1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE2
3 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE42
4 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1
5 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2 --check-prefix=AVX2-SLOW
6 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefix=AVX --check-prefix=AVX2 --check-prefix=AVX2-FAST
7 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512F
8 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512VL
9 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512BW
10 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512BWVL
12 define void @shuffle_v16i8_to_v8i8_1(<16 x i8>* %L, <8 x i8>* %S) nounwind {
13 ; SSE2-LABEL: shuffle_v16i8_to_v8i8_1:
15 ; SSE2-NEXT: movdqa (%rdi), %xmm0
16 ; SSE2-NEXT: pxor %xmm1, %xmm1
17 ; SSE2-NEXT: movdqa %xmm0, %xmm2
18 ; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
19 ; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
20 ; SSE2-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
21 ; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
22 ; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,2,4,5,6,7]
23 ; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
24 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
25 ; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
26 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
27 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
28 ; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
29 ; SSE2-NEXT: packuswb %xmm0, %xmm0
30 ; SSE2-NEXT: movq %xmm0, (%rsi)
33 ; SSE42-LABEL: shuffle_v16i8_to_v8i8_1:
35 ; SSE42-NEXT: movdqa (%rdi), %xmm0
36 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u]
37 ; SSE42-NEXT: movq %xmm0, (%rsi)
40 ; AVX-LABEL: shuffle_v16i8_to_v8i8_1:
42 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
43 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u]
44 ; AVX-NEXT: vmovq %xmm0, (%rsi)
47 ; AVX512-LABEL: shuffle_v16i8_to_v8i8_1:
49 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
50 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u]
51 ; AVX512-NEXT: vmovq %xmm0, (%rsi)
53 %vec = load <16 x i8>, <16 x i8>* %L
54 %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
55 store <8 x i8> %strided.vec, <8 x i8>* %S
59 define void @shuffle_v8i16_to_v4i16_1(<8 x i16>* %L, <4 x i16>* %S) nounwind {
60 ; SSE2-LABEL: shuffle_v8i16_to_v4i16_1:
62 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = mem[3,1,2,3,4,5,6,7]
63 ; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
64 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
65 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
66 ; SSE2-NEXT: movq %xmm0, (%rsi)
69 ; SSE42-LABEL: shuffle_v8i16_to_v4i16_1:
71 ; SSE42-NEXT: movdqa (%rdi), %xmm0
72 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15]
73 ; SSE42-NEXT: movq %xmm0, (%rsi)
76 ; AVX-LABEL: shuffle_v8i16_to_v4i16_1:
78 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
79 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15]
80 ; AVX-NEXT: vmovq %xmm0, (%rsi)
83 ; AVX512-LABEL: shuffle_v8i16_to_v4i16_1:
85 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
86 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15]
87 ; AVX512-NEXT: vmovq %xmm0, (%rsi)
89 %vec = load <8 x i16>, <8 x i16>* %L
90 %strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
91 store <4 x i16> %strided.vec, <4 x i16>* %S
95 define void @shuffle_v4i32_to_v2i32_1(<4 x i32>* %L, <2 x i32>* %S) nounwind {
96 ; SSE-LABEL: shuffle_v4i32_to_v2i32_1:
98 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = mem[1,3,2,3]
99 ; SSE-NEXT: movq %xmm0, (%rsi)
102 ; AVX-LABEL: shuffle_v4i32_to_v2i32_1:
104 ; AVX-NEXT: vpermilps {{.*#+}} xmm0 = mem[1,3,2,3]
105 ; AVX-NEXT: vmovlps %xmm0, (%rsi)
108 ; AVX512-LABEL: shuffle_v4i32_to_v2i32_1:
110 ; AVX512-NEXT: vpermilps {{.*#+}} xmm0 = mem[1,3,2,3]
111 ; AVX512-NEXT: vmovlps %xmm0, (%rsi)
113 %vec = load <4 x i32>, <4 x i32>* %L
114 %strided.vec = shufflevector <4 x i32> %vec, <4 x i32> undef, <2 x i32> <i32 1, i32 3>
