| blob | 3d0244f708a8b6c0077ba06915c41068d5214d81 |
1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX1OR2 --check-prefix=AVX1
3 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX1OR2 --check-prefix=AVX2 --check-prefix=AVX2-SLOW
4 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX1OR2 --check-prefix=AVX2 --check-prefix=AVX2-FAST
5 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl | FileCheck %s --check-prefix=ALL --check-prefix=AVX512VL --check-prefix=AVX512VL-SLOW
6 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX512VL --check-prefix=AVX512VL-FAST
8 define <4 x double> @shuffle_v4f64_0000(<4 x double> %a, <4 x double> %b) {
9 ; AVX1-LABEL: shuffle_v4f64_0000:
10 ; AVX1: # %bb.0:
11 ; AVX1-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
12 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
13 ; AVX1-NEXT: retq
14 ;
15 ; AVX2-LABEL: shuffle_v4f64_0000:
16 ; AVX2: # %bb.0:
17 ; AVX2-NEXT: vbroadcastsd %xmm0, %ymm0
18 ; AVX2-NEXT: retq
19 ;
20 ; AVX512VL-LABEL: shuffle_v4f64_0000:
21 ; AVX512VL: # %bb.0:
22 ; AVX512VL-NEXT: vbroadcastsd %xmm0, %ymm0
23 ; AVX512VL-NEXT: retq
24 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 0, i32 0, i32 0>
25 ret <4 x double> %shuffle
26 }
28 define <4 x double> @shuffle_v4f64_0001(<4 x double> %a, <4 x double> %b) {
29 ; AVX1-LABEL: shuffle_v4f64_0001:
30 ; AVX1: # %bb.0:
31 ; AVX1-NEXT: vmovddup {{.*#+}} xmm1 = xmm0[0,0]
32 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
33 ; AVX1-NEXT: retq
34 ;
35 ; AVX2-LABEL: shuffle_v4f64_0001:
36 ; AVX2: # %bb.0:
37 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
38 ; AVX2-NEXT: retq
39 ;
40 ; AVX512VL-LABEL: shuffle_v4f64_0001:
41 ; AVX512VL: # %bb.0:
42 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
43 ; AVX512VL-NEXT: retq
44 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 0, i32 0, i32 1>
45 ret <4 x double> %shuffle
46 }
48 define <4 x double> @shuffle_v4f64_0020(<4 x double> %a, <4 x double> %b) {
49 ; AVX1-LABEL: shuffle_v4f64_0020:
50 ; AVX1: # %bb.0:
51 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
52 ; AVX1-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
53 ; AVX1-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
54 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
55 ; AVX1-NEXT: retq
56 ;
57 ; AVX2-LABEL: shuffle_v4f64_0020:
58 ; AVX2: # %bb.0:
59 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,0]
60 ; AVX2-NEXT: retq
61 ;
62 ; AVX512VL-LABEL: shuffle_v4f64_0020:
63 ; AVX512VL: # %bb.0:
64 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,0]
65 ; AVX512VL-NEXT: retq
66 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 0, i32 2, i32 0>
67 ret <4 x double> %shuffle
68 }
70 define <4 x double> @shuffle_v4f64_0300(<4 x double> %a, <4 x double> %b) {
71 ; AVX1-LABEL: shuffle_v4f64_0300:
72 ; AVX1: # %bb.0:
73 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
74 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,1,2,2]
75 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
76 ; AVX1-NEXT: retq
77 ;
78 ; AVX2-LABEL: shuffle_v4f64_0300:
79 ; AVX2: # %bb.0:
80 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,0,0]
81 ; AVX2-NEXT: retq
82 ;
83 ; AVX512VL-LABEL: shuffle_v4f64_0300:
84 ; AVX512VL: # %bb.0:
85 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,0,0]
86 ; AVX512VL-NEXT: retq
87 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 3, i32 0, i32 0>
88 ret <4 x double> %shuffle
89 }
91 define <4 x double> @shuffle_v4f64_1000(<4 x double> %a, <4 x double> %b) {
92 ; AVX1-LABEL: shuffle_v4f64_1000:
93 ; AVX1: # %bb.0:
94 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
95 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
96 ; AVX1-NEXT: retq
97 ;
98 ; AVX2-LABEL: shuffle_v4f64_1000:
99 ; AVX2: # %bb.0:
100 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,0,0,0]
101 ; AVX2-NEXT: retq
102 ;
103 ; AVX512VL-LABEL: shuffle_v4f64_1000:
104 ; AVX512VL: # %bb.0:
105 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,0,0,0]
106 ; AVX512VL-NEXT: retq
107 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 0, i32 0>
108 ret <4 x double> %shuffle
109 }
111 define <4 x double> @shuffle_v4f64_2200(<4 x double> %a, <4 x double> %b) {
112 ; AVX1-LABEL: shuffle_v4f64_2200:
113 ; AVX1: # %bb.0:
114 ; AVX1-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
115 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
116 ; AVX1-NEXT: retq
117 ;
118 ; AVX2-LABEL: shuffle_v4f64_2200:
119 ; AVX2: # %bb.0:
120 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,0,0]
121 ; AVX2-NEXT: retq
122 ;
123 ; AVX512VL-LABEL: shuffle_v4f64_2200:
124 ; AVX512VL: # %bb.0:
125 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,0,0]
126 ; AVX512VL-NEXT: retq
127 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 2, i32 2, i32 0, i32 0>
128 ret <4 x double> %shuffle
129 }
131 define <4 x double> @shuffle_v4f64_2222(<4 x double> %a, <4 x double> %b) {
132 ; AVX1-LABEL: shuffle_v4f64_2222:
133 ; AVX1: # %bb.0:
134 ; AVX1-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
135 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
136 ; AVX1-NEXT: retq
137 ;
138 ; AVX2-LABEL: shuffle_v4f64_2222:
139 ; AVX2: # %bb.0:
140 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
141 ; AVX2-NEXT: retq
142 ;
143 ; AVX512VL-LABEL: shuffle_v4f64_2222:
144 ; AVX512VL: # %bb.0:
145 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
146 ; AVX512VL-NEXT: retq
147 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
148 ret <4 x double> %shuffle
149 }
151 define <4 x double> @shuffle_v4f64_2222_bc(<4 x i64> %a, <4 x i64> %b) {
152 ; AVX1-LABEL: shuffle_v4f64_2222_bc:
153 ; AVX1: # %bb.0:
154 ; AVX1-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
155 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
156 ; AVX1-NEXT: retq
157 ;
158 ; AVX2-LABEL: shuffle_v4f64_2222_bc:
159 ; AVX2: # %bb.0:
160 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
161 ; AVX2-NEXT: retq
162 ;
163 ; AVX512VL-LABEL: shuffle_v4f64_2222_bc:
164 ; AVX512VL: # %bb.0:
165 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
166 ; AVX512VL-NEXT: retq
167 %tmp0 = bitcast <4 x i64> %a to <4 x double>
168 %tmp1 = bitcast <4 x i64> %b to <4 x double>
169 %shuffle = shufflevector <4 x double> %tmp0, <4 x double> %tmp1, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
170 ret <4 x double> %shuffle
171 }
173 define <4 x double> @shuffle_v4f64_2233(<4 x double> %a, <4 x double> %b) {
174 ; AVX1-LABEL: shuffle_v4f64_2233:
175 ; AVX1: # %bb.0:
176 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
177 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,3]
178 ; AVX1-NEXT: retq
179 ;
180 ; AVX2-LABEL: shuffle_v4f64_2233:
181 ; AVX2: # %bb.0:
182 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,3,3]
183 ; AVX2-NEXT: retq
184 ;
185 ; AVX512VL-LABEL: shuffle_v4f64_2233:
186 ; AVX512VL: # %bb.0:
187 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,3,3]
188 ; AVX512VL-NEXT: retq
189 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 2, i32 2, i32 3, i32 3>
190 ret <4 x double> %shuffle
191 }
193 define <4 x double> @shuffle_v4f64_3330(<4 x double> %a, <4 x double> %b) {
194 ; AVX1-LABEL: shuffle_v4f64_3330:
195 ; AVX1: # %bb.0:
196 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
197 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
