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 ; AVX1-LABEL: shuffle_v8f32_00000000:
9 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
10 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
13 ; AVX2-LABEL: shuffle_v8f32_00000000:
15 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
17 %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>
18 ret <8 x float> %shuffle
21 define <8 x float> @shuffle_v8f32_00000010(<8 x float> %a, <8 x float> %b) {
22 ; AVX1-LABEL: shuffle_v8f32_00000010:
24 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
25 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
26 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
29 ; AVX2-LABEL: shuffle_v8f32_00000010:
31 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,0,1,0]
32 ; AVX2-NEXT: vpermps %ymm0, %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 0, i32 1, i32 0>
35 ret <8 x float> %shuffle
38 define <8 x float> @shuffle_v8f32_00000200(<8 x float> %a, <8 x float> %b) {
39 ; AVX1-LABEL: shuffle_v8f32_00000200:
41 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
42 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,2,0,0]
43 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
46 ; AVX2-LABEL: shuffle_v8f32_00000200:
48 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,2,0,0]
49 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
51 %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>
52 ret <8 x float> %shuffle
55 define <8 x float> @shuffle_v8f32_00003000(<8 x float> %a, <8 x float> %b) {
56 ; AVX1-LABEL: shuffle_v8f32_00003000:
58 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
59 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,0,0,0]
60 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
63 ; AVX2-LABEL: shuffle_v8f32_00003000:
65 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,0,0,3,0,0,0]
66 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
68 %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>
69 ret <8 x float> %shuffle
72 define <8 x float> @shuffle_v8f32_00040000(<8 x float> %a, <8 x float> %b) {
73 ; AVX1-LABEL: shuffle_v8f32_00040000:
75 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
76 ; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4]
77 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,3,4,4,4,7]
78 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
81 ; AVX2-LABEL: shuffle_v8f32_00040000:
83 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,0,4,0,0,0,0]
84 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
86 %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>
87 ret <8 x float> %shuffle
90 define <8 x float> @shuffle_v8f32_00500000(<8 x float> %a, <8 x float> %b) {
91 ; AVX1-LABEL: shuffle_v8f32_00500000:
93 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
94 ; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[u,u,1,u,4,4,4,4]
95 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
96 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3],ymm1[4,5,6,7]
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: vpermilps {{.*#+}} ymm1 = ymm1[u,2,u,u,4,4,4,4]
113 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,0,4,5,4,4]
114 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4,5,6,7]
117 ; AVX2-LABEL: shuffle_v8f32_06000000:
119 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,6,0,0,0,0,0,0]
120 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
122 %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>
123 ret <8 x float> %shuffle
126 define <8 x float> @shuffle_v8f32_70000000(<8 x float> %a, <8 x float> %b) {
127 ; AVX1-LABEL: shuffle_v8f32_70000000:
129 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
130 ; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[3,u,u,u,4,4,4,4]
131 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
132 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4,5,6,7]
135 ; AVX2-LABEL: shuffle_v8f32_70000000:
137 ; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
138 ; AVX2-NEXT: movl $7, %eax
139 ; AVX2-NEXT: vpinsrd $0, %eax, %xmm1, %xmm1
140 ; AVX2-NEXT: vpxor %ymm2, %ymm2, %ymm2
141 ; AVX2-NEXT: vinserti128 $0, %xmm1, %ymm2, %ymm1
142 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
144 %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>
145 ret <8 x float> %shuffle
148 define <8 x float> @shuffle_v8f32_01014545(<8 x float> %a, <8 x float> %b) {
149 ; ALL-LABEL: shuffle_v8f32_01014545:
151 ; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,2,2]
153 %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>
154 ret <8 x float> %shuffle
157 define <8 x float> @shuffle_v8f32_00112233(<8 x float> %a, <8 x float> %b) {
158 ; AVX1-LABEL: shuffle_v8f32_00112233:
160 ; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm0[0,0,1,1]
161 ; AVX1-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2,2,3,3]
162 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
165 ; AVX2-LABEL: shuffle_v8f32_00112233:
167 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,1,1,2,2,3,3]
168 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
170 %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>
171 ret <8 x float> %shuffle
174 define <8 x float> @shuffle_v8f32_00001111(<8 x float> %a, <8 x float> %b) {
175 ; AVX1-LABEL: shuffle_v8f32_00001111:
177 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
178 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
179 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
182 ; AVX2-LABEL: shuffle_v8f32_00001111:
184 ; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = [0,0,0,0,1,1,1,1]
185 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
187 %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>
188 ret <8 x float> %shuffle
191 define <8 x float> @shuffle_v8f32_81a3c5e7(<8 x float> %a, <8 x float> %b) {
192 ; ALL-LABEL: shuffle_v8f32_81a3c5e7:
194 ; ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
196 %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>
197 ret <8 x float> %shuffle
200 define <8 x float> @shuffle_v8f32_08080808(<8 x float> %a, <8 x float> %b) {
201 ; AVX1-LABEL: shuffle_v8f32_08080808:
203 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,0,2,0]
204 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
