0],ymm0[2,3],ymm1[4,4],ymm0[6,7]
230 %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>
231 ret <8 x float> %shuffle
234 define <8 x float> @shuffle_v8f32_9832dc76(<8 x float> %a, <8 x float> %b) {
235 ; ALL-LABEL: shuffle_v8f32_9832dc76:
237 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[3,2],ymm1[5,4],ymm0[7,6]
239 %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>
240 ret <8 x float> %shuffle
243 define <8 x float> @shuffle_v8f32_9810dc54(<8 x float> %a, <8 x float> %b) {
244 ; ALL-LABEL: shuffle_v8f32_9810dc54:
246 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
248 %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>
249 ret <8 x float> %shuffle
252 define <8 x float> @shuffle_v8f32_08194c5d(<8 x float> %a, <8 x float> %b) {
253 ; ALL-LABEL: shuffle_v8f32_08194c5d:
255 ; ALL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
257 %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>
258 ret <8 x float> %shuffle
261 define <8 x float> @shuffle_v8f32_2a3b6e7f(<8 x float> %a, <8 x float> %b) {
262 ; ALL-LABEL: shuffle_v8f32_2a3b6e7f:
264 ; ALL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
266 %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>
267 ret <8 x float> %shuffle
270 define <8 x float> @shuffle_v8f32_08192a3b(<8 x float> %a, <8 x float> %b) {
271 ; AVX1-LABEL: shuffle_v8f32_08192a3b:
273 ; AVX1-NEXT: vunpckhps {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
274 ; AVX1-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
275 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
278 ; AVX2-LABEL: shuffle_v8f32_08192a3b:
280 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <u,0,u,1,u,2,u,3>
281 ; AVX2-NEXT: vpermps %ymm1, %ymm2, %ymm1
282 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <0,u,1,u,2,u,3,u>
283 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
284 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
286 %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>
287 ret <8 x float> %shuffle
290 define <8 x float> @shuffle_v8f32_08991abb(<8 x float> %a, <8 x float> %b) {
291 ; AVX1-LABEL: shuffle_v8f32_08991abb:
293 ; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm0[0,0],xmm1[0,0]
294 ; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[1,1]
295 ; AVX1-NEXT: vblendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
296 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,2,3,3]
297 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
300 ; AVX2-LABEL: shuffle_v8f32_08991abb:
302 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <0,u,u,u,1,u,u,u>
303 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
304 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <u,0,1,1,u,2,3,3>
305 ; AVX2-NEXT: vpermps %ymm1, %ymm2, %ymm1
306 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
308 %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>
309 ret <8 x float> %shuffle
312 define <8 x float> @shuffle_v8f32_091b2d3f(<8 x float> %a, <8 x float> %b) {
313 ; AVX1-LABEL: shuffle_v8f32_091b2d3f:
315 ; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[0,1,1,3]
316 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,1,3,3]
317 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
318 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
321 ; AVX2-LABEL: shuffle_v8f32_091b2d3f:
323 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <0,u,1,u,2,u,3,u>
324 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
325 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
327 %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>
328 ret <8 x float> %shuffle
331 define <8 x float> @shuffle_v8f32_09ab1def(<8 x float> %a, <8 x float> %b) {
332 ; AVX1-LABEL: shuffle_v8f32_09ab1def:
334 ; AVX1-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
335 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
336 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
339 ; AVX2-LABEL: shuffle_v8f32_09ab1def:
341 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <0,u,u,u,1,u,u,u>
342 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
343 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
345 %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>
346 ret <8 x float> %shuffle
349 define <8 x float> @shuffle_v8f32_00014445(<8 x float> %a, <8 x float> %b) {
350 ; ALL-LABEL: shuffle_v8f32_00014445:
352 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
354 %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>
355 ret <8 x float> %shuffle
358 define <8 x float> @shuffle_v8f32_00204464(<8 x float> %a, <8 x float> %b) {
359 ; ALL-LABEL: shuffle_v8f32_00204464:
361 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
363 %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>
364 ret <8 x float> %shuffle
367 define <8 x float> @shuffle_v8f32_03004744(<8 x float> %a, <8 x float> %b) {
368 ; ALL-LABEL: shuffle_v8f32_03004744:
370 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
372 %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>
373 ret <8 x float> %shuffle
376 define <8 x float> @shuffle_v8f32_10005444(<8 x float> %a, <8 x float> %b) {
377 ; ALL-LABEL: shuffle_v8f32_10005444:
379 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
381 %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>
382 ret <8 x float> %shuffle
385 define <8 x float> @shuffle_v8f32_22006644(<8 x float> %a, <8 x float> %b) {
386 ; ALL-LABEL: shuffle_v8f32_22006644:
388 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
390 %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>
391 ret <8 x float> %shuffle
394 define <8 x float> @shuffle_v8f32_33307774(<8 x float> %a, <8 x float> %b) {
395 ; ALL-LABEL: shuffle_v8f32_33307774:
397 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
399 %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>
400 ret <8 x float> %shuffle
403 define <8 x float> @shuffle_v8f32_32107654(<8 x float> %a, <8 x float> %b) {
404 ; ALL-LABEL: shuffle_v8f32_32107654:
406 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
408 %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>
409 ret <8 x float> %shuffle
412 define <8 x float> @shuffle_v8f32_00234467(<8 x float> %a, <8 x float> %b) {
413 ; ALL-LABEL: shuffle_v8f32_00234467:
415 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
417 %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>
418 ret <8 x float> %shuffle
421 define <8 x float> @shuffle_v8f32_00224466(<8 x float> %a, <8 x float> %b) {
422 ; ALL-LABEL: shuffle_v8f32_00224466:
424 ; ALL-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
426 %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>
427 ret <8 x float> %shuffle
430 define <8 x float> @shuffle_v8f32_10325476(<8 x float> %a, <8 x float> %b) {
431 ; ALL-LABEL: shuffle_v8f32_10325476:
433 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
435 %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>
436 ret <8 x float> %shuffle
439 define <8 x float> @shuffle_v8f32_11335577(<8 x float> %a, <8 x float> %b) {
440 ; ALL-LABEL: shuffle_v8f32_11335577:
442 ; ALL-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
444 %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>
445 ret <8 x float> %shuffle
448 define <8 x float> @shuffle_v8f32_10235467(<8 x float> %a, <8 x float> %b) {
449 ; ALL-LABEL: shuffle_v8f32_10235467:
451 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
453 %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>
454 ret <8 x float> %shuffle
457 define <8 x float> @shuffle_v8f32_10225466(<8 x float> %a, <8 x float> %b) {
458 ; ALL-LABEL: shuffle_v8f32_10225466:
460 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
462 %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>
463 ret <8 x float> %shuffle
466 define <8 x float> @shuffle_v8f32_00015444(<8 x float> %a, <8 x float> %b) {
467 ; ALL-LABEL: shuffle_v8f32_00015444:
469 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,5,4,4,4]
471 %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>
472 ret <8 x float> %shuffle
475 define <8 x float> @shuffle_v8f32_00204644(<8 x float> %a, <8 x float> %b) {
476 ; ALL-LABEL: shuffle_v8f32_00204644:
478 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,6,4,4]
480 %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>
481 ret <8 x float> %shuffle
484 define <8 x float> @shuffle_v8f32_03004474(<8 x float> %a, <8 x float> %b) {
485 ; ALL-LABEL: shuffle_v8f32_03004474:
487 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,4,7,4]
489 %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>
490 ret <8 x float> %shuffle