198 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,3,2]
199 ; AVX1-NEXT: retq
200 ;
201 ; AVX2-LABEL: shuffle_v4f64_3330:
202 ; AVX2: # %bb.0:
203 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,0]
204 ; AVX2-NEXT: retq
205 ;
206 ; AVX512VL-LABEL: shuffle_v4f64_3330:
207 ; AVX512VL: # %bb.0:
208 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,0]
209 ; AVX512VL-NEXT: retq
210 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 3, i32 3, i32 3, i32 0>
211 ret <4 x double> %shuffle
212 }
214 define <4 x double> @shuffle_v4f64_3210(<4 x double> %a, <4 x double> %b) {
215 ; AVX1-LABEL: shuffle_v4f64_3210:
216 ; AVX1: # %bb.0:
217 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
218 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
219 ; AVX1-NEXT: retq
220 ;
221 ; AVX2-LABEL: shuffle_v4f64_3210:
222 ; AVX2: # %bb.0:
223 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,1,0]
224 ; AVX2-NEXT: retq
225 ;
226 ; AVX512VL-LABEL: shuffle_v4f64_3210:
227 ; AVX512VL: # %bb.0:
228 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,1,0]
229 ; AVX512VL-NEXT: retq
230 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
231 ret <4 x double> %shuffle
232 }
234 define <4 x double> @shuffle_v4f64_0023(<4 x double> %a, <4 x double> %b) {
235 ; ALL-LABEL: shuffle_v4f64_0023:
236 ; ALL: # %bb.0:
237 ; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
238 ; ALL-NEXT: retq
240 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 0, i32 2, i32 3>
241 ret <4 x double> %shuffle
242 }
244 define <4 x double> @shuffle_v4f64_0022(<4 x double> %a, <4 x double> %b) {
245 ; ALL-LABEL: shuffle_v4f64_0022:
246 ; ALL: # %bb.0:
247 ; ALL-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
248 ; ALL-NEXT: retq
249 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
250 ret <4 x double> %shuffle
251 }
253 define <4 x double> @shuffle_v4f64mem_0022(<4 x double>* %ptr, <4 x double> %b) {
254 ; ALL-LABEL: shuffle_v4f64mem_0022:
255 ; ALL: # %bb.0:
256 ; ALL-NEXT: vmovddup {{.*#+}} ymm0 = mem[0,0,2,2]
257 ; ALL-NEXT: retq
258 %a = load <4 x double>, <4 x double>* %ptr
259 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
260 ret <4 x double> %shuffle
261 }
263 define <4 x double> @shuffle_v4f64_1032(<4 x double> %a, <4 x double> %b) {
264 ; ALL-LABEL: shuffle_v4f64_1032:
265 ; ALL: # %bb.0:
266 ; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
267 ; ALL-NEXT: retq
268 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
269 ret <4 x double> %shuffle
270 }
272 define <4 x double> @shuffle_v4f64_1133(<4 x double> %a, <4 x double> %b) {
273 ; ALL-LABEL: shuffle_v4f64_1133:
274 ; ALL: # %bb.0:
275 ; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,3,3]
276 ; ALL-NEXT: retq
277 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 1, i32 3, i32 3>
278 ret <4 x double> %shuffle
279 }
281 define <4 x double> @shuffle_v4f64_1023(<4 x double> %a, <4 x double> %b) {
282 ; ALL-LABEL: shuffle_v4f64_1023:
283 ; ALL: # %bb.0:
284 ; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3]
285 ; ALL-NEXT: retq
286 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
287 ret <4 x double> %shuffle
288 }
290 define <4 x double> @shuffle_v4f64_1022(<4 x double> %a, <4 x double> %b) {
291 ; ALL-LABEL: shuffle_v4f64_1022:
292 ; ALL: # %bb.0:
293 ; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
294 ; ALL-NEXT: retq
295 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 2, i32 2>
296 ret <4 x double> %shuffle
297 }
299 define <4 x double> @shuffle_v4f64_0213(<4 x double> %a, <4 x double> %b) {
300 ; AVX1-LABEL: shuffle_v4f64_0213:
301 ; AVX1: # %bb.0:
302 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
303 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
304 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3]
305 ; AVX1-NEXT: retq
306 ;
307 ; AVX2-LABEL: shuffle_v4f64_0213:
308 ; AVX2: # %bb.0:
309 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
310 ; AVX2-NEXT: retq
311 ;
312 ; AVX512VL-LABEL: shuffle_v4f64_0213:
313 ; AVX512VL: # %bb.0:
314 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
315 ; AVX512VL-NEXT: retq
316 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
317 ret <4 x double> %shuffle
318 }
320 define <4 x double> @shuffle_v4f64_0423(<4 x double> %a, <4 x double> %b) {
321 ; ALL-LABEL: shuffle_v4f64_0423:
322 ; ALL: # %bb.0:
323 ; ALL-NEXT: vmovddup {{.*#+}} xmm1 = xmm1[0,0]
324 ; ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
325 ; ALL-NEXT: retq
326 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 4, i32 2, i32 3>
327 ret <4 x double> %shuffle
328 }
330 define <4 x double> @shuffle_v4f64_0462(<4 x double> %a, <4 x double> %b) {
331 ; ALL-LABEL: shuffle_v4f64_0462:
332 ; ALL: # %bb.0:
333 ; ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
334 ; ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
335 ; ALL-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
336 ; ALL-NEXT: retq
337 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 4, i32 6, i32 2>
338 ret <4 x double> %shuffle
339 }
341 define <4 x double> @shuffle_v4f64_0426(<4 x double> %a, <4 x double> %b) {
342 ; ALL-LABEL: shuffle_v4f64_0426:
343 ; ALL: # %bb.0:
344 ; ALL-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
345 ; ALL-NEXT: retq
346 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
347 ret <4 x double> %shuffle
348 }
350 define <4 x double> @shuffle_v4f64_1537(<4 x double> %a, <4 x double> %b) {
351 ; ALL-LABEL: shuffle_v4f64_1537:
352 ; ALL: # %bb.0:
353 ; ALL-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
354 ; ALL-NEXT: retq
355 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
356 ret <4 x double> %shuffle
357 }
359 define <4 x double> @shuffle_v4f64_4062(<4 x double> %a, <4 x double> %b) {
360 ; ALL-LABEL: shuffle_v4f64_4062:
361 ; ALL: # %bb.0:
362 ; ALL-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
363 ; ALL-NEXT: retq
364 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 4, i32 0, i32 6, i32 2>
365 ret <4 x double> %shuffle
366 }
368 define <4 x double> @shuffle_v4f64_5173(<4 x double> %a, <4 x double> %b) {
369 ; ALL-LABEL: shuffle_v4f64_5173:
370 ; ALL: # %bb.0:
371 ; ALL-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
372 ; ALL-NEXT: retq
373 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 5, i32 1, i32 7, i32 3>
374 ret <4 x double> %shuffle
375 }
377 define <4 x double> @shuffle_v4f64_5163(<4 x double> %a, <4 x double> %b) {
378 ; ALL-LABEL: shuffle_v4f64_5163:
379 ; ALL: # %bb.0:
380 ; ALL-NEXT: vshufpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[2],ymm0[3]
381 ; ALL-NEXT: retq
382 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 5, i32 1, i32 6, i32 3>
383 ret <4 x double> %shuffle
384 }
386 define <4 x double> @shuffle_v4f64_0527(<4 x double> %a, <4 x double> %b) {
387 ; ALL-LABEL: shuffle_v4f64_0527:
388 ; ALL: # %bb.0:
389 ; ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
390 ; ALL-NEXT: retq
391 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
392 ret <4 x double> %shuffle
393 }
395 define <4 x double> @shuffle_v4f64_4163(<4 x double> %a, <4 x double> %b) {
396 ; ALL-LABEL: shuffle_v4f64_4163:
397 ; ALL: # %bb.0:
398 ; ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
399 ; ALL-NEXT: retq
400 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 4, i32 1, i32 6, i32 3>