205 ; AVX1-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0,0]
206 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
207 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
210 ; AVX2-LABEL: shuffle_v8f32_08080808:
212 ; AVX2-NEXT: vbroadcastss %xmm1, %ymm1
213 ; AVX2-NEXT: vbroadcastsd %xmm0, %ymm0
214 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
216 %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>
217 ret <8 x float> %shuffle
220 define <8 x float> @shuffle_v8f32_08084c4c(<8 x float> %a, <8 x float> %b) {
221 ; ALL-LABEL: shuffle_v8f32_08084c4c:
223 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
224 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
226 %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>
227 ret <8 x float> %shuffle
230 define <8 x float> @shuffle_v8f32_8823cc67(<8 x float> %a, <8 x float> %b) {
231 ; ALL-LABEL: shuffle_v8f32_8823cc67:
233 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,0],ymm0[2,3],ymm1[4,4],ymm0[6,7]
235 %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>
236 ret <8 x float> %shuffle
239 define <8 x float> @shuffle_v8f32_9832dc76(<8 x float> %a, <8 x float> %b) {
240 ; ALL-LABEL: shuffle_v8f32_9832dc76:
242 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[3,2],ymm1[5,4],ymm0[7,6]
244 %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>
245 ret <8 x float> %shuffle
248 define <8 x float> @shuffle_v8f32_9810dc54(<8 x float> %a, <8 x float> %b) {
249 ; ALL-LABEL: shuffle_v8f32_9810dc54:
251 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
253 %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>
254 ret <8 x float> %shuffle
257 define <8 x float> @shuffle_v8f32_08194c5d(<8 x float> %a, <8 x float> %b) {
258 ; ALL-LABEL: shuffle_v8f32_08194c5d:
260 ; ALL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
262 %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>
263 ret <8 x float> %shuffle
266 define <8 x float> @shuffle_v8f32_2a3b6e7f(<8 x float> %a, <8 x float> %b) {
267 ; ALL-LABEL: shuffle_v8f32_2a3b6e7f:
269 ; ALL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
271 %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>
272 ret <8 x float> %shuffle
275 define <8 x float> @shuffle_v8f32_08192a3b(<8 x float> %a, <8 x float> %b) {
276 ; AVX1-LABEL: shuffle_v8f32_08192a3b:
278 ; AVX1-NEXT: vunpckhps {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
279 ; AVX1-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
280 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
283 ; AVX2-LABEL: shuffle_v8f32_08192a3b:
285 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <u,0,u,1,u,2,u,3>
286 ; AVX2-NEXT: vpermps %ymm1, %ymm2, %ymm1
287 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <0,u,1,u,2,u,3,u>
288 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
289 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
291 %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>
292 ret <8 x float> %shuffle
295 define <8 x float> @shuffle_v8f32_08991abb(<8 x float> %a, <8 x float> %b) {
296 ; AVX1-LABEL: shuffle_v8f32_08991abb:
298 ; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm0[1,0],xmm1[2,0]
299 ; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[3,3]
300 ; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
301 ; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[1,1]
302 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
305 ; AVX2-LABEL: shuffle_v8f32_08991abb:
307 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <0,u,u,u,1,u,u,u>
308 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
309 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <u,0,1,1,u,2,3,3>
310 ; AVX2-NEXT: vpermps %ymm1, %ymm2, %ymm1
311 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
313 %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>
314 ret <8 x float> %shuffle
317 define <8 x float> @shuffle_v8f32_091b2d3f(<8 x float> %a, <8 x float> %b) {
318 ; AVX1-LABEL: shuffle_v8f32_091b2d3f:
320 ; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[0,1,1,3]
321 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,1,3,3]
322 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
323 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
326 ; AVX2-LABEL: shuffle_v8f32_091b2d3f:
328 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <0,u,1,u,2,u,3,u>
329 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
330 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
332 %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>
333 ret <8 x float> %shuffle
336 define <8 x float> @shuffle_v8f32_09ab1def(<8 x float> %a, <8 x float> %b) {
337 ; AVX1-LABEL: shuffle_v8f32_09ab1def:
339 ; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[1,1,2,3]
340 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
341 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
344 ; AVX2-LABEL: shuffle_v8f32_09ab1def:
346 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <0,u,u,u,1,u,u,u>
347 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
348 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
350 %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>
351 ret <8 x float> %shuffle
354 define <8 x float> @shuffle_v8f32_00014445(<8 x float> %a, <8 x float> %b) {
355 ; ALL-LABEL: shuffle_v8f32_00014445:
357 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
359 %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>
360 ret <8 x float> %shuffle
363 define <8 x float> @shuffle_v8f32_00204464(<8 x float> %a, <8 x float> %b) {
364 ; ALL-LABEL: shuffle_v8f32_00204464:
366 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
368 %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>
369 ret <8 x float> %shuffle
372 define <8 x float> @shuffle_v8f32_03004744(<8 x float> %a, <8 x float> %b) {
373 ; ALL-LABEL: shuffle_v8f32_03004744:
375 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
377 %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>
378 ret <8 x float> %shuffle
381 define <8 x float> @shuffle_v8f32_10005444(<8 x float> %a, <8 x float> %b) {
382 ; ALL-LABEL: shuffle_v8f32_10005444:
384 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
386 %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>
387 ret <8 x float> %shuffle