493 define <8 x float> @shuffle_v8f32_10004444(<8 x float> %a, <8 x float> %b) {
494 ; ALL-LABEL: shuffle_v8f32_10004444:
496 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,4,4,4,4]
498 %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>
499 ret <8 x float> %shuffle
502 define <8 x float> @shuffle_v8f32_22006446(<8 x float> %a, <8 x float> %b) {
503 ; ALL-LABEL: shuffle_v8f32_22006446:
505 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,4,4,6]
507 %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>
508 ret <8 x float> %shuffle
511 define <8 x float> @shuffle_v8f32_33307474(<8 x float> %a, <8 x float> %b) {
512 ; ALL-LABEL: shuffle_v8f32_33307474:
514 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,4,7,4]
516 %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>
517 ret <8 x float> %shuffle
520 define <8 x float> @shuffle_v8f32_32104567(<8 x float> %a, <8 x float> %b) {
521 ; ALL-LABEL: shuffle_v8f32_32104567:
523 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7]
525 %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>
526 ret <8 x float> %shuffle
529 define <8 x float> @shuffle_v8f32_00236744(<8 x float> %a, <8 x float> %b) {
530 ; ALL-LABEL: shuffle_v8f32_00236744:
532 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,6,7,4,4]
534 %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>
535 ret <8 x float> %shuffle
538 define <8 x float> @shuffle_v8f32_00226644(<8 x float> %a, <8 x float> %b) {
539 ; ALL-LABEL: shuffle_v8f32_00226644:
541 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,2,6,6,4,4]
543 %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>
544 ret <8 x float> %shuffle
547 define <8 x float> @shuffle_v8f32_10324567(<8 x float> %a, <8 x float> %b) {
548 ; ALL-LABEL: shuffle_v8f32_10324567:
550 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,6,7]
552 %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>
553 ret <8 x float> %shuffle
556 define <8 x float> @shuffle_v8f32_11334567(<8 x float> %a, <8 x float> %b) {
557 ; ALL-LABEL: shuffle_v8f32_11334567:
559 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7]
561 %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>
562 ret <8 x float> %shuffle
565 define <8 x float> @shuffle_v8f32_01235467(<8 x float> %a, <8 x float> %b) {
566 ; ALL-LABEL: shuffle_v8f32_01235467:
568 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,7]
570 %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>
571 ret <8 x float> %shuffle
574 define <8 x float> @shuffle_v8f32_01235466(<8 x float> %a, <8 x float> %b) {
575 ; ALL-LABEL: shuffle_v8f32_01235466:
577 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,6]
579 %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>
580 ret <8 x float> %shuffle
583 define <8 x float> @shuffle_v8f32_002u6u44(<8 x float> %a, <8 x float> %b) {
584 ; ALL-LABEL: shuffle_v8f32_002u6u44:
586 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,u,6,u,4,4]
588 %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>
589 ret <8 x float> %shuffle
592 define <8 x float> @shuffle_v8f32_00uu66uu(<8 x float> %a, <8 x float> %b) {
593 ; ALL-LABEL: shuffle_v8f32_00uu66uu:
595 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,u,u,6,6,u,u]
597 %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>
598 ret <8 x float> %shuffle
601 define <8 x float> @shuffle_v8f32_103245uu(<8 x float> %a, <8 x float> %b) {
602 ; ALL-LABEL: shuffle_v8f32_103245uu:
604 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,u,u]
606 %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>
607 ret <8 x float> %shuffle
610 define <8 x float> @shuffle_v8f32_1133uu67(<8 x float> %a, <8 x float> %b) {
611 ; ALL-LABEL: shuffle_v8f32_1133uu67:
613 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,u,u,6,7]
615 %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>
616 ret <8 x float> %shuffle
619 define <8 x float> @shuffle_v8f32_0uu354uu(<8 x float> %a, <8 x float> %b) {
620 ; ALL-LABEL: shuffle_v8f32_0uu354uu:
622 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,u,u,3,5,4,u,u]
624 %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>
625 ret <8 x float> %shuffle
628 define <8 x float> @shuffle_v8f32_uuu3uu66(<8 x float> %a, <8 x float> %b) {
629 ; ALL-LABEL: shuffle_v8f32_uuu3uu66:
631 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[u,u,u,3,u,u,6,6]
633 %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>
634 ret <8 x float> %shuffle
637 define <8 x float> @shuffle_v8f32_c348cda0(<8 x float> %a, <8 x float> %b) {
638 ; AVX1-LABEL: shuffle_v8f32_c348cda0:
640 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
641 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm2[0,0],ymm0[4,7],ymm2[4,4]
642 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
643 ; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
644 ; AVX1-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3]
645 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
648 ; AVX2-LABEL: shuffle_v8f32_c348cda0:
650 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <u,3,4,u,u,u,u,0>
651 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
652 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <4,u,u,0,4,5,2,u>
653 ; AVX2-NEXT: vpermps %ymm1, %ymm2, %ymm1
654 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
656 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 12, i32 3, i32 4, i32 8, i32 12, i32 13, i32 10, i32 0>
657 ret <8 x float> %shuffle
660 define <8 x float> @shuffle_v8f32_f511235a(<8 x float> %a, <8 x float> %b) {
661 ; AVX1-LABEL: shuffle_v8f32_f511235a:
663 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
664 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
665 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,1,1,4,5,5,5]
666 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
667 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
668 ; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[3,1,2,2,7,5,6,6]
669 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
672 ; AVX2-LABEL: shuffle_v8f32_f511235a:
674 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <7,u,u,u,u,u,u,2>
675 ; AVX2-NEXT: vpermps %ymm1, %ymm2, %ymm1
676 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <u,5,1,1,2,3,5,u>
677 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
678 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
680 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 15, i32 5, i32 1, i32 1, i32 2, i32 3, i32 5, i32 10>
681 ret <8 x float> %shuffle
684 define <8 x float> @shuffle_v8f32_32103210(<8 x float> %a, <8 x float> %b) {
685 ; AVX1-LABEL: shuffle_v8f32_32103210:
687 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
688 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
691 ; AVX2-LABEL: shuffle_v8f32_32103210:
693 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [3,2,1,0,3,2,1,0]
694 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
696 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>
697 ret <8 x float> %shuffle
700 define <8 x float> @shuffle_v8f32_76547654(<8 x float> %a, <8 x float> %b) {
701 ; AVX1-LABEL: shuffle_v8f32_76547654:
703 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
704 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
705 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
708 ; AVX2-LABEL: shuffle_v8f32_76547654:
710 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [7,6,5,4,7,6,5,4]
711 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
713 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 7, i32 6, i32 5, i32 4>
714 ret <8 x float> %shuffle
717 define <8 x float> @shuffle_v8f32_76543210(<8 x float> %a, <8 x float> %b) {
718 ; AVX1-LABEL: shuffle_v8f32_76543210:
720 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
721 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
724 ; AVX2-LABEL: shuffle_v8f32_76543210:
726 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [7,6,5,4,3,2,1,0]
727 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
729 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
730 ret <8 x float> %shuffle
733 define <8 x float> @shuffle_v8f32_3210ba98(<8 x float> %a, <8 x float> %b) {
734 ; ALL-LABEL: shuffle_v8f32_3210ba98:
736 ; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
737 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
739 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 11, i32 10, i32 9, i32 8>
740 ret <8 x float> %shuffle
743 define <8 x float> @shuffle_v8f32_3210fedc(<8 x float> %a, <8 x float> %b) {
744 ; ALL-LABEL: shuffle_v8f32_3210fedc:
746 ; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
747 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
749 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 15, i32 14, i32 13, i32 12>
750 ret <8 x float> %shuffle
753 define <8 x float> @shuffle_v8f32_7654fedc(<8 x float> %a, <8 x float> %b) {
754 ; ALL-LABEL: shuffle_v8f32_7654fedc:
756 ; ALL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
757 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
759 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 15, i32 14, i32 13, i32 12>
760 ret <8 x float> %shuffle
763 define <8 x float> @shuffle_v8f32_fedc7654(<8 x float> %a, <8 x float> %b) {
764 ; ALL-LABEL: shuffle_v8f32_fedc7654:
766 ; ALL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
767 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
769 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 15, i32 14, i32 13, i32 12, i32 7, i32 6, i32 5, i32 4>
770 ret <8 x float> %shuffle
773 define <8 x float> @PR21138(<8 x float> %truc, <8 x float> %tchose) {
774 ; AVX1-LABEL: PR21138:
776 ; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
777 ; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm2[1,3]
778 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
779 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
780 ; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
781 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
784 ; AVX2-LABEL: PR21138:
786 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <u,u,u,u,1,3,5,7>
787 ; AVX2-NEXT: vpermps %ymm1, %ymm2, %ymm1
788 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <1,3,5,7,u,u,u,u>
789 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
790 ; AVX2-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
792 %shuffle = shufflevector <8 x float> %truc, <8 x float> %tchose, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
793 ret <8 x float> %shuffle
796 define <8 x float> @shuffle_v8f32_ba987654(<8 x float> %a, <8 x float> %b) {
797 ; ALL-LABEL: shuffle_v8f32_ba987654:
799 ; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
800 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
802 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4>
803 ret <8 x float> %shuffle
806 define <8 x float> @shuffle_v8f32_ba983210(<8 x float> %a, <8 x float> %b) {
807 ; ALL-LABEL: shuffle_v8f32_ba983210:
809 ; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
810 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
812 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4>
813 ret <8 x float> %shuffle
816 define <8 x float> @shuffle_v8f32_80u1c4u5(<8 x float> %a, <8 x float> %b) {
817 ; ALL-LABEL: shuffle_v8f32_80u1c4u5:
819 ; ALL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
821 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 8, i32 0, i32 undef, i32 1, i32 12, i32 4, i32 undef, i32 5>
822 ret <8 x float> %shuffle
825 define <8 x float> @shuffle_v8f32_a2u3e6f7(<8 x float> %a, <8 x float> %b) {
826 ; ALL-LABEL: shuffle_v8f32_a2u3e6f7:
828 ; ALL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
830 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 10, i32 2, i32 undef, i32 3, i32 14, i32 6, i32 15, i32 7>
831 ret <8 x float> %shuffle
834 define <8 x float> @shuffle_v8f32_uuuu1111(<8 x float> %a, <8 x float> %b) {
835 ; ALL-LABEL: shuffle_v8f32_uuuu1111:
837 ; ALL-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
838 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
840 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 1, i32 1, i32 1, i32 1>
841 ret <8 x float> %shuffle
844 define <8 x float> @shuffle_v8f32_44444444(<8 x float> %a, <8 x float> %b) {
845 ; AVX1-LABEL: shuffle_v8f32_44444444:
847 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
848 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
849 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
852 ; AVX2-LABEL: shuffle_v8f32_44444444:
854 ; AVX2-NEXT: vbroadcastss {{.*}}(%rip), %ymm1
855 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
857 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4>
858 ret <8 x float> %shuffle
861 define <8 x float> @shuffle_v8f32_1188uuuu(<8 x float> %a, <8 x float> %b) {
862 ; ALL-LABEL: shuffle_v8f32_1188uuuu:
864 ; ALL-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,0]
866 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 1, i32 8, i32 8, i32 undef, i32 undef, i32 undef, i32 undef>
867 ret <8 x float> %shuffle
870 define <8 x float> @shuffle_v8f32_uuuu3210(<8 x float> %a, <8 x float> %b) {
871 ; ALL-LABEL: shuffle_v8f32_uuuu3210:
873 ; ALL-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
874 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
876 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 3, i32 2, i32 1, i32 0>
877 ret <8 x float> %shuffle
880 define <8 x float> @shuffle_v8f32_uuuu1188(<8 x float> %a, <8 x float> %b) {
881 ; ALL-LABEL: shuffle_v8f32_uuuu1188:
883 ; ALL-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,0]
884 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
886 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 1, i32 1, i32 8, i32 8>
887 ret <8 x float> %shuffle
890 define <8 x float> @shuffle_v8f32_1111uuuu(<8 x float> %a, <8 x float> %b) {
891 ; ALL-LABEL: shuffle_v8f32_1111uuuu:
893 ; ALL-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
895 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 undef, i32 undef, i32 undef, i32 undef>
896 ret <8 x float> %shuffle
899 define <8 x float> @shuffle_v8f32_5555uuuu(<8 x float> %a, <8 x float> %b) {
900 ; AVX1-LABEL: shuffle_v8f32_5555uuuu:
902 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
903 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
906 ; AVX2-LABEL: shuffle_v8f32_5555uuuu:
908 ; AVX2-NEXT: vbroadcastss {{.*}}(%rip), %ymm1
909 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
911 %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 undef, i32 undef, i32 undef, i32 undef>
912 ret <8 x float> %shuffle
915 define <8 x i32> @shuffle_v8i32_00000000(<8 x i32> %a, <8 x i32> %b) {
916 ; AVX1-LABEL: shuffle_v8i32_00000000:
918 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
919 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
922 ; AVX2-LABEL: shuffle_v8i32_00000000:
924 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
926 %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>
927 ret <8 x i32> %shuffle
930 define <8 x i32> @shuffle_v8i32_00000010(<8 x i32> %a, <8 x i32> %b) {
931 ; AVX1-LABEL: shuffle_v8i32_00000010:
933 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
934 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
935 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
938 ; AVX2-LABEL: shuffle_v8i32_00000010:
940 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,1,0]
941 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
943 %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>
944 ret <8 x i32> %shuffle
947 define <8 x i32> @shuffle_v8i32_00000200(<8 x i32> %a, <8 x i32> %b) {
948 ; AVX1-LABEL: shuffle_v8i32_00000200:
950 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
951 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,2,0,0]
952 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
955 ; AVX2-LABEL: shuffle_v8i32_00000200:
957 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,2,0,0]
958 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
960 %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>
961 ret <8 x i32> %shuffle
964 define <8 x i32> @shuffle_v8i32_00003000(<8 x i32> %a, <8 x i32> %b) {
965 ; AVX1-LABEL: shuffle_v8i32_00003000:
967 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
968 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,0,0,0]
969 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
972 ; AVX2-LABEL: shuffle_v8i32_00003000:
974 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,0,3,0,0,0]
975 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
977 %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>
978 ret <8 x i32> %shuffle
981 define <8 x i32> @shuffle_v8i32_00040000(<8 x i32> %a, <8 x i32> %b) {
982 ; AVX1-LABEL: shuffle_v8i32_00040000:
984 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,3]
985 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
986 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
987 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
990 ; AVX2-LABEL: shuffle_v8i32_00040000:
992 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,4,0,0,0,0]