401 ret <4 x double> %shuffle
402 }
404 define <4 x double> @shuffle_v4f64_0145(<4 x double> %a, <4 x double> %b) {
405 ; ALL-LABEL: shuffle_v4f64_0145:
406 ; ALL: # %bb.0:
407 ; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
408 ; ALL-NEXT: retq
409 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
410 ret <4 x double> %shuffle
411 }
413 define <4 x double> @shuffle_v4f64_4501(<4 x double> %a, <4 x double> %b) {
414 ; ALL-LABEL: shuffle_v4f64_4501:
415 ; ALL: # %bb.0:
416 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
417 ; ALL-NEXT: retq
418 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 4, i32 5, i32 0, i32 1>
419 ret <4 x double> %shuffle
420 }
422 define <4 x double> @shuffle_v4f64_0167(<4 x double> %a, <4 x double> %b) {
423 ; ALL-LABEL: shuffle_v4f64_0167:
424 ; ALL: # %bb.0:
425 ; ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
426 ; ALL-NEXT: retq
427 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 1, i32 6, i32 7>
428 ret <4 x double> %shuffle
429 }
431 define <4 x double> @shuffle_v4f64_1054(<4 x double> %a, <4 x double> %b) {
432 ; ALL-LABEL: shuffle_v4f64_1054:
433 ; ALL: # %bb.0:
434 ; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
435 ; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
436 ; ALL-NEXT: retq
437 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 5, i32 4>
438 ret <4 x double> %shuffle
439 }
441 define <4 x double> @shuffle_v4f64_3254(<4 x double> %a, <4 x double> %b) {
442 ; AVX1OR2-LABEL: shuffle_v4f64_3254:
443 ; AVX1OR2: # %bb.0:
444 ; AVX1OR2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
445 ; AVX1OR2-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
446 ; AVX1OR2-NEXT: retq
447 ;
448 ; AVX512VL-SLOW-LABEL: shuffle_v4f64_3254:
449 ; AVX512VL-SLOW: # %bb.0:
450 ; AVX512VL-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
451 ; AVX512VL-SLOW-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
452 ; AVX512VL-SLOW-NEXT: retq
453 ;
454 ; AVX512VL-FAST-LABEL: shuffle_v4f64_3254:
455 ; AVX512VL-FAST: # %bb.0:
456 ; AVX512VL-FAST-NEXT: vmovapd {{.*#+}} ymm2 = [3,2,5,4]
457 ; AVX512VL-FAST-NEXT: vpermt2pd %ymm1, %ymm2, %ymm0
458 ; AVX512VL-FAST-NEXT: retq
459 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 3, i32 2, i32 5, i32 4>
460 ret <4 x double> %shuffle
461 }
463 define <4 x double> @shuffle_v4f64_3276(<4 x double> %a, <4 x double> %b) {
464 ; AVX1OR2-LABEL: shuffle_v4f64_3276:
465 ; AVX1OR2: # %bb.0:
466 ; AVX1OR2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
467 ; AVX1OR2-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
468 ; AVX1OR2-NEXT: retq
469 ;
470 ; AVX512VL-SLOW-LABEL: shuffle_v4f64_3276:
471 ; AVX512VL-SLOW: # %bb.0:
472 ; AVX512VL-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
473 ; AVX512VL-SLOW-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
474 ; AVX512VL-SLOW-NEXT: retq
475 ;
476 ; AVX512VL-FAST-LABEL: shuffle_v4f64_3276:
477 ; AVX512VL-FAST: # %bb.0:
478 ; AVX512VL-FAST-NEXT: vmovapd {{.*#+}} ymm2 = [3,2,7,6]
479 ; AVX512VL-FAST-NEXT: vpermt2pd %ymm1, %ymm2, %ymm0
480 ; AVX512VL-FAST-NEXT: retq
481 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 3, i32 2, i32 7, i32 6>
482 ret <4 x double> %shuffle
483 }
485 define <4 x double> @shuffle_v4f64_1076(<4 x double> %a, <4 x double> %b) {
486 ; ALL-LABEL: shuffle_v4f64_1076:
487 ; ALL: # %bb.0:
488 ; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
489 ; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
490 ; ALL-NEXT: retq
491 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 7, i32 6>
492 ret <4 x double> %shuffle
493 }
495 define <4 x double> @shuffle_v4f64_0415(<4 x double> %a, <4 x double> %b) {
496 ; AVX1-LABEL: shuffle_v4f64_0415:
497 ; AVX1: # %bb.0:
498 ; AVX1-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm0[1],xmm1[1]
499 ; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
500 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
501 ; AVX1-NEXT: retq
502 ;
503 ; AVX2-LABEL: shuffle_v4f64_0415:
504 ; AVX2: # %bb.0:
505 ; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
506 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
507 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
508 ; AVX2-NEXT: retq
509 ;
510 ; AVX512VL-LABEL: shuffle_v4f64_0415:
511 ; AVX512VL: # %bb.0:
512 ; AVX512VL-NEXT: vmovapd {{.*#+}} ymm2 = [0,4,1,5]
513 ; AVX512VL-NEXT: vpermt2pd %ymm1, %ymm2, %ymm0
514 ; AVX512VL-NEXT: retq
515 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
516 ret <4 x double> %shuffle
517 }
519 define <4 x double> @shuffle_v4f64_u062(<4 x double> %a, <4 x double> %b) {
520 ; ALL-LABEL: shuffle_v4f64_u062:
521 ; ALL: # %bb.0:
522 ; ALL-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
523 ; ALL-NEXT: retq
524 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 undef, i32 0, i32 6, i32 2>
525 ret <4 x double> %shuffle
526 }
528 define <4 x double> @shuffle_v4f64_15uu(<4 x double> %a, <4 x double> %b) {
529 ; ALL-LABEL: shuffle_v4f64_15uu:
530 ; ALL: # %bb.0:
531 ; ALL-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
532 ; ALL-NEXT: retq
533 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 5, i32 undef, i32 undef>
534 ret <4 x double> %shuffle
535 }
537 define <4 x double> @shuffle_v4f64_11uu(<4 x double> %a, <4 x double> %b) {
538 ; ALL-LABEL: shuffle_v4f64_11uu:
539 ; ALL: # %bb.0:
540 ; ALL-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,1]
541 ; ALL-NEXT: retq
542 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 1, i32 undef, i32 undef>
543 ret <4 x double> %shuffle
544 }
546 define <4 x double> @shuffle_v4f64_22uu(<4 x double> %a, <4 x double> %b) {
547 ; AVX1-LABEL: shuffle_v4f64_22uu:
548 ; AVX1: # %bb.0:
549 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
550 ; AVX1-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
551 ; AVX1-NEXT: retq
552 ;
553 ; AVX2-LABEL: shuffle_v4f64_22uu:
554 ; AVX2: # %bb.0:
555 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
556 ; AVX2-NEXT: retq
557 ;
558 ; AVX512VL-LABEL: shuffle_v4f64_22uu:
559 ; AVX512VL: # %bb.0:
560 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
561 ; AVX512VL-NEXT: retq
562 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 2, i32 2, i32 undef, i32 undef>
563 ret <4 x double> %shuffle
564 }
566 define <4 x double> @shuffle_v4f64_3333(<4 x double> %a, <4 x double> %b) {
567 ; AVX1-LABEL: shuffle_v4f64_3333:
568 ; AVX1: # %bb.0:
569 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,3,3]
570 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
571 ; AVX1-NEXT: retq
572 ;
573 ; AVX2-LABEL: shuffle_v4f64_3333:
574 ; AVX2: # %bb.0:
575 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
576 ; AVX2-NEXT: retq
577 ;
578 ; AVX512VL-LABEL: shuffle_v4f64_3333:
579 ; AVX512VL: # %bb.0:
580 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
581 ; AVX512VL-NEXT: retq
582 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
583 ret <4 x double> %shuffle
584 }
586 define <4 x double> @shuffle_v4f64_0z3z(<4 x double> %a, <4 x double> %b) {
587 ; AVX1-LABEL: shuffle_v4f64_0z3z:
588 ; AVX1: # %bb.0:
589 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,2]
590 ; AVX1-NEXT: vxorpd %xmm1, %xmm1, %xmm1
591 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
592 ; AVX1-NEXT: retq
593 ;
594 ; AVX2-SLOW-LABEL: shuffle_v4f64_0z3z:
595 ; AVX2-SLOW: # %bb.0:
596 ; AVX2-SLOW-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,2]
597 ; AVX2-SLOW-NEXT: vxorpd %xmm1, %xmm1, %xmm1
598 ; AVX2-SLOW-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
599 ; AVX2-SLOW-NEXT: retq
600 ;
601 ; AVX2-FAST-LABEL: shuffle_v4f64_0z3z:
602 ; AVX2-FAST: # %bb.0:
603 ; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero
604 ; AVX2-FAST-NEXT: retq
605 ;
606 ; AVX512VL-SLOW-LABEL: shuffle_v4f64_0z3z:
607 ; AVX512VL-SLOW: # %bb.0:
608 ; AVX512VL-SLOW-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,2]
609 ; AVX512VL-SLOW-NEXT: vxorpd %xmm1, %xmm1, %xmm1
610 ; AVX512VL-SLOW-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
611 ; AVX512VL-SLOW-NEXT: retq
612 ;
613 ; AVX512VL-FAST-LABEL: shuffle_v4f64_0z3z:
614 ; AVX512VL-FAST: # %bb.0:
615 ; AVX512VL-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero
616 ; AVX512VL-FAST-NEXT: retq
617 %shuffle = shufflevector <4 x double> %a, <4 x double> <double 0.000000e+00, double undef, double undef, double undef>, <4 x i32> <i32 0, i32 4, i32 3, i32 4>
618 ret <4 x double> %shuffle
619 }
621 define <4 x double> @shuffle_v4f64_1z2z(<4 x double> %a, <4 x double> %b) {
622 ; AVX1-LABEL: shuffle_v4f64_1z2z:
623 ; AVX1: # %bb.0:
624 ; AVX1-NEXT: vxorpd %xmm1, %xmm1, %xmm1
625 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
626 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
627 ; AVX1-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm0[0]
628 ; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
629 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
630 ; AVX1-NEXT: retq
631 ;
632 ; AVX2-SLOW-LABEL: shuffle_v4f64_1z2z:
633 ; AVX2-SLOW: # %bb.0:
634 ; AVX2-SLOW-NEXT: vxorps %xmm1, %xmm1, %xmm1
635 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
636 ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,0,2,0]
637 ; AVX2-SLOW-NEXT: retq
638 ;
639 ; AVX2-FAST-LABEL: shuffle_v4f64_1z2z:
640 ; AVX2-FAST: # %bb.0:
641 ; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19,20,21,22,23],zero,zero,zero,zero,zero,zero,zero,zero
642 ; AVX2-FAST-NEXT: retq
643 ;
644 ; AVX512VL-SLOW-LABEL: shuffle_v4f64_1z2z:
645 ; AVX512VL-SLOW: # %bb.0:
646 ; AVX512VL-SLOW-NEXT: vxorps %xmm1, %xmm1, %xmm1
647 ; AVX512VL-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
648 ; AVX512VL-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,0,2,0]
649 ; AVX512VL-SLOW-NEXT: retq
650 ;
651 ; AVX512VL-FAST-LABEL: shuffle_v4f64_1z2z:
652 ; AVX512VL-FAST: # %bb.0:
653 ; AVX512VL-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19,20,21,22,23],zero,zero,zero,zero,zero,zero,zero,zero
654 ; AVX512VL-FAST-NEXT: retq
655 %1 = shufflevector <4 x double> %a, <4 x double> <double 0.000000e+00, double undef, double undef, double undef>, <4 x i32> <i32 1, i32 4, i32 2, i32 4>
656 ret <4 x double> %1
657 }
659 define <4 x i64> @shuffle_v4i64_0000(<4 x i64> %a, <4 x i64> %b) {
660 ; AVX1-LABEL: shuffle_v4i64_0000:
661 ; AVX1: # %bb.0:
662 ; AVX1-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
663 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
664 ; AVX1-NEXT: retq
665 ;
666 ; AVX2-LABEL: shuffle_v4i64_0000:
667 ; AVX2: # %bb.0:
668 ; AVX2-NEXT: vbroadcastsd %xmm0, %ymm0
669 ; AVX2-NEXT: retq
670 ;
671 ; AVX512VL-LABEL: shuffle_v4i64_0000:
672 ; AVX512VL: # %bb.0:
673 ; AVX512VL-NEXT: vbroadcastsd %xmm0, %ymm0
674 ; AVX512VL-NEXT: retq
675 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 0, i32 0, i32 0>
676 ret <4 x i64> %shuffle
677 }
679 define <4 x i64> @shuffle_v4i64_0001(<4 x i64> %a, <4 x i64> %b) {
680 ; AVX1-LABEL: shuffle_v4i64_0001:
681 ; AVX1: # %bb.0:
682 ; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,1,0,1]
683 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
684 ; AVX1-NEXT: retq
685 ;
686 ; AVX2-LABEL: shuffle_v4i64_0001:
687 ; AVX2: # %bb.0:
688 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
689 ; AVX2-NEXT: retq
690 ;
691 ; AVX512VL-LABEL: shuffle_v4i64_0001:
692 ; AVX512VL: # %bb.0:
693 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
694 ; AVX512VL-NEXT: retq
695 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 0, i32 0, i32 1>
696 ret <4 x i64> %shuffle
697 }
699 define <4 x i64> @shuffle_v4i64_0020(<4 x i64> %a, <4 x i64> %b) {
700 ; AVX1-LABEL: shuffle_v4i64_0020:
701 ; AVX1: # %bb.0:
702 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
703 ; AVX1-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
704 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
705 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
706 ; AVX1-NEXT: retq
707 ;
708 ; AVX2-LABEL: shuffle_v4i64_0020:
709 ; AVX2: # %bb.0:
710 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,0]
711 ; AVX2-NEXT: retq
712 ;
713 ; AVX512VL-LABEL: shuffle_v4i64_0020:
714 ; AVX512VL: # %bb.0:
715 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,0]
716 ; AVX512VL-NEXT: retq
717 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 0, i32 2, i32 0>
718 ret <4 x i64> %shuffle
719 }
721 define <4 x i64> @shuffle_v4i64_0112(<4 x i64> %a, <4 x i64> %b) {
722 ; AVX1-LABEL: shuffle_v4i64_0112:
723 ; AVX1: # %bb.0:
724 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
725 ; AVX1-NEXT: vpalignr {{.*#+}} xmm1 = xmm0[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
726 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
727 ; AVX1-NEXT: retq
728 ;
729 ; AVX2-LABEL: shuffle_v4i64_0112:
730 ; AVX2: # %bb.0:
731 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,2]
732 ; AVX2-NEXT: retq
733 ;
734 ; AVX512VL-LABEL: shuffle_v4i64_0112:
735 ; AVX512VL: # %bb.0:
736 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,2]
737 ; AVX512VL-NEXT: retq
738 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 1, i32 1, i32 2>
739 ret <4 x i64> %shuffle
740 }
742 define <4 x i64> @shuffle_v4i64_0300(<4 x i64> %a, <4 x i64> %b) {
743 ; AVX1-LABEL: shuffle_v4i64_0300:
744 ; AVX1: # %bb.0:
745 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
746 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,1,2,2]
747 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
748 ; AVX1-NEXT: retq
749 ;
750 ; AVX2-LABEL: shuffle_v4i64_0300:
751 ; AVX2: # %bb.0:
752 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,0,0]
753 ; AVX2-NEXT: retq
754 ;
755 ; AVX512VL-LABEL: shuffle_v4i64_0300:
756 ; AVX512VL: # %bb.0:
757 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,0,0]
758 ; AVX512VL-NEXT: retq
759 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 3, i32 0, i32 0>
760 ret <4 x i64> %shuffle
761 }
763 define <4 x i64> @shuffle_v4i64_1000(<4 x i64> %a, <4 x i64> %b) {
764 ; AVX1-LABEL: shuffle_v4i64_1000:
765 ; AVX1: # %bb.0:
766 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
767 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
768 ; AVX1-NEXT: retq
769 ;
770 ; AVX2-LABEL: shuffle_v4i64_1000:
771 ; AVX2: # %bb.0:
772 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,0,0,0]
773 ; AVX2-NEXT: retq
774 ;
775 ; AVX512VL-LABEL: shuffle_v4i64_1000:
776 ; AVX512VL: # %bb.0:
777 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,0,0,0]
778 ; AVX512VL-NEXT: retq
779 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 0, i32 0, i32 0>
780 ret <4 x i64> %shuffle
781 }
783 define <4 x i64> @shuffle_v4i64_2200(<4 x i64> %a, <4 x i64> %b) {
784 ; AVX1-LABEL: shuffle_v4i64_2200:
785 ; AVX1: # %bb.0:
786 ; AVX1-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