115 store <2 x i32> %strided.vec, <2 x i32>* %S
119 define void @shuffle_v16i8_to_v4i8_1(<16 x i8>* %L, <4 x i8>* %S) nounwind {
120 ; SSE2-LABEL: shuffle_v16i8_to_v4i8_1:
122 ; SSE2-NEXT: movdqa (%rdi), %xmm0
123 ; SSE2-NEXT: pxor %xmm1, %xmm1
124 ; SSE2-NEXT: movdqa %xmm0, %xmm2
125 ; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
126 ; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
127 ; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
128 ; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
129 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
130 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
131 ; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
132 ; SSE2-NEXT: packuswb %xmm0, %xmm0
133 ; SSE2-NEXT: movd %xmm0, (%rsi)
136 ; SSE42-LABEL: shuffle_v16i8_to_v4i8_1:
138 ; SSE42-NEXT: movdqa (%rdi), %xmm0
139 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u]
140 ; SSE42-NEXT: movd %xmm0, (%rsi)
143 ; AVX-LABEL: shuffle_v16i8_to_v4i8_1:
145 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
146 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u]
147 ; AVX-NEXT: vmovd %xmm0, (%rsi)
150 ; AVX512-LABEL: shuffle_v16i8_to_v4i8_1:
152 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
153 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u]
154 ; AVX512-NEXT: vmovd %xmm0, (%rsi)
156 %vec = load <16 x i8>, <16 x i8>* %L
157 %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
158 store <4 x i8> %strided.vec, <4 x i8>* %S
162 define void @shuffle_v16i8_to_v4i8_2(<16 x i8>* %L, <4 x i8>* %S) nounwind {
163 ; SSE2-LABEL: shuffle_v16i8_to_v4i8_2:
165 ; SSE2-NEXT: movdqa (%rdi), %xmm0
166 ; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
167 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
168 ; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
169 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
170 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
171 ; SSE2-NEXT: packuswb %xmm0, %xmm0
172 ; SSE2-NEXT: movd %xmm0, (%rsi)
175 ; SSE42-LABEL: shuffle_v16i8_to_v4i8_2:
177 ; SSE42-NEXT: movdqa (%rdi), %xmm0
178 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u]
179 ; SSE42-NEXT: movd %xmm0, (%rsi)
182 ; AVX-LABEL: shuffle_v16i8_to_v4i8_2:
184 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
185 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u]
186 ; AVX-NEXT: vmovd %xmm0, (%rsi)
189 ; AVX512-LABEL: shuffle_v16i8_to_v4i8_2:
191 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
192 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u]
193 ; AVX512-NEXT: vmovd %xmm0, (%rsi)
195 %vec = load <16 x i8>, <16 x i8>* %L
196 %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
197 store <4 x i8> %strided.vec, <4 x i8>* %S
201 define void @shuffle_v16i8_to_v4i8_3(<16 x i8>* %L, <4 x i8>* %S) nounwind {
202 ; SSE2-LABEL: shuffle_v16i8_to_v4i8_3:
204 ; SSE2-NEXT: movdqa (%rdi), %xmm0
205 ; SSE2-NEXT: pxor %xmm1, %xmm1
206 ; SSE2-NEXT: movdqa %xmm0, %xmm2
207 ; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
208 ; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
209 ; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
210 ; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
211 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
212 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
213 ; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
214 ; SSE2-NEXT: packuswb %xmm0, %xmm0
215 ; SSE2-NEXT: movd %xmm0, (%rsi)
218 ; SSE42-LABEL: shuffle_v16i8_to_v4i8_3:
220 ; SSE42-NEXT: movdqa (%rdi), %xmm0
221 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u]
222 ; SSE42-NEXT: movd %xmm0, (%rsi)
225 ; AVX-LABEL: shuffle_v16i8_to_v4i8_3:
227 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
228 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u]
229 ; AVX-NEXT: vmovd %xmm0, (%rsi)
232 ; AVX512-LABEL: shuffle_v16i8_to_v4i8_3:
234 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
235 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u]