993 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
995 %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>
996 ret <8 x i32> %shuffle
999 define <8 x i32> @shuffle_v8i32_00500000(<8 x i32> %a, <8 x i32> %b) {
1000 ; AVX1-LABEL: shuffle_v8i32_00500000:
1002 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
1003 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
1004 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,4,4]
1007 ; AVX2-LABEL: shuffle_v8i32_00500000:
1009 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,5,0,0,0,0,0]
1010 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1012 %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>
1013 ret <8 x i32> %shuffle
1016 define <8 x i32> @shuffle_v8i32_06000000(<8 x i32> %a, <8 x i32> %b) {
1017 ; AVX1-LABEL: shuffle_v8i32_06000000:
1019 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
1020 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
1021 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,2,0,0,4,4,4,4]
1024 ; AVX2-LABEL: shuffle_v8i32_06000000:
1026 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,6,0,0,0,0,0,0]
1027 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1029 %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>
1030 ret <8 x i32> %shuffle
1033 define <8 x i32> @shuffle_v8i32_70000000(<8 x i32> %a, <8 x i32> %b) {
1034 ; AVX1-LABEL: shuffle_v8i32_70000000:
1036 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
1037 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
1038 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,0,0,0,4,4,4,4]
1041 ; AVX2-LABEL: shuffle_v8i32_70000000:
1043 ; AVX2-NEXT: movl $7, %eax
1044 ; AVX2-NEXT: vmovd %eax, %xmm1
1045 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1047 %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>
1048 ret <8 x i32> %shuffle
1051 define <8 x i32> @shuffle_v8i32_01014545(<8 x i32> %a, <8 x i32> %b) {
1052 ; AVX1-LABEL: shuffle_v8i32_01014545:
1054 ; AVX1-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
1057 ; AVX2-LABEL: shuffle_v8i32_01014545:
1059 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
1061 %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>
1062 ret <8 x i32> %shuffle
1065 define <8 x i32> @shuffle_v8i32_00112233(<8 x i32> %a, <8 x i32> %b) {
1066 ; AVX1-LABEL: shuffle_v8i32_00112233:
1068 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,1,1]
1069 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
1070 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
1073 ; AVX2-LABEL: shuffle_v8i32_00112233:
1075 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,1,1,2,2,3,3]
1076 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1078 %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>
1079 ret <8 x i32> %shuffle
1082 define <8 x i32> @shuffle_v8i32_00001111(<8 x i32> %a, <8 x i32> %b) {
1083 ; AVX1-LABEL: shuffle_v8i32_00001111:
1085 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
1086 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
1087 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
1090 ; AVX2-LABEL: shuffle_v8i32_00001111:
1092 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,0,1,1,1,1]
1093 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1095 %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>
1096 ret <8 x i32> %shuffle
1099 define <8 x i32> @shuffle_v8i32_81a3c5e7(<8 x i32> %a, <8 x i32> %b) {
1100 ; AVX1-LABEL: shuffle_v8i32_81a3c5e7:
1102 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
1105 ; AVX2-LABEL: shuffle_v8i32_81a3c5e7:
1107 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
1109 %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>
1110 ret <8 x i32> %shuffle
1113 define <8 x i32> @shuffle_v8i32_08080808(<8 x i32> %a, <8 x i32> %b) {
1114 ; AVX1-LABEL: shuffle_v8i32_08080808:
1116 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,0,2,0]
1117 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
1118 ; AVX1-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
1119 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1120 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1123 ; AVX2-LABEL: shuffle_v8i32_08080808:
1125 ; AVX2-NEXT: vpbroadcastd %xmm1, %ymm1
1126 ; AVX2-NEXT: vpbroadcastq %xmm0, %ymm0
1127 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1129 %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>
1130 ret <8 x i32> %shuffle
1133 define <8 x i32> @shuffle_v8i32_08084c4c(<8 x i32> %a, <8 x i32> %b) {
1134 ; AVX1-LABEL: shuffle_v8i32_08084c4c:
1136 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
1137 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
1140 ; AVX2-LABEL: shuffle_v8i32_08084c4c:
1142 ; AVX2-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,0,2,0,4,4,6,4]
1143 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
1144 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1146 %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>
1147 ret <8 x i32> %shuffle
1150 define <8 x i32> @shuffle_v8i32_8823cc67(<8 x i32> %a, <8 x i32> %b) {
1151 ; AVX1-LABEL: shuffle_v8i32_8823cc67:
1153 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,0],ymm0[2,3],ymm1[4,4],ymm0[6,7]
1156 ; AVX2-LABEL: shuffle_v8i32_8823cc67:
1158 ; AVX2-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,0,2,3,4,4,6,7]
1159 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
1161 %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>
1162 ret <8 x i32> %shuffle
1165 define <8 x i32> @shuffle_v8i32_9832dc76(<8 x i32> %a, <8 x i32> %b) {
1166 ; AVX1-LABEL: shuffle_v8i32_9832dc76:
1168 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[3,2],ymm1[5,4],ymm0[7,6]
1171 ; AVX2-LABEL: shuffle_v8i32_9832dc76:
1173 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
1174 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
1176 %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>
1177 ret <8 x i32> %shuffle
1180 define <8 x i32> @shuffle_v8i32_9810dc54(<8 x i32> %a, <8 x i32> %b) {
1181 ; AVX1-LABEL: shuffle_v8i32_9810dc54:
1183 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
1186 ; AVX2-LABEL: shuffle_v8i32_9810dc54:
1188 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,1,0,4,5,5,4]
1189 ; AVX2-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,0,2,3,5,4,6,7]
1190 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
1192 %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>
1193 ret <8 x i32> %shuffle
1196 define <8 x i32> @shuffle_v8i32_08194c5d(<8 x i32> %a, <8 x i32> %b) {
1197 ; AVX1-LABEL: shuffle_v8i32_08194c5d:
1199 ; AVX1-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
1202 ; AVX2-LABEL: shuffle_v8i32_08194c5d:
1204 ; AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
1206 %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>
1207 ret <8 x i32> %shuffle
1210 define <8 x i32> @shuffle_v8i32_2a3b6e7f(<8 x i32> %a, <8 x i32> %b) {
1211 ; AVX1-LABEL: shuffle_v8i32_2a3b6e7f:
1213 ; AVX1-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
1216 ; AVX2-LABEL: shuffle_v8i32_2a3b6e7f:
1218 ; AVX2-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
1220 %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>
1221 ret <8 x i32> %shuffle
1224 define <8 x i32> @shuffle_v8i32_08192a3b(<8 x i32> %a, <8 x i32> %b) {
1225 ; AVX1-LABEL: shuffle_v8i32_08192a3b:
1227 ; AVX1-NEXT: vunpckhps {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
1228 ; AVX1-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
1229 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1232 ; AVX2-LABEL: shuffle_v8i32_08192a3b:
1234 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <u,0,u,1,u,2,u,3>
1235 ; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1
1236 ; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
1237 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1239 %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>
1240 ret <8 x i32> %shuffle
1243 define <8 x i32> @shuffle_v8i32_08991abb(<8 x i32> %a, <8 x i32> %b) {
1244 ; AVX1-LABEL: shuffle_v8i32_08991abb:
1246 ; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm0[0,0],xmm1[0,0]
1247 ; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[1,1]
1248 ; AVX1-NEXT: vblendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
1249 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,2,3,3]
1250 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
1253 ; AVX2-LABEL: shuffle_v8i32_08991abb:
1255 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <0,u,u,u,1,u,u,u>
1256 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