390 define <8 x float> @shuffle_v8f32_22006644(<8 x float> %a, <8 x float> %b) {
391 ; ALL-LABEL: shuffle_v8f32_22006644:
393 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
395 %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>
396 ret <8 x float> %shuffle
399 define <8 x float> @shuffle_v8f32_33307774(<8 x float> %a, <8 x float> %b) {
400 ; ALL-LABEL: shuffle_v8f32_33307774:
402 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
404 %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>
405 ret <8 x float> %shuffle
408 define <8 x float> @shuffle_v8f32_32107654(<8 x float> %a, <8 x float> %b) {
409 ; ALL-LABEL: shuffle_v8f32_32107654:
411 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
413 %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>
414 ret <8 x float> %shuffle
417 define <8 x float> @shuffle_v8f32_00234467(<8 x float> %a, <8 x float> %b) {
418 ; ALL-LABEL: shuffle_v8f32_00234467:
420 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
422 %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>
423 ret <8 x float> %shuffle
426 define <8 x float> @shuffle_v8f32_00224466(<8 x float> %a, <8 x float> %b) {
427 ; ALL-LABEL: shuffle_v8f32_00224466:
429 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
431 %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>
432 ret <8 x float> %shuffle
435 define <8 x float> @shuffle_v8f32_10325476(<8 x float> %a, <8 x float> %b) {
436 ; ALL-LABEL: shuffle_v8f32_10325476:
438 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
440 %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>
441 ret <8 x float> %shuffle
444 define <8 x float> @shuffle_v8f32_11335577(<8 x float> %a, <8 x float> %b) {
445 ; ALL-LABEL: shuffle_v8f32_11335577:
447 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
449 %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>
450 ret <8 x float> %shuffle
453 define <8 x float> @shuffle_v8f32_10235467(<8 x float> %a, <8 x float> %b) {
454 ; ALL-LABEL: shuffle_v8f32_10235467:
456 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
458 %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>
459 ret <8 x float> %shuffle
462 define <8 x float> @shuffle_v8f32_10225466(<8 x float> %a, <8 x float> %b) {
463 ; ALL-LABEL: shuffle_v8f32_10225466:
465 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
467 %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>
468 ret <8 x float> %shuffle
471 define <8 x float> @shuffle_v8f32_00015444(<8 x float> %a, <8 x float> %b) {
472 ; ALL-LABEL: shuffle_v8f32_00015444:
474 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,5,4,4,4]
476 %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>
477 ret <8 x float> %shuffle
480 define <8 x float> @shuffle_v8f32_00204644(<8 x float> %a, <8 x float> %b) {
481 ; ALL-LABEL: shuffle_v8f32_00204644:
483 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,6,4,4]
485 %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>
486 ret <8 x float> %shuffle
489 define <8 x float> @shuffle_v8f32_03004474(<8 x float> %a, <8 x float> %b) {
490 ; ALL-LABEL: shuffle_v8f32_03004474:
492 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,4,7,4]
494 %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>
495 ret <8 x float> %shuffle
498 define <8 x float> @shuffle_v8f32_10004444(<8 x float> %a, <8 x float> %b) {
499 ; ALL-LABEL: shuffle_v8f32_10004444:
501 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,4,4,4,4]
503 %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>
504 ret <8 x float> %shuffle
507 define <8 x float> @shuffle_v8f32_22006446(<8 x float> %a, <8 x float> %b) {
508 ; ALL-LABEL: shuffle_v8f32_22006446:
510 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,4,4,6]
512 %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>
513 ret <8 x float> %shuffle
516 define <8 x float> @shuffle_v8f32_33307474(<8 x float> %a, <8 x float> %b) {
517 ; ALL-LABEL: shuffle_v8f32_33307474:
519 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,4,7,4]
521 %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>
522 ret <8 x float> %shuffle
525 define <8 x float> @shuffle_v8f32_32104567(<8 x float> %a, <8 x float> %b) {
526 ; ALL-LABEL: shuffle_v8f32_32104567:
528 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7]
530 %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>
531 ret <8 x float> %shuffle
534 define <8 x float> @shuffle_v8f32_00236744(<8 x float> %a, <8 x float> %b) {
535 ; ALL-LABEL: shuffle_v8f32_00236744:
537 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,6,7,4,4]
539 %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>
540 ret <8 x float> %shuffle
543 define <8 x float> @shuffle_v8f32_00226644(<8 x float> %a, <8 x float> %b) {
544 ; ALL-LABEL: shuffle_v8f32_00226644:
546 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,2,6,6,4,4]
548 %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>
549 ret <8 x float> %shuffle
552 define <8 x float> @shuffle_v8f32_10324567(<8 x float> %a, <8 x float> %b) {
553 ; ALL-LABEL: shuffle_v8f32_10324567:
555 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,6,7]
557 %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>
558 ret <8 x float> %shuffle
561 define <8 x float> @shuffle_v8f32_11334567(<8 x float> %a, <8 x float> %b) {
562 ; ALL-LABEL: shuffle_v8f32_11334567:
564 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7]
566 %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>
567 ret <8 x float> %shuffle
570 define <8 x float> @shuffle_v8f32_01235467(<8 x float> %a, <8 x float> %b) {
571 ; ALL-LABEL: shuffle_v8f32_01235467:
573 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,7]
575 %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>
576 ret <8 x float> %shuffle
579 define <8 x float> @shuffle_v8f32_01235466(<8 x float> %a, <8 x float> %b) {
580 ; ALL-LABEL: shuffle_v8f32_01235466:
582 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,6]
584 %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>
585 ret <8 x float> %shuffle
588 define <8 x float> @shuffle_v8f32_002u6u44(<8 x float> %a, <8 x float> %b) {
589 ; ALL-LABEL: shuffle_v8f32_002u6u44:
591 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,u,6,u,4,4]
593 %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>
594 ret <8 x float> %shuffle