787 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
788 ; AVX1-NEXT: retq
789 ;
790 ; AVX2-LABEL: shuffle_v4i64_2200:
791 ; AVX2: # %bb.0:
792 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,0,0]
793 ; AVX2-NEXT: retq
794 ;
795 ; AVX512VL-LABEL: shuffle_v4i64_2200:
796 ; AVX512VL: # %bb.0:
797 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,0,0]
798 ; AVX512VL-NEXT: retq
799 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 2, i32 2, i32 0, i32 0>
800 ret <4 x i64> %shuffle
801 }
803 define <4 x i64> @shuffle_v4i64_3330(<4 x i64> %a, <4 x i64> %b) {
804 ; AVX1-LABEL: shuffle_v4i64_3330:
805 ; AVX1: # %bb.0:
806 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
807 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
808 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,3,2]
809 ; AVX1-NEXT: retq
810 ;
811 ; AVX2-LABEL: shuffle_v4i64_3330:
812 ; AVX2: # %bb.0:
813 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,0]
814 ; AVX2-NEXT: retq
815 ;
816 ; AVX512VL-LABEL: shuffle_v4i64_3330:
817 ; AVX512VL: # %bb.0:
818 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,0]
819 ; AVX512VL-NEXT: retq
820 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 3, i32 3, i32 3, i32 0>
821 ret <4 x i64> %shuffle
822 }
824 define <4 x i64> @shuffle_v4i64_3210(<4 x i64> %a, <4 x i64> %b) {
825 ; AVX1-LABEL: shuffle_v4i64_3210:
826 ; AVX1: # %bb.0:
827 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
828 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
829 ; AVX1-NEXT: retq
830 ;
831 ; AVX2-LABEL: shuffle_v4i64_3210:
832 ; AVX2: # %bb.0:
833 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,1,0]
834 ; AVX2-NEXT: retq
835 ;
836 ; AVX512VL-LABEL: shuffle_v4i64_3210:
837 ; AVX512VL: # %bb.0:
838 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,1,0]
839 ; AVX512VL-NEXT: retq
840 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
841 ret <4 x i64> %shuffle
842 }
844 define <4 x i64> @shuffle_v4i64_0213(<4 x i64> %a, <4 x i64> %b) {
845 ; AVX1-LABEL: shuffle_v4i64_0213:
846 ; AVX1: # %bb.0:
847 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
848 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
849 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3]
850 ; AVX1-NEXT: retq
851 ;
852 ; AVX2-LABEL: shuffle_v4i64_0213:
853 ; AVX2: # %bb.0:
854 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
855 ; AVX2-NEXT: retq
856 ;
857 ; AVX512VL-LABEL: shuffle_v4i64_0213:
858 ; AVX512VL: # %bb.0:
859 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
860 ; AVX512VL-NEXT: retq
861 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
862 ret <4 x i64> %shuffle
863 }
865 define <4 x i64> @shuffle_v4i64_0124(<4 x i64> %a, <4 x i64> %b) {
866 ; AVX1-LABEL: shuffle_v4i64_0124:
867 ; AVX1: # %bb.0:
868 ; AVX1-NEXT: vmovddup {{.*#+}} xmm1 = xmm1[0,0]
869 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
870 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
871 ; AVX1-NEXT: retq
872 ;
873 ; AVX2-LABEL: shuffle_v4i64_0124:
874 ; AVX2: # %bb.0:
875 ; AVX2-NEXT: vbroadcastsd %xmm1, %ymm1
876 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
877 ; AVX2-NEXT: retq
878 ;
879 ; AVX512VL-SLOW-LABEL: shuffle_v4i64_0124:
880 ; AVX512VL-SLOW: # %bb.0:
881 ; AVX512VL-SLOW-NEXT: vbroadcastsd %xmm1, %ymm1
882 ; AVX512VL-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
883 ; AVX512VL-SLOW-NEXT: retq
884 ;
885 ; AVX512VL-FAST-LABEL: shuffle_v4i64_0124:
886 ; AVX512VL-FAST: # %bb.0:
887 ; AVX512VL-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [0,1,2,4]
888 ; AVX512VL-FAST-NEXT: vpermt2q %ymm1, %ymm2, %ymm0
889 ; AVX512VL-FAST-NEXT: retq
890 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 1, i32 2, i32 4>
891 ret <4 x i64> %shuffle
892 }
894 define <4 x i64> @shuffle_v4i64_0142(<4 x i64> %a, <4 x i64> %b) {
895 ; AVX1-LABEL: shuffle_v4i64_0142:
896 ; AVX1: # %bb.0:
897 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
898 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,1,2,2]
899 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3]
900 ; AVX1-NEXT: retq
901 ;
902 ; AVX2-LABEL: shuffle_v4i64_0142:
903 ; AVX2: # %bb.0:
904 ; AVX2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
905 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,2,2]
906 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
907 ; AVX2-NEXT: retq
908 ;
909 ; AVX512VL-LABEL: shuffle_v4i64_0142:
910 ; AVX512VL: # %bb.0:
911 ; AVX512VL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
912 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,2,2]
913 ; AVX512VL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
914 ; AVX512VL-NEXT: retq
915 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 1, i32 4, i32 2>
916 ret <4 x i64> %shuffle
917 }
919 define <4 x i64> @shuffle_v4i64_0412(<4 x i64> %a, <4 x i64> %b) {
920 ; AVX1-LABEL: shuffle_v4i64_0412:
921 ; AVX1: # %bb.0:
922 ; AVX1-NEXT: vmovddup {{.*#+}} xmm1 = xmm1[0,0]
923 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
924 ; AVX1-NEXT: vpalignr {{.*#+}} xmm2 = xmm0[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
925 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
926 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
927 ; AVX1-NEXT: retq
928 ;
929 ; AVX2-LABEL: shuffle_v4i64_0412:
930 ; AVX2: # %bb.0:
931 ; AVX2-NEXT: vmovddup {{.*#+}} xmm1 = xmm1[0,0]
932 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,2]
933 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
934 ; AVX2-NEXT: retq
935 ;
936 ; AVX512VL-LABEL: shuffle_v4i64_0412:
937 ; AVX512VL: # %bb.0:
938 ; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm2 = [0,4,1,2]
939 ; AVX512VL-NEXT: vpermt2q %ymm1, %ymm2, %ymm0
940 ; AVX512VL-NEXT: retq
941 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 4, i32 1, i32 2>
942 ret <4 x i64> %shuffle
943 }
945 define <4 x i64> @shuffle_v4i64_4012(<4 x i64> %a, <4 x i64> %b) {
946 ; AVX1-LABEL: shuffle_v4i64_4012:
947 ; AVX1: # %bb.0:
948 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
949 ; AVX1-NEXT: vpalignr {{.*#+}} xmm2 = xmm0[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
950 ; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
951 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
952 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
953 ; AVX1-NEXT: retq
954 ;
955 ; AVX2-LABEL: shuffle_v4i64_4012:
956 ; AVX2: # %bb.0:
957 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,2]
958 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
959 ; AVX2-NEXT: retq
960 ;
961 ; AVX512VL-SLOW-LABEL: shuffle_v4i64_4012:
962 ; AVX512VL-SLOW: # %bb.0:
963 ; AVX512VL-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,2]
964 ; AVX512VL-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
965 ; AVX512VL-SLOW-NEXT: retq
966 ;
967 ; AVX512VL-FAST-LABEL: shuffle_v4i64_4012:
968 ; AVX512VL-FAST: # %bb.0:
969 ; AVX512VL-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [4,0,1,2]
970 ; AVX512VL-FAST-NEXT: vpermt2q %ymm1, %ymm2, %ymm0
971 ; AVX512VL-FAST-NEXT: retq
972 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 4, i32 0, i32 1, i32 2>
973 ret <4 x i64> %shuffle
974 }
976 define <4 x i64> @shuffle_v4i64_0145(<4 x i64> %a, <4 x i64> %b) {
977 ; ALL-LABEL: shuffle_v4i64_0145:
978 ; ALL: # %bb.0:
979 ; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