1257 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <u,0,1,1,u,2,3,3>
1258 ; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1
1259 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1261 %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>
1262 ret <8 x i32> %shuffle
1265 define <8 x i32> @shuffle_v8i32_091b2d3f(<8 x i32> %a, <8 x i32> %b) {
1266 ; AVX1-LABEL: shuffle_v8i32_091b2d3f:
1268 ; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[0,1,1,3]
1269 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,1,3,3]
1270 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
1271 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1274 ; AVX2-LABEL: shuffle_v8i32_091b2d3f:
1276 ; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
1277 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1279 %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>
1280 ret <8 x i32> %shuffle
1283 define <8 x i32> @shuffle_v8i32_09ab1def(<8 x i32> %a, <8 x i32> %b) {
1284 ; AVX1-LABEL: shuffle_v8i32_09ab1def:
1286 ; AVX1-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
1287 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1288 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1291 ; AVX2-LABEL: shuffle_v8i32_09ab1def:
1293 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <0,u,u,u,1,u,u,u>
1294 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
1295 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1297 %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>
1298 ret <8 x i32> %shuffle
1301 define <8 x i32> @shuffle_v8i32_00014445(<8 x i32> %a, <8 x i32> %b) {
1302 ; AVX1-LABEL: shuffle_v8i32_00014445:
1304 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
1307 ; AVX2-LABEL: shuffle_v8i32_00014445:
1309 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
1311 %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>
1312 ret <8 x i32> %shuffle
1315 define <8 x i32> @shuffle_v8i32_00204464(<8 x i32> %a, <8 x i32> %b) {
1316 ; AVX1-LABEL: shuffle_v8i32_00204464:
1318 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
1321 ; AVX2-LABEL: shuffle_v8i32_00204464:
1323 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
1325 %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>
1326 ret <8 x i32> %shuffle
1329 define <8 x i32> @shuffle_v8i32_03004744(<8 x i32> %a, <8 x i32> %b) {
1330 ; AVX1-LABEL: shuffle_v8i32_03004744:
1332 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
1335 ; AVX2-LABEL: shuffle_v8i32_03004744:
1337 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
1339 %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>
1340 ret <8 x i32> %shuffle
1343 define <8 x i32> @shuffle_v8i32_10005444(<8 x i32> %a, <8 x i32> %b) {
1344 ; AVX1-LABEL: shuffle_v8i32_10005444:
1346 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
1349 ; AVX2-LABEL: shuffle_v8i32_10005444:
1351 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
1353 %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>
1354 ret <8 x i32> %shuffle
1357 define <8 x i32> @shuffle_v8i32_22006644(<8 x i32> %a, <8 x i32> %b) {
1358 ; AVX1-LABEL: shuffle_v8i32_22006644:
1360 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
1363 ; AVX2-LABEL: shuffle_v8i32_22006644:
1365 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
1367 %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>
1368 ret <8 x i32> %shuffle
1371 define <8 x i32> @shuffle_v8i32_33307774(<8 x i32> %a, <8 x i32> %b) {
1372 ; AVX1-LABEL: shuffle_v8i32_33307774:
1374 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
1377 ; AVX2-LABEL: shuffle_v8i32_33307774:
1379 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
1381 %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>
1382 ret <8 x i32> %shuffle
1385 define <8 x i32> @shuffle_v8i32_32107654(<8 x i32> %a, <8 x i32> %b) {
1386 ; AVX1-LABEL: shuffle_v8i32_32107654:
1388 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1391 ; AVX2-LABEL: shuffle_v8i32_32107654:
1393 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1395 %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>
1396 ret <8 x i32> %shuffle
1399 define <8 x i32> @shuffle_v8i32_00234467(<8 x i32> %a, <8 x i32> %b) {
1400 ; AVX1-LABEL: shuffle_v8i32_00234467:
1402 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
1405 ; AVX2-LABEL: shuffle_v8i32_00234467:
1407 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
1409 %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>
1410 ret <8 x i32> %shuffle
1413 define <8 x i32> @shuffle_v8i32_00224466(<8 x i32> %a, <8 x i32> %b) {
1414 ; AVX1-LABEL: shuffle_v8i32_00224466:
1416 ; AVX1-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
1419 ; AVX2-LABEL: shuffle_v8i32_00224466:
1421 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
1423 %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>
1424 ret <8 x i32> %shuffle
1427 define <8 x i32> @shuffle_v8i32_10325476(<8 x i32> %a, <8 x i32> %b) {
1428 ; AVX1-LABEL: shuffle_v8i32_10325476:
1430 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
1433 ; AVX2-LABEL: shuffle_v8i32_10325476:
1435 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
1437 %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>
1438 ret <8 x i32> %shuffle
1441 define <8 x i32> @shuffle_v8i32_11335577(<8 x i32> %a, <8 x i32> %b) {
1442 ; AVX1-LABEL: shuffle_v8i32_11335577:
1444 ; AVX1-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
1447 ; AVX2-LABEL: shuffle_v8i32_11335577:
1449 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
1451 %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>
1452 ret <8 x i32> %shuffle
1455 define <8 x i32> @shuffle_v8i32_10235467(<8 x i32> %a, <8 x i32> %b) {
1456 ; AVX1-LABEL: shuffle_v8i32_10235467:
1458 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
1461 ; AVX2-LABEL: shuffle_v8i32_10235467:
1463 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
1465 %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>
1466 ret <8 x i32> %shuffle
1469 define <8 x i32> @shuffle_v8i32_10225466(<8 x i32> %a, <8 x i32> %b) {
1470 ; AVX1-LABEL: shuffle_v8i32_10225466:
1472 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
1475 ; AVX2-LABEL: shuffle_v8i32_10225466:
1477 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
1479 %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>
1480 ret <8 x i32> %shuffle
1483 define <8 x i32> @shuffle_v8i32_00015444(<8 x i32> %a, <8 x i32> %b) {
1484 ; AVX1-LABEL: shuffle_v8i32_00015444:
1486 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,5,4,4,4]
1489 ; AVX2-LABEL: shuffle_v8i32_00015444:
1491 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,1,5,4,4,4]
1492 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1494 %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>
1495 ret <8 x i32> %shuffle
1498 define <8 x i32> @shuffle_v8i32_00204644(<8 x i32> %a, <8 x i32> %b) {
1499 ; AVX1-LABEL: shuffle_v8i32_00204644:
1501 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,6,4,4]
1504 ; AVX2-LABEL: shuffle_v8i32_00204644:
1506 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,2,0,4,6,4,4]
1507 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1509 %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>
1510 ret <8 x i32> %shuffle
1513 define <8 x i32> @shuffle_v8i32_03004474(<8 x i32> %a, <8 x i32> %b) {
1514 ; AVX1-LABEL: shuffle_v8i32_03004474:
1516 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,4,7,4]
1519 ; AVX2-LABEL: shuffle_v8i32_03004474:
1521 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,3,0,0,4,4,7,4]
1522 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1524 %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>
1525 ret <8 x i32> %shuffle
1528 define <8 x i32> @shuffle_v8i32_10004444(<8 x i32> %a, <8 x i32> %b) {
1529 ; AVX1-LABEL: shuffle_v8i32_10004444:
1531 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,4,4,4,4]
1534 ; AVX2-LABEL: shuffle_v8i32_10004444:
1536 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [1,0,0,0,4,4,4,4]
1537 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1539 %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>
1540 ret <8 x i32> %shuffle
1543 define <8 x i32> @shuffle_v8i32_22006446(<8 x i32> %a, <8 x i32> %b) {
1544 ; AVX1-LABEL: shuffle_v8i32_22006446:
1546 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,4,4,6]
1549 ; AVX2-LABEL: shuffle_v8i32_22006446:
1551 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [2,2,0,0,6,4,4,6]
1552 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1554 %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>
1555 ret <8 x i32> %shuffle