597 define <8 x float> @shuffle_v8f32_00uu66uu(<8 x float> %a, <8 x float> %b) {
598 ; ALL-LABEL: shuffle_v8f32_00uu66uu:
600 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,u,u,6,6,u,u]
602 %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>
603 ret <8 x float> %shuffle
606 define <8 x float> @shuffle_v8f32_103245uu(<8 x float> %a, <8 x float> %b) {
607 ; ALL-LABEL: shuffle_v8f32_103245uu:
609 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,u,u]
611 %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>
612 ret <8 x float> %shuffle
615 define <8 x float> @shuffle_v8f32_1133uu67(<8 x float> %a, <8 x float> %b) {
616 ; ALL-LABEL: shuffle_v8f32_1133uu67:
618 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,u,u,6,7]
620 %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>
621 ret <8 x float> %shuffle
624 define <8 x float> @shuffle_v8f32_0uu354uu(<8 x float> %a, <8 x float> %b) {
625 ; ALL-LABEL: shuffle_v8f32_0uu354uu:
627 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,u,u,3,5,4,u,u]
629 %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>
630 ret <8 x float> %shuffle
633 define <8 x float> @shuffle_v8f32_uuu3uu66(<8 x float> %a, <8 x float> %b) {
634 ; ALL-LABEL: shuffle_v8f32_uuu3uu66:
636 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[u,u,u,3,u,u,6,6]
638 %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>
639 ret <8 x float> %shuffle
642 define <8 x float> @shuffle_v8f32_c348cda0(<8 x float> %a, <8 x float> %b) {
643 ; AVX1-LABEL: shuffle_v8f32_c348cda0:
645 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
646 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm2[0,0],ymm0[4,7],ymm2[4,4]
647 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
648 ; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
649 ; AVX1-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3]
650 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
653 ; AVX2-LABEL: shuffle_v8f32_c348cda0:
655 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <u,3,4,u,u,u,u,0>
656 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
657 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <4,u,u,0,4,5,2,u>
658 ; AVX2-NEXT: vpermps %ymm1, %ymm2, %ymm1
659 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
661 %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>
662 ret <8 x float> %shuffle
665 define <8 x float> @shuffle_v8f32_f511235a(<8 x float> %a, <8 x float> %b) {
666 ; AVX1-LABEL: shuffle_v8f32_f511235a:
668 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
669 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
670 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,1,1,4,5,5,5]
671 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
672 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
673 ; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[3,1,2,2,7,5,6,6]
674 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
677 ; AVX2-LABEL: shuffle_v8f32_f511235a:
679 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <7,u,u,u,u,u,u,2>
680 ; AVX2-NEXT: vpermps %ymm1, %ymm2, %ymm1
681 ; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <u,5,1,1,2,3,5,u>
682 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
683 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
685 %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>
686 ret <8 x float> %shuffle
689 define <8 x i32> @shuffle_v8i32_00000000(<8 x i32> %a, <8 x i32> %b) {
690 ; AVX1-LABEL: shuffle_v8i32_00000000:
692 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
693 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
696 ; AVX2-LABEL: shuffle_v8i32_00000000:
698 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
700 %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>
701 ret <8 x i32> %shuffle
704 define <8 x i32> @shuffle_v8i32_00000010(<8 x i32> %a, <8 x i32> %b) {
705 ; AVX1-LABEL: shuffle_v8i32_00000010:
707 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
708 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
709 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
712 ; AVX2-LABEL: shuffle_v8i32_00000010:
714 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,1,0]
715 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
717 %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>
718 ret <8 x i32> %shuffle
721 define <8 x i32> @shuffle_v8i32_00000200(<8 x i32> %a, <8 x i32> %b) {
722 ; AVX1-LABEL: shuffle_v8i32_00000200:
724 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
725 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,2,0,0]
726 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
729 ; AVX2-LABEL: shuffle_v8i32_00000200:
731 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,2,0,0]
732 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
734 %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>
735 ret <8 x i32> %shuffle
738 define <8 x i32> @shuffle_v8i32_00003000(<8 x i32> %a, <8 x i32> %b) {
739 ; AVX1-LABEL: shuffle_v8i32_00003000:
741 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
742 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,0,0,0]
743 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
746 ; AVX2-LABEL: shuffle_v8i32_00003000:
748 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,0,3,0,0,0]
749 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
751 %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>
752 ret <8 x i32> %shuffle
755 define <8 x i32> @shuffle_v8i32_00040000(<8 x i32> %a, <8 x i32> %b) {
756 ; AVX1-LABEL: shuffle_v8i32_00040000:
758 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
759 ; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4]
760 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,3,4,4,4,7]
761 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
764 ; AVX2-LABEL: shuffle_v8i32_00040000:
766 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,4,0,0,0,0]
767 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
769 %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>
770 ret <8 x i32> %shuffle
773 define <8 x i32> @shuffle_v8i32_00500000(<8 x i32> %a, <8 x i32> %b) {
774 ; AVX1-LABEL: shuffle_v8i32_00500000:
776 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
777 ; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[u,u,1,u,4,4,4,4]