980 ; ALL-NEXT: retq
981 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
982 ret <4 x i64> %shuffle
983 }
985 define <4 x i64> @shuffle_v4i64_0451(<4 x i64> %a, <4 x i64> %b) {
986 ; AVX1-LABEL: shuffle_v4i64_0451:
987 ; AVX1: # %bb.0:
988 ; AVX1-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm1[1],xmm0[1]
989 ; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
990 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
991 ; AVX1-NEXT: retq
992 ;
993 ; AVX2-LABEL: shuffle_v4i64_0451:
994 ; AVX2: # %bb.0:
995 ; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,0,1,3]
996 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,2,1]
997 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5],ymm0[6,7]
998 ; AVX2-NEXT: retq
999 ;
1000 ; AVX512VL-LABEL: shuffle_v4i64_0451:
1001 ; AVX512VL: # %bb.0:
1002 ; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm2 = [0,4,5,1]
1003 ; AVX512VL-NEXT: vpermt2q %ymm1, %ymm2, %ymm0
1004 ; AVX512VL-NEXT: retq
1005 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 4, i32 5, i32 1>
1006 ret <4 x i64> %shuffle
1007 }
1009 define <4 x i64> @shuffle_v4i64_4501(<4 x i64> %a, <4 x i64> %b) {
1010 ; ALL-LABEL: shuffle_v4i64_4501:
1011 ; ALL: # %bb.0:
1012 ; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
1013 ; ALL-NEXT: retq
1014 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 4, i32 5, i32 0, i32 1>
1015 ret <4 x i64> %shuffle
1016 }
1018 define <4 x i64> @shuffle_v4i64_4015(<4 x i64> %a, <4 x i64> %b) {
1019 ; AVX1-LABEL: shuffle_v4i64_4015:
1020 ; AVX1: # %bb.0:
1021 ; AVX1-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm0[1],xmm1[1]
1022 ; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm1[0],xmm0[0]
1023 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1024 ; AVX1-NEXT: retq
1025 ;
1026 ; AVX2-LABEL: shuffle_v4i64_4015:
1027 ; AVX2: # %bb.0:
1028 ; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
1029 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,3]
1030 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
1031 ; AVX2-NEXT: retq
1032 ;
1033 ; AVX512VL-LABEL: shuffle_v4i64_4015:
1034 ; AVX512VL: # %bb.0:
1035 ; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm2 = [4,0,1,5]
1036 ; AVX512VL-NEXT: vpermt2q %ymm1, %ymm2, %ymm0
1037 ; AVX512VL-NEXT: retq
1038 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 4, i32 0, i32 1, i32 5>
1039 ret <4 x i64> %shuffle
1040 }
1042 define <4 x i64> @shuffle_v4i64_2u35(<4 x i64> %a, <4 x i64> %b) {
1043 ; AVX1-LABEL: shuffle_v4i64_2u35:
1044 ; AVX1: # %bb.0:
1045 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
1046 ; AVX1-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm0[1],xmm1[1]
1047 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
1048 ; AVX1-NEXT: retq
1049 ;
1050 ; AVX2-LABEL: shuffle_v4i64_2u35:
1051 ; AVX2: # %bb.0:
1052 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
1053 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,1]
1054 ; AVX2-NEXT: retq
1055 ;
1056 ; AVX512VL-SLOW-LABEL: shuffle_v4i64_2u35:
1057 ; AVX512VL-SLOW: # %bb.0:
1058 ; AVX512VL-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
1059 ; AVX512VL-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,1]
1060 ; AVX512VL-SLOW-NEXT: retq
1061 ;
1062 ; AVX512VL-FAST-LABEL: shuffle_v4i64_2u35:
1063 ; AVX512VL-FAST: # %bb.0:
1064 ; AVX512VL-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [2,5,3,5]
1065 ; AVX512VL-FAST-NEXT: vpermt2q %ymm1, %ymm2, %ymm0
1066 ; AVX512VL-FAST-NEXT: retq
1067 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 2, i32 undef, i32 3, i32 5>
1068 ret <4 x i64> %shuffle
1069 }
1071 define <4 x i64> @shuffle_v4i64_1251(<4 x i64> %a, <4 x i64> %b) {
1072 ; AVX1-LABEL: shuffle_v4i64_1251:
1073 ; AVX1: # %bb.0:
1074 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
1075 ; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm2[0],ymm0[2],ymm2[3]
1076 ; AVX1-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
1077 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
1078 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3]
1079 ; AVX1-NEXT: retq
1080 ;
1081 ; AVX2-LABEL: shuffle_v4i64_1251:
1082 ; AVX2: # %bb.0:
1083 ; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,3]
1084 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,2,2,1]
1085 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
1086 ; AVX2-NEXT: retq
1087 ;
1088 ; AVX512VL-LABEL: shuffle_v4i64_1251:
1089 ; AVX512VL: # %bb.0:
1090 ; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm2 = [1,2,5,1]
1091 ; AVX512VL-NEXT: vpermt2q %ymm1, %ymm2, %ymm0
1092 ; AVX512VL-NEXT: retq
1093 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 2, i32 5, i32 1>
1094 ret <4 x i64> %shuffle
1095 }
1097 define <4 x i64> @shuffle_v4i64_1054(<4 x i64> %a, <4 x i64> %b) {
1098 ; AVX1-LABEL: shuffle_v4i64_1054:
1099 ; AVX1: # %bb.0:
1100 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
1101 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
1102 ; AVX1-NEXT: retq
1103 ;
1104 ; AVX2-LABEL: shuffle_v4i64_1054:
1105 ; AVX2: # %bb.0:
1106 ; AVX2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
1107 ; AVX2-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
1108 ; AVX2-NEXT: retq
1109 ;
1110 ; AVX512VL-LABEL: shuffle_v4i64_1054:
1111 ; AVX512VL: # %bb.0:
1112 ; AVX512VL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
1113 ; AVX512VL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
1114 ; AVX512VL-NEXT: retq
1115 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 0, i32 5, i32 4>
1116 ret <4 x i64> %shuffle
1117 }
1119 define <4 x i64> @shuffle_v4i64_3254(<4 x i64> %a, <4 x i64> %b) {
1120 ; AVX1-LABEL: shuffle_v4i64_3254:
1121 ; AVX1: # %bb.0:
1122 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
1123 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
1124 ; AVX1-NEXT: retq
1125 ;
1126 ; AVX2-LABEL: shuffle_v4i64_3254:
1127 ; AVX2: # %bb.0:
1128 ; AVX2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
1129 ; AVX2-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
1130 ; AVX2-NEXT: retq
1131 ;
1132 ; AVX512VL-SLOW-LABEL: shuffle_v4i64_3254:
1133 ; AVX512VL-SLOW: # %bb.0:
1134 ; AVX512VL-SLOW-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
1135 ; AVX512VL-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
1136 ; AVX512VL-SLOW-NEXT: retq
1137 ;
1138 ; AVX512VL-FAST-LABEL: shuffle_v4i64_3254:
1139 ; AVX512VL-FAST: # %bb.0:
1140 ; AVX512VL-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [3,2,5,4]
1141 ; AVX512VL-FAST-NEXT: vpermt2q %ymm1, %ymm2, %ymm0
1142 ; AVX512VL-FAST-NEXT: retq
1143 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 3, i32 2, i32 5, i32 4>
1144 ret <4 x i64> %shuffle
1145 }
1147 define <4 x i64> @shuffle_v4i64_3276(<4 x i64> %a, <4 x i64> %b) {
1148 ; AVX1-LABEL: shuffle_v4i64_3276:
1149 ; AVX1: # %bb.0:
1150 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
1151 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
1152 ; AVX1-NEXT: retq
1153 ;
1154 ; AVX2-LABEL: shuffle_v4i64_3276:
1155 ; AVX2: # %bb.0:
1156 ; AVX2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
1157 ; AVX2-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
1158 ; AVX2-NEXT: retq
1159 ;
1160 ; AVX512VL-SLOW-LABEL: shuffle_v4i64_3276:
1161 ; AVX512VL-SLOW: # %bb.0:
1162 ; AVX512VL-SLOW-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