236 ; AVX512-NEXT: vmovd %xmm0, (%rsi)
238 %vec = load <16 x i8>, <16 x i8>* %L
239 %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
240 store <4 x i8> %strided.vec, <4 x i8>* %S
244 define void @shuffle_v8i16_to_v2i16_1(<8 x i16>* %L, <2 x i16>* %S) nounwind {
245 ; SSE-LABEL: shuffle_v8i16_to_v2i16_1:
247 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = mem[0,2,2,3]
248 ; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
249 ; SSE-NEXT: movd %xmm0, (%rsi)
252 ; AVX1-LABEL: shuffle_v8i16_to_v2i16_1:
254 ; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = mem[0,2,2,3]
255 ; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
256 ; AVX1-NEXT: vmovd %xmm0, (%rsi)
259 ; AVX2-SLOW-LABEL: shuffle_v8i16_to_v2i16_1:
260 ; AVX2-SLOW: # %bb.0:
261 ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[0,2,2,3]
262 ; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
263 ; AVX2-SLOW-NEXT: vmovd %xmm0, (%rsi)
264 ; AVX2-SLOW-NEXT: retq
266 ; AVX2-FAST-LABEL: shuffle_v8i16_to_v2i16_1:
267 ; AVX2-FAST: # %bb.0:
268 ; AVX2-FAST-NEXT: vmovdqa (%rdi), %xmm0
269 ; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
270 ; AVX2-FAST-NEXT: vmovd %xmm0, (%rsi)
271 ; AVX2-FAST-NEXT: retq
273 ; AVX512F-LABEL: shuffle_v8i16_to_v2i16_1:
275 ; AVX512F-NEXT: vpshufd {{.*#+}} xmm0 = mem[0,2,2,3]
276 ; AVX512F-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
277 ; AVX512F-NEXT: vmovd %xmm0, (%rsi)
280 ; AVX512VL-LABEL: shuffle_v8i16_to_v2i16_1:
282 ; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
283 ; AVX512VL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
284 ; AVX512VL-NEXT: vmovd %xmm0, (%rsi)
285 ; AVX512VL-NEXT: retq
287 ; AVX512BW-LABEL: shuffle_v8i16_to_v2i16_1:
289 ; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
290 ; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
291 ; AVX512BW-NEXT: vmovd %xmm0, (%rsi)
292 ; AVX512BW-NEXT: retq
294 ; AVX512BWVL-LABEL: shuffle_v8i16_to_v2i16_1:
295 ; AVX512BWVL: # %bb.0:
296 ; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
297 ; AVX512BWVL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
298 ; AVX512BWVL-NEXT: vmovd %xmm0, (%rsi)
299 ; AVX512BWVL-NEXT: retq
300 %vec = load <8 x i16>, <8 x i16>* %L
301 %strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <2 x i32> <i32 1, i32 5>
302 store <2 x i16> %strided.vec, <2 x i16>* %S
306 define void @shuffle_v8i16_to_v2i16_2(<8 x i16>* %L, <2 x i16>* %S) nounwind {
307 ; SSE-LABEL: shuffle_v8i16_to_v2i16_2:
309 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = mem[3,1,2,3]
310 ; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
311 ; SSE-NEXT: movd %xmm0, (%rsi)
314 ; AVX1-LABEL: shuffle_v8i16_to_v2i16_2:
316 ; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = mem[3,1,2,3]
317 ; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
318 ; AVX1-NEXT: vmovd %xmm0, (%rsi)
321 ; AVX2-SLOW-LABEL: shuffle_v8i16_to_v2i16_2:
322 ; AVX2-SLOW: # %bb.0:
323 ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[3,1,2,3]
324 ; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
325 ; AVX2-SLOW-NEXT: vmovd %xmm0, (%rsi)
326 ; AVX2-SLOW-NEXT: retq
328 ; AVX2-FAST-LABEL: shuffle_v8i16_to_v2i16_2:
329 ; AVX2-FAST: # %bb.0:
330 ; AVX2-FAST-NEXT: vmovdqa (%rdi), %xmm0
331 ; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,12,13,4,5,6,7,8,9,10,11,12,13,14,15]
332 ; AVX2-FAST-NEXT: vmovd %xmm0, (%rsi)
333 ; AVX2-FAST-NEXT: retq
335 ; AVX512F-LABEL: shuffle_v8i16_to_v2i16_2:
337 ; AVX512F-NEXT: vpshufd {{.*#+}} xmm0 = mem[3,1,2,3]
338 ; AVX512F-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
339 ; AVX512F-NEXT: vmovd %xmm0, (%rsi)
342 ; AVX512VL-LABEL: shuffle_v8i16_to_v2i16_2:
344 ; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
345 ; AVX512VL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,12,13,4,5,6,7,8,9,10,11,12,13,14,15]
346 ; AVX512VL-NEXT: vmovd %xmm0, (%rsi)
347 ; AVX512VL-NEXT: retq
349 ; AVX512BW-LABEL: shuffle_v8i16_to_v2i16_2:
351 ; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