778 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
779 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3],ymm1[4,5,6,7]
782 ; AVX2-LABEL: shuffle_v8i32_00500000:
784 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,5,0,0,0,0,0]
785 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
787 %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>
788 ret <8 x i32> %shuffle
791 define <8 x i32> @shuffle_v8i32_06000000(<8 x i32> %a, <8 x i32> %b) {
792 ; AVX1-LABEL: shuffle_v8i32_06000000:
794 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
795 ; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[u,2,u,u,4,4,4,4]
796 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,0,4,5,4,4]
797 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4,5,6,7]
800 ; AVX2-LABEL: shuffle_v8i32_06000000:
802 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,6,0,0,0,0,0,0]
803 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
805 %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>
806 ret <8 x i32> %shuffle
809 define <8 x i32> @shuffle_v8i32_70000000(<8 x i32> %a, <8 x i32> %b) {
810 ; AVX1-LABEL: shuffle_v8i32_70000000:
812 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
813 ; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[3,u,u,u,4,4,4,4]
814 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
815 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4,5,6,7]
818 ; AVX2-LABEL: shuffle_v8i32_70000000:
820 ; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
821 ; AVX2-NEXT: movl $7, %eax
822 ; AVX2-NEXT: vpinsrd $0, %eax, %xmm1, %xmm1
823 ; AVX2-NEXT: vpxor %ymm2, %ymm2, %ymm2
824 ; AVX2-NEXT: vinserti128 $0, %xmm1, %ymm2, %ymm1
825 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
827 %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>
828 ret <8 x i32> %shuffle
831 define <8 x i32> @shuffle_v8i32_01014545(<8 x i32> %a, <8 x i32> %b) {
832 ; AVX1-LABEL: shuffle_v8i32_01014545:
834 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,2,2]
837 ; AVX2-LABEL: shuffle_v8i32_01014545:
839 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
841 %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>
842 ret <8 x i32> %shuffle
845 define <8 x i32> @shuffle_v8i32_00112233(<8 x i32> %a, <8 x i32> %b) {
846 ; AVX1-LABEL: shuffle_v8i32_00112233:
848 ; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm0[0,0,1,1]
849 ; AVX1-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2,2,3,3]
850 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
853 ; AVX2-LABEL: shuffle_v8i32_00112233:
855 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,1,1,2,2,3,3]
856 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
858 %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>
859 ret <8 x i32> %shuffle
862 define <8 x i32> @shuffle_v8i32_00001111(<8 x i32> %a, <8 x i32> %b) {
863 ; AVX1-LABEL: shuffle_v8i32_00001111:
865 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
866 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
867 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
870 ; AVX2-LABEL: shuffle_v8i32_00001111:
872 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,0,1,1,1,1]
873 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
875 %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>
876 ret <8 x i32> %shuffle
879 define <8 x i32> @shuffle_v8i32_81a3c5e7(<8 x i32> %a, <8 x i32> %b) {
880 ; AVX1-LABEL: shuffle_v8i32_81a3c5e7:
882 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
885 ; AVX2-LABEL: shuffle_v8i32_81a3c5e7:
887 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
889 %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>
890 ret <8 x i32> %shuffle
893 define <8 x i32> @shuffle_v8i32_08080808(<8 x i32> %a, <8 x i32> %b) {
894 ; AVX1-LABEL: shuffle_v8i32_08080808:
896 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,0,2,0]
897 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
898 ; AVX1-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0,0]
899 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
900 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
903 ; AVX2-LABEL: shuffle_v8i32_08080808:
905 ; AVX2-NEXT: vpbroadcastd %xmm1, %ymm1
906 ; AVX2-NEXT: vpbroadcastq %xmm0, %ymm0
907 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
909 %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>
910 ret <8 x i32> %shuffle
913 define <8 x i32> @shuffle_v8i32_08084c4c(<8 x i32> %a, <8 x i32> %b) {
914 ; AVX1-LABEL: shuffle_v8i32_08084c4c:
916 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
917 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
920 ; AVX2-LABEL: shuffle_v8i32_08084c4c:
922 ; AVX2-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,0,2,0,4,4,6,4]
923 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
924 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
926 %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>
927 ret <8 x i32> %shuffle
930 define <8 x i32> @shuffle_v8i32_8823cc67(<8 x i32> %a, <8 x i32> %b) {
931 ; AVX1-LABEL: shuffle_v8i32_8823cc67:
933 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,0],ymm0[2,3],ymm1[4,4],ymm0[6,7]
936 ; AVX2-LABEL: shuffle_v8i32_8823cc67:
938 ; AVX2-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,0,2,3,4,4,6,7]
939 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
941 %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>
942 ret <8 x i32> %shuffle
945 define <8 x i32> @shuffle_v8i32_9832dc76(<8 x i32> %a, <8 x i32> %b) {
946 ; AVX1-LABEL: shuffle_v8i32_9832dc76:
948 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[3,2],ymm1[5,4],ymm0[7,6]
951 ; AVX2-LABEL: shuffle_v8i32_9832dc76:
953 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
954 ; AVX2-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,0,2,3,5,4,6,7]
955 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
957 %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>
958 ret <8 x i32> %shuffle
961 define <8 x i32> @shuffle_v8i32_9810dc54(<8 x i32> %a, <8 x i32> %b) {
962 ; AVX1-LABEL: shuffle_v8i32_9810dc54:
964 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
967 ; AVX2-LABEL: shuffle_v8i32_9810dc54:
969 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,1,0,4,5,5,4]
970 ; AVX2-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,0,2,3,5,4,6,7]