1163 ; AVX512VL-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
1164 ; AVX512VL-SLOW-NEXT: retq
1165 ;
1166 ; AVX512VL-FAST-LABEL: shuffle_v4i64_3276:
1167 ; AVX512VL-FAST: # %bb.0:
1168 ; AVX512VL-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [3,2,7,6]
1169 ; AVX512VL-FAST-NEXT: vpermt2q %ymm1, %ymm2, %ymm0
1170 ; AVX512VL-FAST-NEXT: retq
1171 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 3, i32 2, i32 7, i32 6>
1172 ret <4 x i64> %shuffle
1173 }
1175 define <4 x i64> @shuffle_v4i64_1076(<4 x i64> %a, <4 x i64> %b) {
1176 ; AVX1-LABEL: shuffle_v4i64_1076:
1177 ; AVX1: # %bb.0:
1178 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
1179 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
1180 ; AVX1-NEXT: retq
1181 ;
1182 ; AVX2-LABEL: shuffle_v4i64_1076:
1183 ; AVX2: # %bb.0:
1184 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
1185 ; AVX2-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
1186 ; AVX2-NEXT: retq
1187 ;
1188 ; AVX512VL-LABEL: shuffle_v4i64_1076:
1189 ; AVX512VL: # %bb.0:
1190 ; AVX512VL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
1191 ; AVX512VL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
1192 ; AVX512VL-NEXT: retq
1193 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 0, i32 7, i32 6>
1194 ret <4 x i64> %shuffle
1195 }
1197 define <4 x i64> @shuffle_v4i64_0415(<4 x i64> %a, <4 x i64> %b) {
1198 ; AVX1-LABEL: shuffle_v4i64_0415:
1199 ; AVX1: # %bb.0:
1200 ; AVX1-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm0[1],xmm1[1]
1201 ; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
1202 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1203 ; AVX1-NEXT: retq
1204 ;
1205 ; AVX2-LABEL: shuffle_v4i64_0415:
1206 ; AVX2: # %bb.0:
1207 ; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
1208 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
1209 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
1210 ; AVX2-NEXT: retq
1211 ;
1212 ; AVX512VL-LABEL: shuffle_v4i64_0415:
1213 ; AVX512VL: # %bb.0:
1214 ; AVX512VL-NEXT: vmovdqa {{.*#+}} ymm2 = [0,4,1,5]
1215 ; AVX512VL-NEXT: vpermt2q %ymm1, %ymm2, %ymm0
1216 ; AVX512VL-NEXT: retq
1217 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
1218 ret <4 x i64> %shuffle
1219 }
1221 define <4 x i64> @shuffle_v4i64_z4z6(<4 x i64> %a) {
1222 ; AVX1-LABEL: shuffle_v4i64_z4z6:
1223 ; AVX1: # %bb.0:
1224 ; AVX1-NEXT: vxorps %xmm1, %xmm1, %xmm1
1225 ; AVX1-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
1226 ; AVX1-NEXT: retq
1227 ;
1228 ; AVX2-LABEL: shuffle_v4i64_z4z6:
1229 ; AVX2: # %bb.0:
1230 ; AVX2-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,2,3,4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19,20,21,22,23]
1231 ; AVX2-NEXT: retq
1232 ;
1233 ; AVX512VL-LABEL: shuffle_v4i64_z4z6:
1234 ; AVX512VL: # %bb.0:
1235 ; AVX512VL-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,2,3,4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19,20,21,22,23]
1236 ; AVX512VL-NEXT: retq
1237 %shuffle = shufflevector <4 x i64> zeroinitializer, <4 x i64> %a, <4 x i32> <i32 0, i32 4, i32 0, i32 6>
1238 ret <4 x i64> %shuffle
1239 }
1241 define <4 x i64> @shuffle_v4i64_5zuz(<4 x i64> %a) {
1242 ; AVX1-LABEL: shuffle_v4i64_5zuz:
1243 ; AVX1: # %bb.0:
1244 ; AVX1-NEXT: vxorps %xmm1, %xmm1, %xmm1
1245 ; AVX1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
1246 ; AVX1-NEXT: retq
1247 ;
1248 ; AVX2-LABEL: shuffle_v4i64_5zuz:
1249 ; AVX2: # %bb.0:
1250 ; AVX2-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero
1251 ; AVX2-NEXT: retq
1252 ;
1253 ; AVX512VL-LABEL: shuffle_v4i64_5zuz:
1254 ; AVX512VL: # %bb.0:
1255 ; AVX512VL-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero
1256 ; AVX512VL-NEXT: retq
1257 %shuffle = shufflevector <4 x i64> zeroinitializer, <4 x i64> %a, <4 x i32> <i32 5, i32 0, i32 undef, i32 0>
1258 ret <4 x i64> %shuffle
1259 }
1261 define <4 x i64> @shuffle_v4i64_40u2(<4 x i64> %a, <4 x i64> %b) {
1262 ; ALL-LABEL: shuffle_v4i64_40u2:
1263 ; ALL: # %bb.0:
1264 ; ALL-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
1265 ; ALL-NEXT: retq
1266 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 4, i32 0, i32 undef, i32 2>
1267 ret <4 x i64> %shuffle
1268 }
1270 define <4 x i64> @shuffle_v4i64_15uu(<4 x i64> %a, <4 x i64> %b) {
1271 ; ALL-LABEL: shuffle_v4i64_15uu:
1272 ; ALL: # %bb.0:
1273 ; ALL-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
1274 ; ALL-NEXT: retq
1275 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 5, i32 undef, i32 undef>
1276 ret <4 x i64> %shuffle
1277 }
1279 define <4 x i64> @shuffle_v4i64_11uu(<4 x i64> %a, <4 x i64> %b) {
1280 ; ALL-LABEL: shuffle_v4i64_11uu:
1281 ; ALL: # %bb.0:
1282 ; ALL-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
1283 ; ALL-NEXT: retq
1284 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 1, i32 undef, i32 undef>
1285 ret <4 x i64> %shuffle
1286 }
1288 define <4 x i64> @shuffle_v4i64_22uu(<4 x i64> %a, <4 x i64> %b) {
1289 ; AVX1-LABEL: shuffle_v4i64_22uu:
1290 ; AVX1: # %bb.0:
1291 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
1292 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
1293 ; AVX1-NEXT: retq
1294 ;
1295 ; AVX2-LABEL: shuffle_v4i64_22uu:
1296 ; AVX2: # %bb.0:
1297 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
1298 ; AVX2-NEXT: retq
1299 ;
1300 ; AVX512VL-LABEL: shuffle_v4i64_22uu:
1301 ; AVX512VL: # %bb.0:
1302 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
1303 ; AVX512VL-NEXT: retq
1304 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 2, i32 2, i32 undef, i32 undef>
1305 ret <4 x i64> %shuffle
1306 }
1308 define <4 x i64> @shuffle_v4i64_3333(<4 x i64> %a, <4 x i64> %b) {
1309 ; AVX1-LABEL: shuffle_v4i64_3333:
1310 ; AVX1: # %bb.0:
1311 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,3,3]
1312 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
1313 ; AVX1-NEXT: retq
1314 ;
1315 ; AVX2-LABEL: shuffle_v4i64_3333:
1316 ; AVX2: # %bb.0:
1317 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
1318 ; AVX2-NEXT: retq
1319 ;
1320 ; AVX512VL-LABEL: shuffle_v4i64_3333:
1321 ; AVX512VL: # %bb.0:
1322 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
1323 ; AVX512VL-NEXT: retq
1324 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
1325 ret <4 x i64> %shuffle
1326 }
1328 define <4 x i64> @shuffle_v4i64_1z3z(<4 x i64> %a, <4 x i64> %b) {
1329 ; AVX1-LABEL: shuffle_v4i64_1z3z:
1330 ; AVX1: # %bb.0:
1331 ; AVX1-NEXT: vxorps %xmm1, %xmm1, %xmm1
1332 ; AVX1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
1333 ; AVX1-NEXT: retq
1334 ;
1335 ; AVX2-LABEL: shuffle_v4i64_1z3z:
1336 ; AVX2: # %bb.0:
1337 ; AVX2-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero
1338 ; AVX2-NEXT: retq
1339 ;
1340 ; AVX512VL-LABEL: shuffle_v4i64_1z3z:
1341 ; AVX512VL: # %bb.0:
1342 ; AVX512VL-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero
1343 ; AVX512VL-NEXT: retq
1344 %shuffle = shufflevector <4 x i64> %a, <4 x i64> <i64 0, i64 undef, i64 undef, i64 undef>, <4 x i32> <i32 1, i32 4, i32 3, i32 4>
1345 ret <4 x i64> %shuffle
1346 }
1348 define <4 x i64> @stress_test1(<4 x i64> %a, <4 x i64> %b) {
1349 ; ALL-LABEL: stress_test1:
1350 ; ALL: retq
1351 %c = shufflevector <4 x i64> %b, <4 x i64> undef, <4 x i32> <i32 3, i32 1, i32 1, i32 0>
1352 %d = shufflevector <4 x i64> %c, <4 x i64> undef, <4 x i32> <i32 3, i32 undef, i32 2, i32 undef>