1558 define <8 x i32> @shuffle_v8i32_33307474(<8 x i32> %a, <8 x i32> %b) {
1559 ; AVX1-LABEL: shuffle_v8i32_33307474:
1561 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,4,7,4]
1564 ; AVX2-LABEL: shuffle_v8i32_33307474:
1566 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [3,3,3,0,7,4,7,4]
1567 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1569 %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>
1570 ret <8 x i32> %shuffle
1573 define <8 x i32> @shuffle_v8i32_32104567(<8 x i32> %a, <8 x i32> %b) {
1574 ; AVX1-LABEL: shuffle_v8i32_32104567:
1576 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7]
1579 ; AVX2-LABEL: shuffle_v8i32_32104567:
1581 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [3,2,1,0,4,5,6,7]
1582 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1584 %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>
1585 ret <8 x i32> %shuffle
1588 define <8 x i32> @shuffle_v8i32_00236744(<8 x i32> %a, <8 x i32> %b) {
1589 ; AVX1-LABEL: shuffle_v8i32_00236744:
1591 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,6,7,4,4]
1594 ; AVX2-LABEL: shuffle_v8i32_00236744:
1596 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,2,3,6,7,4,4]
1597 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1599 %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>
1600 ret <8 x i32> %shuffle
1603 define <8 x i32> @shuffle_v8i32_00226644(<8 x i32> %a, <8 x i32> %b) {
1604 ; AVX1-LABEL: shuffle_v8i32_00226644:
1606 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,2,6,6,4,4]
1609 ; AVX2-LABEL: shuffle_v8i32_00226644:
1611 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,2,2,6,6,4,4]
1612 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1614 %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>
1615 ret <8 x i32> %shuffle
1618 define <8 x i32> @shuffle_v8i32_10324567(<8 x i32> %a, <8 x i32> %b) {
1619 ; AVX1-LABEL: shuffle_v8i32_10324567:
1621 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,6,7]
1624 ; AVX2-LABEL: shuffle_v8i32_10324567:
1626 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [1,0,3,2,4,5,6,7]
1627 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1629 %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>
1630 ret <8 x i32> %shuffle
1633 define <8 x i32> @shuffle_v8i32_11334567(<8 x i32> %a, <8 x i32> %b) {
1634 ; AVX1-LABEL: shuffle_v8i32_11334567:
1636 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7]
1639 ; AVX2-LABEL: shuffle_v8i32_11334567:
1641 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [1,1,3,3,4,5,6,7]
1642 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1644 %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>
1645 ret <8 x i32> %shuffle
1648 define <8 x i32> @shuffle_v8i32_01235467(<8 x i32> %a, <8 x i32> %b) {
1649 ; AVX1-LABEL: shuffle_v8i32_01235467:
1651 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,7]
1654 ; AVX2-LABEL: shuffle_v8i32_01235467:
1656 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,1,2,3,5,4,6,7]
1657 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1659 %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>
1660 ret <8 x i32> %shuffle
1663 define <8 x i32> @shuffle_v8i32_01235466(<8 x i32> %a, <8 x i32> %b) {
1664 ; AVX1-LABEL: shuffle_v8i32_01235466:
1666 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,6]
1669 ; AVX2-LABEL: shuffle_v8i32_01235466:
1671 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,1,2,3,5,4,6,6]
1672 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1674 %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>
1675 ret <8 x i32> %shuffle
1678 define <8 x i32> @shuffle_v8i32_002u6u44(<8 x i32> %a, <8 x i32> %b) {
1679 ; AVX1-LABEL: shuffle_v8i32_002u6u44:
1681 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,u,6,u,4,4]
1684 ; AVX2-LABEL: shuffle_v8i32_002u6u44:
1686 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <0,0,2,u,6,u,4,4>
1687 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1689 %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>
1690 ret <8 x i32> %shuffle
1693 define <8 x i32> @shuffle_v8i32_00uu66uu(<8 x i32> %a, <8 x i32> %b) {
1694 ; AVX1-LABEL: shuffle_v8i32_00uu66uu:
1696 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,u,u,6,6,u,u]
1699 ; AVX2-LABEL: shuffle_v8i32_00uu66uu:
1701 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <0,0,u,u,6,6,u,u>
1702 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1704 %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>
1705 ret <8 x i32> %shuffle
1708 define <8 x i32> @shuffle_v8i32_103245uu(<8 x i32> %a, <8 x i32> %b) {
1709 ; AVX1-LABEL: shuffle_v8i32_103245uu:
1711 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,u,u]
1714 ; AVX2-LABEL: shuffle_v8i32_103245uu:
1716 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <1,0,3,2,4,5,u,u>
1717 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1719 %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>
1720 ret <8 x i32> %shuffle
1723 define <8 x i32> @shuffle_v8i32_1133uu67(<8 x i32> %a, <8 x i32> %b) {
1724 ; AVX1-LABEL: shuffle_v8i32_1133uu67:
1726 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,u,u,6,7]
1729 ; AVX2-LABEL: shuffle_v8i32_1133uu67:
1731 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <1,1,3,3,u,u,6,7>
1732 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1734 %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>
1735 ret <8 x i32> %shuffle
1738 define <8 x i32> @shuffle_v8i32_0uu354uu(<8 x i32> %a, <8 x i32> %b) {
1739 ; AVX1-LABEL: shuffle_v8i32_0uu354uu:
1741 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,u,u,3,5,4,u,u]
1744 ; AVX2-LABEL: shuffle_v8i32_0uu354uu:
1746 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <0,u,u,3,5,4,u,u>
1747 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1749 %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>
1750 ret <8 x i32> %shuffle
1753 define <8 x i32> @shuffle_v8i32_uuu3uu66(<8 x i32> %a, <8 x i32> %b) {
1754 ; AVX1-LABEL: shuffle_v8i32_uuu3uu66:
1756 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[u,u,u,3,u,u,6,6]
1759 ; AVX2-LABEL: shuffle_v8i32_uuu3uu66:
1761 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,3,u,u,6,6>
1762 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1764 %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>
1765 ret <8 x i32> %shuffle
1768 define <8 x i32> @shuffle_v8i32_6caa87e5(<8 x i32> %a, <8 x i32> %b) {
1769 ; AVX1-LABEL: shuffle_v8i32_6caa87e5:
1771 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
1772 ; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[2,2],ymm2[4,4],ymm1[6,6]
1773 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
1774 ; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
1775 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1776 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
1779 ; AVX2-LABEL: shuffle_v8i32_6caa87e5:
1781 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <u,4,2,2,0,u,6,u>
1782 ; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1
1783 ; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,1,3,2]
1784 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
1786 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 6, i32 12, i32 10, i32 10, i32 8, i32 7, i32 14, i32 5>
1787 ret <8 x i32> %shuffle
1790 define <8 x i32> @shuffle_v8i32_32103210(<8 x i32> %a, <8 x i32> %b) {
1791 ; AVX1-LABEL: shuffle_v8i32_32103210:
1793 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
1794 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1797 ; AVX2-LABEL: shuffle_v8i32_32103210:
1799 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [3,2,1,0,3,2,1,0]
1800 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1802 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>
1803 ret <8 x i32> %shuffle
1806 define <8 x i32> @shuffle_v8i32_76547654(<8 x i32> %a, <8 x i32> %b) {
1807 ; AVX1-LABEL: shuffle_v8i32_76547654:
1809 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
1810 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
1811 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1814 ; AVX2-LABEL: shuffle_v8i32_76547654:
1816 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [7,6,5,4,7,6,5,4]
1817 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1819 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 7, i32 6, i32 5, i32 4>
1820 ret <8 x i32> %shuffle
1823 define <8 x i32> @shuffle_v8i32_76543210(<8 x i32> %a, <8 x i32> %b) {
1824 ; AVX1-LABEL: shuffle_v8i32_76543210:
1826 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