971 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
973 %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>
974 ret <8 x i32> %shuffle
977 define <8 x i32> @shuffle_v8i32_08194c5d(<8 x i32> %a, <8 x i32> %b) {
978 ; AVX1-LABEL: shuffle_v8i32_08194c5d:
980 ; AVX1-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
983 ; AVX2-LABEL: shuffle_v8i32_08194c5d:
985 ; AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
987 %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>
988 ret <8 x i32> %shuffle
991 define <8 x i32> @shuffle_v8i32_2a3b6e7f(<8 x i32> %a, <8 x i32> %b) {
992 ; AVX1-LABEL: shuffle_v8i32_2a3b6e7f:
994 ; AVX1-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
997 ; AVX2-LABEL: shuffle_v8i32_2a3b6e7f:
999 ; AVX2-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
1001 %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>
1002 ret <8 x i32> %shuffle
1005 define <8 x i32> @shuffle_v8i32_08192a3b(<8 x i32> %a, <8 x i32> %b) {
1006 ; AVX1-LABEL: shuffle_v8i32_08192a3b:
1008 ; AVX1-NEXT: vunpckhps {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
1009 ; AVX1-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
1010 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1013 ; AVX2-LABEL: shuffle_v8i32_08192a3b:
1015 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <u,0,u,1,u,2,u,3>
1016 ; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1
1017 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <0,u,1,u,2,u,3,u>
1018 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
1019 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1021 %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>
1022 ret <8 x i32> %shuffle
1025 define <8 x i32> @shuffle_v8i32_08991abb(<8 x i32> %a, <8 x i32> %b) {
1026 ; AVX1-LABEL: shuffle_v8i32_08991abb:
1028 ; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm0[1,0],xmm1[2,0]
1029 ; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[3,3]
1030 ; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
1031 ; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[1,1]
1032 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1035 ; AVX2-LABEL: shuffle_v8i32_08991abb:
1037 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <0,u,u,u,1,u,u,u>
1038 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
1039 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <u,0,1,1,u,2,3,3>
1040 ; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1
1041 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1043 %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>
1044 ret <8 x i32> %shuffle
1047 define <8 x i32> @shuffle_v8i32_091b2d3f(<8 x i32> %a, <8 x i32> %b) {
1048 ; AVX1-LABEL: shuffle_v8i32_091b2d3f:
1050 ; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[0,1,1,3]
1051 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,1,3,3]
1052 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
1053 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1056 ; AVX2-LABEL: shuffle_v8i32_091b2d3f:
1058 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <0,u,1,u,2,u,3,u>
1059 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
1060 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1062 %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>
1063 ret <8 x i32> %shuffle
1066 define <8 x i32> @shuffle_v8i32_09ab1def(<8 x i32> %a, <8 x i32> %b) {
1067 ; AVX1-LABEL: shuffle_v8i32_09ab1def:
1069 ; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[1,1,2,3]
1070 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1071 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1074 ; AVX2-LABEL: shuffle_v8i32_09ab1def:
1076 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <0,u,u,u,1,u,u,u>
1077 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
1078 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1080 %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>
1081 ret <8 x i32> %shuffle
1084 define <8 x i32> @shuffle_v8i32_00014445(<8 x i32> %a, <8 x i32> %b) {
1085 ; AVX1-LABEL: shuffle_v8i32_00014445:
1087 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
1090 ; AVX2-LABEL: shuffle_v8i32_00014445:
1092 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
1094 %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>
1095 ret <8 x i32> %shuffle
1098 define <8 x i32> @shuffle_v8i32_00204464(<8 x i32> %a, <8 x i32> %b) {
1099 ; AVX1-LABEL: shuffle_v8i32_00204464:
1101 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
1104 ; AVX2-LABEL: shuffle_v8i32_00204464:
1106 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
1108 %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>
1109 ret <8 x i32> %shuffle
1112 define <8 x i32> @shuffle_v8i32_03004744(<8 x i32> %a, <8 x i32> %b) {
1113 ; AVX1-LABEL: shuffle_v8i32_03004744:
1115 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
1118 ; AVX2-LABEL: shuffle_v8i32_03004744:
1120 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
1122 %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>
1123 ret <8 x i32> %shuffle
1126 define <8 x i32> @shuffle_v8i32_10005444(<8 x i32> %a, <8 x i32> %b) {
1127 ; AVX1-LABEL: shuffle_v8i32_10005444:
1129 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
1132 ; AVX2-LABEL: shuffle_v8i32_10005444:
1134 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
1136 %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>
1137 ret <8 x i32> %shuffle
1140 define <8 x i32> @shuffle_v8i32_22006644(<8 x i32> %a, <8 x i32> %b) {
1141 ; AVX1-LABEL: shuffle_v8i32_22006644:
1143 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
1146 ; AVX2-LABEL: shuffle_v8i32_22006644:
1148 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
1150 %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>
1151 ret <8 x i32> %shuffle
1154 define <8 x i32> @shuffle_v8i32_33307774(<8 x i32> %a, <8 x i32> %b) {
1155 ; AVX1-LABEL: shuffle_v8i32_33307774:
1157 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
1160 ; AVX2-LABEL: shuffle_v8i32_33307774:
1162 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
1164 %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>
1165 ret <8 x i32> %shuffle
1168 define <8 x i32> @shuffle_v8i32_32107654(<8 x i32> %a, <8 x i32> %b) {
1169 ; AVX1-LABEL: shuffle_v8i32_32107654:
1171 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1174 ; AVX2-LABEL: shuffle_v8i32_32107654:
1176 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1178 %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>
1179 ret <8 x i32> %shuffle
1182 define <8 x i32> @shuffle_v8i32_00234467(<8 x i32> %a, <8 x i32> %b) {
1183 ; AVX1-LABEL: shuffle_v8i32_00234467:
1185 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
1188 ; AVX2-LABEL: shuffle_v8i32_00234467:
1190 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
1192 %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>
1193 ret <8 x i32> %shuffle
1196 define <8 x i32> @shuffle_v8i32_00224466(<8 x i32> %a, <8 x i32> %b) {
1197 ; AVX1-LABEL: shuffle_v8i32_00224466:
1199 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
1202 ; AVX2-LABEL: shuffle_v8i32_00224466:
1204 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
1206 %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>
1207 ret <8 x i32> %shuffle
1210 define <8 x i32> @shuffle_v8i32_10325476(<8 x i32> %a, <8 x i32> %b) {
1211 ; AVX1-LABEL: shuffle_v8i32_10325476:
1213 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
1216 ; AVX2-LABEL: shuffle_v8i32_10325476:
1218 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
1220 %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>
1221 ret <8 x i32> %shuffle
1224 define <8 x i32> @shuffle_v8i32_11335577(<8 x i32> %a, <8 x i32> %b) {
1225 ; AVX1-LABEL: shuffle_v8i32_11335577:
1227 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
1230 ; AVX2-LABEL: shuffle_v8i32_11335577:
1232 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
1234 %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>
1235 ret <8 x i32> %shuffle
1238 define <8 x i32> @shuffle_v8i32_10235467(<8 x i32> %a, <8 x i32> %b) {
1239 ; AVX1-LABEL: shuffle_v8i32_10235467:
1241 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
1244 ; AVX2-LABEL: shuffle_v8i32_10235467:
1246 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
1248 %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>
1249 ret <8 x i32> %shuffle
1252 define <8 x i32> @shuffle_v8i32_10225466(<8 x i32> %a, <8 x i32> %b) {
1253 ; AVX1-LABEL: shuffle_v8i32_10225466:
1255 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
1258 ; AVX2-LABEL: shuffle_v8i32_10225466:
1260 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
1262 %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>
1263 ret <8 x i32> %shuffle
1266 define <8 x i32> @shuffle_v8i32_00015444(<8 x i32> %a, <8 x i32> %b) {
1267 ; AVX1-LABEL: shuffle_v8i32_00015444:
1269 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,5,4,4,4]
1272 ; AVX2-LABEL: shuffle_v8i32_00015444:
1274 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,0,1,5,4,4,4]
1275 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1277 %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>
1278 ret <8 x i32> %shuffle
1281 define <8 x i32> @shuffle_v8i32_00204644(<8 x i32> %a, <8 x i32> %b) {
1282 ; AVX1-LABEL: shuffle_v8i32_00204644:
1284 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,6,4,4]
1287 ; AVX2-LABEL: shuffle_v8i32_00204644:
1289 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,2,0,4,6,4,4]
1290 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1292 %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>
1293 ret <8 x i32> %shuffle
1296 define <8 x i32> @shuffle_v8i32_03004474(<8 x i32> %a, <8 x i32> %b) {
1297 ; AVX1-LABEL: shuffle_v8i32_03004474:
1299 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,4,7,4]
1302 ; AVX2-LABEL: shuffle_v8i32_03004474:
1304 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,3,0,0,4,4,7,4]
1305 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1307 %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>
1308 ret <8 x i32> %shuffle
1311 define <8 x i32> @shuffle_v8i32_10004444(<8 x i32> %a, <8 x i32> %b) {
1312 ; AVX1-LABEL: shuffle_v8i32_10004444:
1314 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,4,4,4,4]
1317 ; AVX2-LABEL: shuffle_v8i32_10004444:
1319 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [1,0,0,0,4,4,4,4]
1320 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1322 %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>
1323 ret <8 x i32> %shuffle
1326 define <8 x i32> @shuffle_v8i32_22006446(<8 x i32> %a, <8 x i32> %b) {
1327 ; AVX1-LABEL: shuffle_v8i32_22006446:
1329 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,4,4,6]
1332 ; AVX2-LABEL: shuffle_v8i32_22006446:
1334 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [2,2,0,0,6,4,4,6]
1335 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1337 %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>
1338 ret <8 x i32> %shuffle
1341 define <8 x i32> @shuffle_v8i32_33307474(<8 x i32> %a, <8 x i32> %b) {
1342 ; AVX1-LABEL: shuffle_v8i32_33307474:
1344 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,4,7,4]
1347 ; AVX2-LABEL: shuffle_v8i32_33307474:
1349 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [3,3,3,0,7,4,7,4]
1350 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1352 %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>
1353 ret <8 x i32> %shuffle
1356 define <8 x i32> @shuffle_v8i32_32104567(<8 x i32> %a, <8 x i32> %b) {
1357 ; AVX1-LABEL: shuffle_v8i32_32104567:
1359 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7]
1362 ; AVX2-LABEL: shuffle_v8i32_32104567:
1364 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [3,2,1,0,4,5,6,7]
1365 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1367 %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>
1368 ret <8 x i32> %shuffle
1371 define <8 x i32> @shuffle_v8i32_00236744(<8 x i32> %a, <8 x i32> %b) {
1372 ; AVX1-LABEL: shuffle_v8i32_00236744:
1374 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,6,7,4,4]
1377 ; AVX2-LABEL: shuffle_v8i32_00236744:
1379 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,2,3,6,7,4,4]
1380 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1382 %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>
1383 ret <8 x i32> %shuffle
1386 define <8 x i32> @shuffle_v8i32_00226644(<8 x i32> %a, <8 x i32> %b) {
1387 ; AVX1-LABEL: shuffle_v8i32_00226644:
1389 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,2,6,6,4,4]
1392 ; AVX2-LABEL: shuffle_v8i32_00226644:
1394 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,0,2,2,6,6,4,4]
1395 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1397 %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>
1398 ret <8 x i32> %shuffle
1401 define <8 x i32> @shuffle_v8i32_10324567(<8 x i32> %a, <8 x i32> %b) {
1402 ; AVX1-LABEL: shuffle_v8i32_10324567:
1404 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,6,7]
1407 ; AVX2-LABEL: shuffle_v8i32_10324567:
1409 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [1,0,3,2,4,5,6,7]
1410 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1412 %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>
1413 ret <8 x i32> %shuffle
1416 define <8 x i32> @shuffle_v8i32_11334567(<8 x i32> %a, <8 x i32> %b) {
1417 ; AVX1-LABEL: shuffle_v8i32_11334567:
1419 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7]
1422 ; AVX2-LABEL: shuffle_v8i32_11334567:
1424 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [1,1,3,3,4,5,6,7]
1425 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1427 %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>
1428 ret <8 x i32> %shuffle
1431 define <8 x i32> @shuffle_v8i32_01235467(<8 x i32> %a, <8 x i32> %b) {
1432 ; AVX1-LABEL: shuffle_v8i32_01235467:
1434 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,7]
1437 ; AVX2-LABEL: shuffle_v8i32_01235467:
1439 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,1,2,3,5,4,6,7]
1440 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1442 %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>
1443 ret <8 x i32> %shuffle
1446 define <8 x i32> @shuffle_v8i32_01235466(<8 x i32> %a, <8 x i32> %b) {
1447 ; AVX1-LABEL: shuffle_v8i32_01235466:
1449 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,6]
1452 ; AVX2-LABEL: shuffle_v8i32_01235466:
1454 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [0,1,2,3,5,4,6,6]
1455 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1457 %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>
1458 ret <8 x i32> %shuffle
1461 define <8 x i32> @shuffle_v8i32_002u6u44(<8 x i32> %a, <8 x i32> %b) {
1462 ; AVX1-LABEL: shuffle_v8i32_002u6u44:
1464 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,u,6,u,4,4]
1467 ; AVX2-LABEL: shuffle_v8i32_002u6u44:
1469 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <0,0,2,u,6,u,4,4>
1470 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1472 %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>
1473 ret <8 x i32> %shuffle
1476 define <8 x i32> @shuffle_v8i32_00uu66uu(<8 x i32> %a, <8 x i32> %b) {
1477 ; AVX1-LABEL: shuffle_v8i32_00uu66uu:
1479 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,u,u,6,6,u,u]
1482 ; AVX2-LABEL: shuffle_v8i32_00uu66uu:
1484 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <0,0,u,u,6,6,u,u>
1485 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1487 %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>
1488 ret <8 x i32> %shuffle
1491 define <8 x i32> @shuffle_v8i32_103245uu(<8 x i32> %a, <8 x i32> %b) {
1492 ; AVX1-LABEL: shuffle_v8i32_103245uu:
1494 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,u,u]
1497 ; AVX2-LABEL: shuffle_v8i32_103245uu:
1499 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <1,0,3,2,4,5,u,u>
1500 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1502 %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>
1503 ret <8 x i32> %shuffle
1506 define <8 x i32> @shuffle_v8i32_1133uu67(<8 x i32> %a, <8 x i32> %b) {
1507 ; AVX1-LABEL: shuffle_v8i32_1133uu67:
1509 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,u,u,6,7]
1512 ; AVX2-LABEL: shuffle_v8i32_1133uu67:
1514 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <1,1,3,3,u,u,6,7>
1515 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1517 %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>
1518 ret <8 x i32> %shuffle
1521 define <8 x i32> @shuffle_v8i32_0uu354uu(<8 x i32> %a, <8 x i32> %b) {
1522 ; AVX1-LABEL: shuffle_v8i32_0uu354uu:
1524 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,u,u,3,5,4,u,u]
1527 ; AVX2-LABEL: shuffle_v8i32_0uu354uu:
1529 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <0,u,u,3,5,4,u,u>
1530 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1532 %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>
1533 ret <8 x i32> %shuffle
1536 define <8 x i32> @shuffle_v8i32_uuu3uu66(<8 x i32> %a, <8 x i32> %b) {
1537 ; AVX1-LABEL: shuffle_v8i32_uuu3uu66:
1539 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[u,u,u,3,u,u,6,6]
1542 ; AVX2-LABEL: shuffle_v8i32_uuu3uu66:
1544 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,3,u,u,6,6>
1545 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1547 %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>
1548 ret <8 x i32> %shuffle
1551 define <8 x i32> @shuffle_v8i32_6caa87e5(<8 x i32> %a, <8 x i32> %b) {
1552 ; AVX1-LABEL: shuffle_v8i32_6caa87e5:
1554 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
1555 ; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[2,2],ymm2[4,4],ymm1[6,6]
1556 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
1557 ; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
1558 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1559 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
1562 ; AVX2-LABEL: shuffle_v8i32_6caa87e5:
1564 ; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <u,4,2,2,0,u,6,u>
1565 ; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1
1566 ; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,1,3,2]
1567 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
1569 %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>
1570 ret <8 x i32> %shuffle
1573 define <8 x float> @splat_mem_v8f32_2(float* %p) {
1574 ; ALL-LABEL: splat_mem_v8f32_2:
1576 ; ALL-NEXT: vbroadcastss (%rdi), %ymm0
1579 %2 = insertelement <4 x float> undef, float %1, i32 0
1580 %3 = shufflevector <4 x float> %2, <4 x float> undef, <8 x i32> zeroinitializer
1584 define <8 x float> @splat_v8f32(<4 x float> %r) {
1585 ; AVX1-LABEL: splat_v8f32:
1587 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
1588 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1591 ; AVX2-LABEL: splat_v8f32:
1593 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
1595 %1 = shufflevector <4 x float> %r, <4 x float> undef, <8 x i32> zeroinitializer