352 ; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,12,13,4,5,6,7,8,9,10,11,12,13,14,15]
353 ; AVX512BW-NEXT: vmovd %xmm0, (%rsi)
354 ; AVX512BW-NEXT: retq
356 ; AVX512BWVL-LABEL: shuffle_v8i16_to_v2i16_2:
357 ; AVX512BWVL: # %bb.0:
358 ; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
359 ; AVX512BWVL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,12,13,4,5,6,7,8,9,10,11,12,13,14,15]
360 ; AVX512BWVL-NEXT: vmovd %xmm0, (%rsi)
361 ; AVX512BWVL-NEXT: retq
362 %vec = load <8 x i16>, <8 x i16>* %L
363 %strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <2 x i32> <i32 2, i32 6>
364 store <2 x i16> %strided.vec, <2 x i16>* %S
368 define void @shuffle_v8i16_to_v2i16_3(<8 x i16>* %L, <2 x i16>* %S) nounwind {
369 ; SSE-LABEL: shuffle_v8i16_to_v2i16_3:
371 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = mem[3,1,2,3]
372 ; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
373 ; SSE-NEXT: movd %xmm0, (%rsi)
376 ; AVX1-LABEL: shuffle_v8i16_to_v2i16_3:
378 ; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = mem[3,1,2,3]
379 ; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
380 ; AVX1-NEXT: vmovd %xmm0, (%rsi)
383 ; AVX2-SLOW-LABEL: shuffle_v8i16_to_v2i16_3:
384 ; AVX2-SLOW: # %bb.0:
385 ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[3,1,2,3]
386 ; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
387 ; AVX2-SLOW-NEXT: vmovd %xmm0, (%rsi)
388 ; AVX2-SLOW-NEXT: retq
390 ; AVX2-FAST-LABEL: shuffle_v8i16_to_v2i16_3:
391 ; AVX2-FAST: # %bb.0:
392 ; AVX2-FAST-NEXT: vmovdqa (%rdi), %xmm0
393 ; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[6,7,14,15,4,5,6,7,8,9,10,11,12,13,14,15]
394 ; AVX2-FAST-NEXT: vmovd %xmm0, (%rsi)
395 ; AVX2-FAST-NEXT: retq
397 ; AVX512F-LABEL: shuffle_v8i16_to_v2i16_3:
399 ; AVX512F-NEXT: vpshufd {{.*#+}} xmm0 = mem[3,1,2,3]
400 ; AVX512F-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
401 ; AVX512F-NEXT: vmovd %xmm0, (%rsi)
404 ; AVX512VL-LABEL: shuffle_v8i16_to_v2i16_3:
406 ; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
407 ; AVX512VL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[6,7,14,15,4,5,6,7,8,9,10,11,12,13,14,15]
408 ; AVX512VL-NEXT: vmovd %xmm0, (%rsi)
409 ; AVX512VL-NEXT: retq
411 ; AVX512BW-LABEL: shuffle_v8i16_to_v2i16_3:
413 ; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
414 ; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[6,7,14,15,4,5,6,7,8,9,10,11,12,13,14,15]
415 ; AVX512BW-NEXT: vmovd %xmm0, (%rsi)
416 ; AVX512BW-NEXT: retq
418 ; AVX512BWVL-LABEL: shuffle_v8i16_to_v2i16_3:
419 ; AVX512BWVL: # %bb.0:
420 ; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
421 ; AVX512BWVL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[6,7,14,15,4,5,6,7,8,9,10,11,12,13,14,15]
422 ; AVX512BWVL-NEXT: vmovd %xmm0, (%rsi)
423 ; AVX512BWVL-NEXT: retq
424 %vec = load <8 x i16>, <8 x i16>* %L
425 %strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <2 x i32> <i32 3, i32 7>
426 store <2 x i16> %strided.vec, <2 x i16>* %S
430 define void @shuffle_v16i8_to_v2i8_1(<16 x i8>* %L, <2 x i8>* %S) nounwind {
431 ; SSE2-LABEL: shuffle_v16i8_to_v2i8_1:
433 ; SSE2-NEXT: movdqa (%rdi), %xmm0
434 ; SSE2-NEXT: pxor %xmm1, %xmm1
435 ; SSE2-NEXT: movdqa %xmm0, %xmm2
436 ; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
437 ; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
438 ; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
439 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
440 ; SSE2-NEXT: packuswb %xmm0, %xmm0
441 ; SSE2-NEXT: movd %xmm0, %eax
442 ; SSE2-NEXT: movw %ax, (%rsi)
445 ; SSE42-LABEL: shuffle_v16i8_to_v2i8_1:
447 ; SSE42-NEXT: movdqa (%rdi), %xmm0
448 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
449 ; SSE42-NEXT: pextrw $0, %xmm0, (%rsi)
452 ; AVX-LABEL: shuffle_v16i8_to_v2i8_1:
454 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
455 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
456 ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi)
459 ; AVX512-LABEL: shuffle_v16i8_to_v2i8_1:
461 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
462 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
463 ; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi)
465 %vec = load <16 x i8>, <16 x i8>* %L
466 %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 1, i32 9>
467 store <2 x i8> %strided.vec, <2 x i8>* %S
471 define void @shuffle_v16i8_to_v2i8_2(<16 x i8>* %L, <2 x i8>* %S) nounwind {
472 ; SSE2-LABEL: shuffle_v16i8_to_v2i8_2:
474 ; SSE2-NEXT: movdqa (%rdi), %xmm0
475 ; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
476 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
477 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
478 ; SSE2-NEXT: packuswb %xmm0, %xmm0
479 ; SSE2-NEXT: movd %xmm0, %eax
480 ; SSE2-NEXT: movw %ax, (%rsi)
483 ; SSE42-LABEL: shuffle_v16i8_to_v2i8_2:
485 ; SSE42-NEXT: movdqa (%rdi), %xmm0
486 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
487 ; SSE42-NEXT: pextrw $0, %xmm0, (%rsi)
490 ; AVX-LABEL: shuffle_v16i8_to_v2i8_2:
492 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
493 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
494 ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi)
497 ; AVX512-LABEL: shuffle_v16i8_to_v2i8_2:
499 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
500 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
501 ; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi)
503 %vec = load <16 x i8>, <16 x i8>* %L
504 %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 2, i32 10>
505 store <2 x i8> %strided.vec, <2 x i8>* %S
509 define void @shuffle_v16i8_to_v2i8_3(<16 x i8>* %L, <2 x i8>* %S) nounwind {
510 ; SSE2-LABEL: shuffle_v16i8_to_v2i8_3:
512 ; SSE2-NEXT: movdqa (%rdi), %xmm0
513 ; SSE2-NEXT: pxor %xmm1, %xmm1
514 ; SSE2-NEXT: movdqa %xmm0, %xmm2
515 ; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
516 ; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
517 ; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
518 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
519 ; SSE2-NEXT: packuswb %xmm0, %xmm0
520 ; SSE2-NEXT: movd %xmm0, %eax
521 ; SSE2-NEXT: movw %ax, (%rsi)
524 ; SSE42-LABEL: shuffle_v16i8_to_v2i8_3:
526 ; SSE42-NEXT: movdqa (%rdi), %xmm0
527 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
528 ; SSE42-NEXT: pextrw $0, %xmm0, (%rsi)
531 ; AVX-LABEL: shuffle_v16i8_to_v2i8_3:
533 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
534 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
535 ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi)
538 ; AVX512-LABEL: shuffle_v16i8_to_v2i8_3:
540 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
541 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
542 ; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi)
544 %vec = load <16 x i8>, <16 x i8>* %L
545 %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 3, i32 11>
546 store <2 x i8> %strided.vec, <2 x i8>* %S
550 define void @shuffle_v16i8_to_v2i8_4(<16 x i8>* %L, <2 x i8>* %S) nounwind {
551 ; SSE2-LABEL: shuffle_v16i8_to_v2i8_4:
553 ; SSE2-NEXT: movdqa (%rdi), %xmm0
554 ; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
555 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
556 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
557 ; SSE2-NEXT: packuswb %xmm0, %xmm0
558 ; SSE2-NEXT: movd %xmm0, %eax
559 ; SSE2-NEXT: movw %ax, (%rsi)
562 ; SSE42-LABEL: shuffle_v16i8_to_v2i8_4:
564 ; SSE42-NEXT: movdqa (%rdi), %xmm0
565 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
566 ; SSE42-NEXT: pextrw $0, %xmm0, (%rsi)
569 ; AVX-LABEL: shuffle_v16i8_to_v2i8_4:
571 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
572 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
573 ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi)
576 ; AVX512-LABEL: shuffle_v16i8_to_v2i8_4:
578 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
579 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
580 ; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi)
582 %vec = load <16 x i8>, <16 x i8>* %L