1827 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1830 ; AVX2-LABEL: shuffle_v8i32_76543210:
1832 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [7,6,5,4,3,2,1,0]
1833 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1835 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
1836 ret <8 x i32> %shuffle
1839 define <8 x i32> @shuffle_v8i32_3210ba98(<8 x i32> %a, <8 x i32> %b) {
1840 ; AVX1-LABEL: shuffle_v8i32_3210ba98:
1842 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
1843 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1846 ; AVX2-LABEL: shuffle_v8i32_3210ba98:
1848 ; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
1849 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1851 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 11, i32 10, i32 9, i32 8>
1852 ret <8 x i32> %shuffle
1855 define <8 x i32> @shuffle_v8i32_3210fedc(<8 x i32> %a, <8 x i32> %b) {
1856 ; AVX1-LABEL: shuffle_v8i32_3210fedc:
1858 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
1859 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1862 ; AVX2-LABEL: shuffle_v8i32_3210fedc:
1864 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
1865 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1867 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 15, i32 14, i32 13, i32 12>
1868 ret <8 x i32> %shuffle
1871 define <8 x i32> @shuffle_v8i32_7654fedc(<8 x i32> %a, <8 x i32> %b) {
1872 ; AVX1-LABEL: shuffle_v8i32_7654fedc:
1874 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
1875 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1878 ; AVX2-LABEL: shuffle_v8i32_7654fedc:
1880 ; AVX2-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
1881 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1883 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 15, i32 14, i32 13, i32 12>
1884 ret <8 x i32> %shuffle
1887 define <8 x i32> @shuffle_v8i32_fedc7654(<8 x i32> %a, <8 x i32> %b) {
1888 ; AVX1-LABEL: shuffle_v8i32_fedc7654:
1890 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
1891 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1894 ; AVX2-LABEL: shuffle_v8i32_fedc7654:
1896 ; AVX2-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
1897 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1899 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 15, i32 14, i32 13, i32 12, i32 7, i32 6, i32 5, i32 4>
1900 ret <8 x i32> %shuffle
1903 define <8 x i32> @shuffle_v8i32_ba987654(<8 x i32> %a, <8 x i32> %b) {
1904 ; AVX1-LABEL: shuffle_v8i32_ba987654:
1906 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
1907 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1910 ; AVX2-LABEL: shuffle_v8i32_ba987654:
1912 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
1913 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1915 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4>
1916 ret <8 x i32> %shuffle
1919 define <8 x i32> @shuffle_v8i32_ba983210(<8 x i32> %a, <8 x i32> %b) {
1920 ; AVX1-LABEL: shuffle_v8i32_ba983210:
1922 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
1923 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1926 ; AVX2-LABEL: shuffle_v8i32_ba983210:
1928 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
1929 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1931 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4>
1932 ret <8 x i32> %shuffle
1935 define <8 x i32> @shuffle_v8i32_zuu8zuuc(<8 x i32> %a) {
1936 ; AVX1-LABEL: shuffle_v8i32_zuu8zuuc:
1938 ; AVX1-NEXT: vxorps %xmm1, %xmm1, %xmm1
1939 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
1942 ; AVX2-LABEL: shuffle_v8i32_zuu8zuuc:
1944 ; AVX2-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,2,3],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19]
1946 %shuffle = shufflevector <8 x i32> zeroinitializer, <8 x i32> %a, <8 x i32> <i32 0, i32 undef, i32 undef, i32 8, i32 0, i32 undef, i32 undef, i32 12>
1947 ret <8 x i32> %shuffle
1950 define <8 x i32> @shuffle_v8i32_9ubzdefz(<8 x i32> %a) {
1951 ; AVX1-LABEL: shuffle_v8i32_9ubzdefz:
1953 ; AVX1-NEXT: vxorps %xmm1, %xmm1, %xmm1
1954 ; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
1955 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,2],ymm1[2,0],ymm0[5,6],ymm1[6,4]
1958 ; AVX2-LABEL: shuffle_v8i32_9ubzdefz:
1960 ; AVX2-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,ymm0[20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero
1962 %shuffle = shufflevector <8 x i32> zeroinitializer, <8 x i32> %a, <8 x i32> <i32 9, i32 undef, i32 11, i32 0, i32 13, i32 14, i32 15, i32 0>
1963 ret <8 x i32> %shuffle
1966 define <8 x i32> @shuffle_v8i32_80u1b4uu(<8 x i32> %a, <8 x i32> %b) {
1967 ; AVX1-LABEL: shuffle_v8i32_80u1b4uu:
1969 ; AVX1-NEXT: vunpcklps {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
1972 ; AVX2-LABEL: shuffle_v8i32_80u1b4uu:
1974 ; AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
1976 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 8, i32 0, i32 undef, i32 1, i32 12, i32 4, i32 undef, i32 undef>
1977 ret <8 x i32> %shuffle
1980 define <8 x i32> @shuffle_v8i32_uuuu1111(<8 x i32> %a, <8 x i32> %b) {
1981 ; AVX1-LABEL: shuffle_v8i32_uuuu1111:
1983 ; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
1984 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1987 ; AVX2-LABEL: shuffle_v8i32_uuuu1111:
1989 ; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
1990 ; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
1992 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 1, i32 1, i32 1, i32 1>
1993 ret <8 x i32> %shuffle
1996 define <8 x i32> @shuffle_v8i32_2222uuuu(<8 x i32> %a, <8 x i32> %b) {
1997 ; ALL-LABEL: shuffle_v8i32_2222uuuu:
1999 ; ALL-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
2001 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 undef, i32 undef, i32 undef, i32 undef>
2002 ret <8 x i32> %shuffle
2005 define <8 x i32> @shuffle_v8i32_2A3Buuuu(<8 x i32> %a, <8 x i32> %b) {
2006 ; ALL-LABEL: shuffle_v8i32_2A3Buuuu:
2008 ; ALL-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
2010 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 undef, i32 undef, i32 undef, i32 undef>
2011 ret <8 x i32> %shuffle
2014 define <8 x i32> @shuffle_v8i32_44444444(<8 x i32> %a, <8 x i32> %b) {
2015 ; AVX1-LABEL: shuffle_v8i32_44444444:
2017 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
2018 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
2019 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
2022 ; AVX2-LABEL: shuffle_v8i32_44444444:
2024 ; AVX2-NEXT: vpbroadcastd {{.*}}(%rip), %ymm1
2025 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
2027 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4>
2028 ret <8 x i32> %shuffle
2031 define <8 x i32> @shuffle_v8i32_5555uuuu(<8 x i32> %a, <8 x i32> %b) {
2032 ; AVX1-LABEL: shuffle_v8i32_5555uuuu:
2034 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
2035 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
2038 ; AVX2-LABEL: shuffle_v8i32_5555uuuu:
2040 ; AVX2-NEXT: vpbroadcastd {{.*}}(%rip), %ymm1
2041 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
2043 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 undef, i32 undef, i32 undef, i32 undef>
2044 ret <8 x i32> %shuffle
2047 define <8 x float> @splat_mem_v8f32_2(float* %p) {
2048 ; ALL-LABEL: splat_mem_v8f32_2:
2050 ; ALL-NEXT: vbroadcastss (%rdi), %ymm0
2052 %1 = load float, float* %p
2053 %2 = insertelement <4 x float> undef, float %1, i32 0
2054 %3 = shufflevector <4 x float> %2, <4 x float> undef, <8 x i32> zeroinitializer
2058 define <8 x float> @splat_v8f32(<4 x float> %r) {
2059 ; AVX1-LABEL: splat_v8f32:
2061 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
2062 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
2065 ; AVX2-LABEL: splat_v8f32:
2067 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
2069 %1 = shufflevector <4 x float> %r, <4 x float> undef, <8 x i32> zeroinitializer
2074 ; Shuffle to logical bit shifts
2077 define <8 x i32> @shuffle_v8i32_z0U2zUz6(<8 x i32> %a) {
2078 ; AVX1-LABEL: shuffle_v8i32_z0U2zUz6:
2080 ; AVX1-NEXT: vxorps %xmm1, %xmm1, %xmm1
2081 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2],ymm1[0,2],ymm0[4,6],ymm1[4,6]
2082 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0,3,1,6,4,7,5]
2085 ; AVX2-LABEL: shuffle_v8i32_z0U2zUz6:
2087 ; AVX2-NEXT: vpsllq $32, %ymm0, %ymm0
2089 %shuffle = shufflevector <8 x i32> %a, <8 x i32> zeroinitializer, <8 x i32> <i32 8, i32 0, i32 undef, i32 2, i32 8, i32 undef, i32 8, i32 6>
2090 ret <8 x i32> %shuffle
2093 define <8 x i32> @shuffle_v8i32_1U3z5zUU(<8 x i32> %a) {