583 %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 4, i32 12>
584 store <2 x i8> %strided.vec, <2 x i8>* %S
588 define void @shuffle_v16i8_to_v2i8_5(<16 x i8>* %L, <2 x i8>* %S) nounwind {
589 ; SSE2-LABEL: shuffle_v16i8_to_v2i8_5:
591 ; SSE2-NEXT: movdqa (%rdi), %xmm0
592 ; SSE2-NEXT: pxor %xmm1, %xmm1
593 ; SSE2-NEXT: movdqa %xmm0, %xmm2
594 ; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
595 ; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
596 ; SSE2-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
597 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
598 ; SSE2-NEXT: packuswb %xmm0, %xmm0
599 ; SSE2-NEXT: movd %xmm0, %eax
600 ; SSE2-NEXT: movw %ax, (%rsi)
603 ; SSE42-LABEL: shuffle_v16i8_to_v2i8_5:
605 ; SSE42-NEXT: movdqa (%rdi), %xmm0
606 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
607 ; SSE42-NEXT: pextrw $0, %xmm0, (%rsi)
610 ; AVX-LABEL: shuffle_v16i8_to_v2i8_5:
612 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
613 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
614 ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi)
617 ; AVX512-LABEL: shuffle_v16i8_to_v2i8_5:
619 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
620 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
621 ; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi)
623 %vec = load <16 x i8>, <16 x i8>* %L
624 %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 5, i32 13>
625 store <2 x i8> %strided.vec, <2 x i8>* %S
629 define void @shuffle_v16i8_to_v2i8_6(<16 x i8>* %L, <2 x i8>* %S) nounwind {
630 ; SSE2-LABEL: shuffle_v16i8_to_v2i8_6:
632 ; SSE2-NEXT: movdqa (%rdi), %xmm0
633 ; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
634 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
635 ; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
636 ; SSE2-NEXT: packuswb %xmm0, %xmm0
637 ; SSE2-NEXT: movd %xmm0, %eax
638 ; SSE2-NEXT: movw %ax, (%rsi)
641 ; SSE42-LABEL: shuffle_v16i8_to_v2i8_6:
643 ; SSE42-NEXT: movdqa (%rdi), %xmm0
644 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
645 ; SSE42-NEXT: pextrw $0, %xmm0, (%rsi)
648 ; AVX-LABEL: shuffle_v16i8_to_v2i8_6:
650 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
651 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
652 ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi)
655 ; AVX512-LABEL: shuffle_v16i8_to_v2i8_6:
657 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
658 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
659 ; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi)
661 %vec = load <16 x i8>, <16 x i8>* %L
662 %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 6, i32 14>
663 store <2 x i8> %strided.vec, <2 x i8>* %S
667 define void @shuffle_v16i8_to_v2i8_7(<16 x i8>* %L, <2 x i8>* %S) nounwind {
668 ; SSE2-LABEL: shuffle_v16i8_to_v2i8_7:
670 ; SSE2-NEXT: movdqa (%rdi), %xmm0
671 ; SSE2-NEXT: pxor %xmm1, %xmm1
672 ; SSE2-NEXT: movdqa %xmm0, %xmm2
673 ; SSE2-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
674 ; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
675 ; SSE2-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
676 ; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
677 ; SSE2-NEXT: packuswb %xmm0, %xmm0
678 ; SSE2-NEXT: movd %xmm0, %eax
679 ; SSE2-NEXT: movw %ax, (%rsi)
682 ; SSE42-LABEL: shuffle_v16i8_to_v2i8_7:
684 ; SSE42-NEXT: movdqa (%rdi), %xmm0
685 ; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
686 ; SSE42-NEXT: pextrw $0, %xmm0, (%rsi)
689 ; AVX-LABEL: shuffle_v16i8_to_v2i8_7:
691 ; AVX-NEXT: vmovdqa (%rdi), %xmm0
692 ; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
693 ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi)
696 ; AVX512-LABEL: shuffle_v16i8_to_v2i8_7:
698 ; AVX512-NEXT: vmovdqa (%rdi), %xmm0
699 ; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
700 ; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi)
702 %vec = load <16 x i8>, <16 x i8>* %L
703 %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 7, i32 15>
704 store <2 x i8> %strided.vec, <2 x i8>* %S