2094 ; AVX1-LABEL: shuffle_v8i32_1U3z5zUU:
2096 ; AVX1-NEXT: vxorps %xmm1, %xmm1, %xmm1
2097 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm1[1,3],ymm0[5,7],ymm1[5,7]
2098 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
2101 ; AVX2-LABEL: shuffle_v8i32_1U3z5zUU:
2103 ; AVX2-NEXT: vpsrlq $32, %ymm0, %ymm0
2105 %shuffle = shufflevector <8 x i32> %a, <8 x i32> zeroinitializer, <8 x i32> <i32 1, i32 undef, i32 3, i32 8, i32 5, i32 8, i32 undef, i32 undef>
2106 ret <8 x i32> %shuffle
2109 define <8 x i32> @shuffle_v8i32_B012F456(<8 x i32> %a, <8 x i32> %b) {
2110 ; AVX1-LABEL: shuffle_v8i32_B012F456:
2112 ; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
2113 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,2],ymm0[1,2],ymm1[4,6],ymm0[5,6]
2116 ; AVX2-LABEL: shuffle_v8i32_B012F456:
2118 ; AVX2-NEXT: vpalignr {{.*#+}} ymm0 = ymm1[12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10,11],ymm1[28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26,27]
2120 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 11, i32 0, i32 1, i32 2, i32 15, i32 4, i32 5, i32 6>
2121 ret <8 x i32> %shuffle
2124 define <8 x i32> @shuffle_v8i32_1238567C(<8 x i32> %a, <8 x i32> %b) {
2125 ; AVX1-LABEL: shuffle_v8i32_1238567C:
2127 ; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm0[3,0],ymm1[4,4],ymm0[7,4]
2128 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,2],ymm1[2,0],ymm0[5,6],ymm1[6,4]
2131 ; AVX2-LABEL: shuffle_v8i32_1238567C:
2133 ; AVX2-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0,1,2,3],ymm0[20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16,17,18,19]
2135 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 2, i32 3, i32 8, i32 5, i32 6, i32 7, i32 12>
2136 ret <8 x i32> %shuffle
2139 define <8 x i32> @shuffle_v8i32_9AB0DEF4(<8 x i32> %a, <8 x i32> %b) {
2140 ; AVX1-LABEL: shuffle_v8i32_9AB0DEF4:
2142 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[3,0],ymm0[4,4],ymm1[7,4]
2143 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,2],ymm0[2,0],ymm1[5,6],ymm0[6,4]
2146 ; AVX2-LABEL: shuffle_v8i32_9AB0DEF4:
2148 ; AVX2-NEXT: vpalignr {{.*#+}} ymm0 = ymm1[4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1,2,3],ymm1[20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17,18,19]
2150 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 9, i32 10, i32 11, i32 0, i32 13, i32 14, i32 15, i32 4>
2151 ret <8 x i32> %shuffle
2154 define <8 x i32> @shuffle_v8i32_389A7CDE(<8 x i32> %a, <8 x i32> %b) {
2155 ; AVX1-LABEL: shuffle_v8i32_389A7CDE:
2157 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
2158 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2],ymm1[1,2],ymm0[4,6],ymm1[5,6]
2161 ; AVX2-LABEL: shuffle_v8i32_389A7CDE:
2163 ; AVX2-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
2165 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 8, i32 9, i32 10, i32 7, i32 12, i32 13, i32 14>
2166 ret <8 x i32> %shuffle
2169 define <8 x i32> @shuffle_v8i32_30127456(<8 x i32> %a, <8 x i32> %b) {
2170 ; AVX1-LABEL: shuffle_v8i32_30127456:
2172 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,0,1,2,7,4,5,6]
2175 ; AVX2-LABEL: shuffle_v8i32_30127456:
2177 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,0,1,2,7,4,5,6]
2179 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 0, i32 1, i32 2, i32 7, i32 4, i32 5, i32 6>
2180 ret <8 x i32> %shuffle
2183 define <8 x i32> @shuffle_v8i32_12305674(<8 x i32> %a, <8 x i32> %b) {
2184 ; AVX1-LABEL: shuffle_v8i32_12305674:
2186 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,2,3,0,5,6,7,4]
2189 ; AVX2-LABEL: shuffle_v8i32_12305674:
2191 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,2,3,0,5,6,7,4]
2193 %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 2, i32 3, i32 0, i32 5, i32 6, i32 7, i32 4>
2194 ret <8 x i32> %shuffle
2197 define <8x float> @concat_v2f32_1(<2 x float>* %tmp64, <2 x float>* %tmp65) {
2198 ; ALL-LABEL: concat_v2f32_1:
2199 ; ALL: # BB#0: # %entry
2200 ; ALL-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
2201 ; ALL-NEXT: vmovhpd (%rsi), %xmm0, %xmm0
2204 %tmp74 = load <2 x float>, <2 x float>* %tmp65, align 8
2205 %tmp72 = load <2 x float>, <2 x float>* %tmp64, align 8
2206 %tmp73 = shufflevector <2 x float> %tmp72, <2 x float> undef, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
2207 %tmp75 = shufflevector <2 x float> %tmp74, <2 x float> undef, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
2208 %tmp76 = shufflevector <8 x float> %tmp73, <8 x float> %tmp75, <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 undef>
2209 ret <8 x float> %tmp76
2212 define <8x float> @concat_v2f32_2(<2 x float>* %tmp64, <2 x float>* %tmp65) {
2213 ; ALL-LABEL: concat_v2f32_2:
2214 ; ALL: # BB#0: # %entry
2215 ; ALL-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
2216 ; ALL-NEXT: vmovhpd (%rsi), %xmm0, %xmm0
2219 %tmp74 = load <2 x float>, <2 x float>* %tmp65, align 8
2220 %tmp72 = load <2 x float>, <2 x float>* %tmp64, align 8
2221 %tmp76 = shufflevector <2 x float> %tmp72, <2 x float> %tmp74, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
2222 ret <8 x float> %tmp76
2225 define <8x float> @concat_v2f32_3(<2 x float>* %tmp64, <2 x float>* %tmp65) {
2226 ; ALL-LABEL: concat_v2f32_3:
2227 ; ALL: # BB#0: # %entry
2228 ; ALL-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
2229 ; ALL-NEXT: vmovhpd (%rsi), %xmm0, %xmm0
2232 %tmp74 = load <2 x float>, <2 x float>* %tmp65, align 8
2233 %tmp72 = load <2 x float>, <2 x float>* %tmp64, align 8
2234 %tmp76 = shufflevector <2 x float> %tmp72, <2 x float> %tmp74, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
2235 %res = shufflevector <4 x float> %tmp76, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
2236 ret <8 x float> %res
2239 define <8 x i32> @insert_mem_and_zero_v8i32(i32* %ptr) {
2240 ; ALL-LABEL: insert_mem_and_zero_v8i32:
2242 ; ALL-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
2244 %a = load i32, i32* %ptr
2245 %v = insertelement <8 x i32> undef, i32 %a, i32 0
2246 %shuffle = shufflevector <8 x i32> %v, <8 x i32> zeroinitializer, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
2247 ret <8 x i32> %shuffle
2250 define <8 x i32> @concat_v8i32_0123CDEF(<8 x i32> %a, <8 x i32> %b) {
2251 ; AVX1-LABEL: concat_v8i32_0123CDEF:
2253 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
2256 ; AVX2-LABEL: concat_v8i32_0123CDEF:
2258 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
2260 %alo = shufflevector <8 x i32> %a, <8 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
2261 %bhi = shufflevector <8 x i32> %b, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
2262 %shuf = shufflevector <4 x i32> %alo, <4 x i32> %bhi, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
2266 define <8 x i32> @concat_v8i32_4567CDEF_bc(<8 x i32> %a0, <8 x i32> %a1) {
2267 ; ALL-LABEL: concat_v8i32_4567CDEF_bc:
2269 ; ALL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
2271 %a0hi = shufflevector <8 x i32> %a0, <8 x i32> %a1, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
2272 %a1hi = shufflevector <8 x i32> %a0, <8 x i32> %a1, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
2273 %bc0hi = bitcast <4 x i32> %a0hi to <2 x i64>
2274 %bc1hi = bitcast <4 x i32> %a1hi to <2 x i64>
2275 %shuffle64 = shufflevector <2 x i64> %bc0hi, <2 x i64> %bc1hi, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
2276 %shuffle32 = bitcast <4 x i64> %shuffle64 to <8 x i32>
2277 ret <8 x i32> %shuffle32
2280 define <8 x float> @concat_v8f32_4567CDEF_bc(<8 x float> %f0, <8 x float> %f1) {
2281 ; ALL-LABEL: concat_v8f32_4567CDEF_bc:
2283 ; ALL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
2285 %a0 = bitcast <8 x float> %f0 to <4 x i64>
2286 %a1 = bitcast <8 x float> %f1 to <8 x i32>
2287 %a0hi = shufflevector <4 x i64> %a0, <4 x i64> undef, <2 x i32> <i32 2, i32 3>
2288 %a1hi = shufflevector <8 x i32> %a1, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
2289 %bc0hi = bitcast <2 x i64> %a0hi to <2 x i64>
2290 %bc1hi = bitcast <4 x i32> %a1hi to <2 x i64>
2291 %shuffle64 = shufflevector <2 x i64> %bc0hi, <2 x i64> %bc1hi, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
2292 %shuffle32 = bitcast <4 x i64> %shuffle64 to <8 x float>
2293 ret <8 x float> %shuffle32
2296 define <8 x i32> @insert_dup_mem_v8i32(i32* %ptr) {
2297 ; ALL-LABEL: insert_dup_mem_v8i32:
2299 ; ALL-NEXT: vbroadcastss (%rdi), %ymm0
2301 %tmp = load i32, i32* %ptr, align 4
2302 %tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 0
2303 %tmp2 = shufflevector <4 x i32> %tmp1, <4 x i32> undef, <8 x i32> zeroinitializer