1353 %e = shufflevector <4 x i64> %b, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 undef>
1354 %f = shufflevector <4 x i64> %d, <4 x i64> %e, <4 x i32> <i32 5, i32 1, i32 1, i32 0>
1356 ret <4 x i64> %f
1357 }
1359 define <4 x i64> @insert_reg_and_zero_v4i64(i64 %a) {
1360 ; ALL-LABEL: insert_reg_and_zero_v4i64:
1361 ; ALL: # %bb.0:
1362 ; ALL-NEXT: vmovq %rdi, %xmm0
1363 ; ALL-NEXT: retq
1364 %v = insertelement <4 x i64> undef, i64 %a, i64 0
1365 %shuffle = shufflevector <4 x i64> %v, <4 x i64> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
1366 ret <4 x i64> %shuffle
1367 }
1369 define <4 x i64> @insert_mem_and_zero_v4i64(i64* %ptr) {
1370 ; ALL-LABEL: insert_mem_and_zero_v4i64:
1371 ; ALL: # %bb.0:
1372 ; ALL-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
1373 ; ALL-NEXT: retq
1374 %a = load i64, i64* %ptr
1375 %v = insertelement <4 x i64> undef, i64 %a, i64 0
1376 %shuffle = shufflevector <4 x i64> %v, <4 x i64> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
1377 ret <4 x i64> %shuffle
1378 }
1380 define <4 x double> @insert_reg_and_zero_v4f64(double %a) {
1381 ; ALL-LABEL: insert_reg_and_zero_v4f64:
1382 ; ALL: # %bb.0:
1383 ; ALL-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
1384 ; ALL-NEXT: vxorps %xmm1, %xmm1, %xmm1
1385 ; ALL-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
1386 ; ALL-NEXT: retq
1387 %v = insertelement <4 x double> undef, double %a, i32 0
1388 %shuffle = shufflevector <4 x double> %v, <4 x double> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
1389 ret <4 x double> %shuffle
1390 }
1392 define <4 x double> @insert_mem_and_zero_v4f64(double* %ptr) {
1393 ; ALL-LABEL: insert_mem_and_zero_v4f64:
1394 ; ALL: # %bb.0:
1395 ; ALL-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
1396 ; ALL-NEXT: retq
1397 %a = load double, double* %ptr
1398 %v = insertelement <4 x double> undef, double %a, i32 0
1399 %shuffle = shufflevector <4 x double> %v, <4 x double> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
1400 ret <4 x double> %shuffle
1401 }
1403 define <4 x double> @splat_mem_v4f64(double* %ptr) {
1404 ; ALL-LABEL: splat_mem_v4f64:
1405 ; ALL: # %bb.0:
1406 ; ALL-NEXT: vbroadcastsd (%rdi), %ymm0
1407 ; ALL-NEXT: retq
1408 %a = load double, double* %ptr
1409 %v = insertelement <4 x double> undef, double %a, i32 0
1410 %shuffle = shufflevector <4 x double> %v, <4 x double> undef, <4 x i32> <i32 0, i32 0, i32 0, i32 0>
1411 ret <4 x double> %shuffle
1412 }
1414 define <4 x i64> @splat_mem_v4i64(i64* %ptr) {
1415 ; ALL-LABEL: splat_mem_v4i64:
1416 ; ALL: # %bb.0:
1417 ; ALL-NEXT: vbroadcastsd (%rdi), %ymm0
1418 ; ALL-NEXT: retq
1419 %a = load i64, i64* %ptr
1420 %v = insertelement <4 x i64> undef, i64 %a, i64 0
1421 %shuffle = shufflevector <4 x i64> %v, <4 x i64> undef, <4 x i32> <i32 0, i32 0, i32 0, i32 0>
1422 ret <4 x i64> %shuffle
1423 }
1425 define <4 x double> @splat_mem_v4f64_2(double* %p) {
1426 ; ALL-LABEL: splat_mem_v4f64_2:
1427 ; ALL: # %bb.0:
1428 ; ALL-NEXT: vbroadcastsd (%rdi), %ymm0
1429 ; ALL-NEXT: retq
1430 %1 = load double, double* %p
1431 %2 = insertelement <2 x double> undef, double %1, i32 0
1432 %3 = shufflevector <2 x double> %2, <2 x double> undef, <4 x i32> zeroinitializer
1433 ret <4 x double> %3
1434 }
1436 define <4 x double> @splat_v4f64(<2 x double> %r) {
1437 ; AVX1-LABEL: splat_v4f64:
1438 ; AVX1: # %bb.0:
1439 ; AVX1-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
1440 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1441 ; AVX1-NEXT: retq
1442 ;
1443 ; AVX2-LABEL: splat_v4f64:
1444 ; AVX2: # %bb.0:
1445 ; AVX2-NEXT: vbroadcastsd %xmm0, %ymm0
1446 ; AVX2-NEXT: retq
1447 ;
1448 ; AVX512VL-LABEL: splat_v4f64:
1449 ; AVX512VL: # %bb.0:
1450 ; AVX512VL-NEXT: vbroadcastsd %xmm0, %ymm0
1451 ; AVX512VL-NEXT: retq
1452 %1 = shufflevector <2 x double> %r, <2 x double> undef, <4 x i32> zeroinitializer
1453 ret <4 x double> %1
1454 }
1456 define <4 x i64> @splat_mem_v4i64_from_v2i64(<2 x i64>* %ptr) {
1457 ; ALL-LABEL: splat_mem_v4i64_from_v2i64:
1458 ; ALL: # %bb.0:
1459 ; ALL-NEXT: vbroadcastsd (%rdi), %ymm0
1460 ; ALL-NEXT: retq
1461 %v = load <2 x i64>, <2 x i64>* %ptr
1462 %shuffle = shufflevector <2 x i64> %v, <2 x i64> undef, <4 x i32> <i32 0, i32 0, i32 0, i32 0>
1463 ret <4 x i64> %shuffle
1464 }
1466 define <4 x double> @splat_mem_v4f64_from_v2f64(<2 x double>* %ptr) {
1467 ; ALL-LABEL: splat_mem_v4f64_from_v2f64:
1468 ; ALL: # %bb.0:
1469 ; ALL-NEXT: vbroadcastsd (%rdi), %ymm0
1470 ; ALL-NEXT: retq
1471 %v = load <2 x double>, <2 x double>* %ptr
1472 %shuffle = shufflevector <2 x double> %v, <2 x double> undef, <4 x i32> <i32 0, i32 0, i32 0, i32 0>
1473 ret <4 x double> %shuffle
1474 }
1476 define <4 x i64> @splat128_mem_v4i64_from_v2i64(<2 x i64>* %ptr) {
1477 ; AVX1OR2-LABEL: splat128_mem_v4i64_from_v2i64:
1478 ; AVX1OR2: # %bb.0:
1479 ; AVX1OR2-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
1480 ; AVX1OR2-NEXT: retq
1481 ;
1482 ; AVX512VL-LABEL: splat128_mem_v4i64_from_v2i64:
1483 ; AVX512VL: # %bb.0:
1484 ; AVX512VL-NEXT: vbroadcasti128 {{.*#+}} ymm0 = mem[0,1,0,1]
1485 ; AVX512VL-NEXT: retq
1486 %v = load <2 x i64>, <2 x i64>* %ptr
1487 %shuffle = shufflevector <2 x i64> %v, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
1488 ret <4 x i64> %shuffle
1489 }
1491 define <4 x double> @splat128_mem_v4f64_from_v2f64(<2 x double>* %ptr) {
1492 ; ALL-LABEL: splat128_mem_v4f64_from_v2f64:
1493 ; ALL: # %bb.0:
1494 ; ALL-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
1495 ; ALL-NEXT: retq
1496 %v = load <2 x double>, <2 x double>* %ptr
1497 %shuffle = shufflevector <2 x double> %v, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
1498 ret <4 x double> %shuffle
1499 }
1501 define <4 x double> @broadcast_v4f64_0000_from_v2i64(<2 x i64> %a0) {
1502 ; AVX1-LABEL: broadcast_v4f64_0000_from_v2i64:
1503 ; AVX1: # %bb.0:
1504 ; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
1505 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1506 ; AVX1-NEXT: retq
1507 ;
1508 ; AVX2-LABEL: broadcast_v4f64_0000_from_v2i64:
1509 ; AVX2: # %bb.0:
1510 ; AVX2-NEXT: vbroadcastsd %xmm0, %ymm0
1511 ; AVX2-NEXT: retq
1512 ;
1513 ; AVX512VL-LABEL: broadcast_v4f64_0000_from_v2i64:
1514 ; AVX512VL: # %bb.0:
1515 ; AVX512VL-NEXT: vbroadcastsd %xmm0, %ymm0
1516 ; AVX512VL-NEXT: retq
1517 %1 = shufflevector <2 x i64> %a0, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
1518 %2 = bitcast <4 x i64> %1 to <4 x double>
1519 %3 = shufflevector <4 x double> %2, <4 x double> undef, <4 x i32> zeroinitializer
1520 ret <4 x double> %3
1521 }
1523 define <4 x double> @bitcast_v4f64_0426(<4 x double> %a, <4 x double> %b) {
1524 ; ALL-LABEL: bitcast_v4f64_0426:
1525 ; ALL: # %bb.0:
1526 ; ALL-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
1527 ; ALL-NEXT: retq
1528 %shuffle64 = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 4, i32 0, i32 6, i32 2>
1529 %bitcast32 = bitcast <4 x double> %shuffle64 to <8 x float>
1530 %shuffle32 = shufflevector <8 x float> %bitcast32, <8 x float> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
1531 %bitcast16 = bitcast <8 x float> %shuffle32 to <16 x i16>
1532 %shuffle16 = shufflevector <16 x i16> %bitcast16, <16 x i16> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 6, i32 7, i32 4, i32 5, i32 10, i32 11, i32 8, i32 9, i32 14, i32 15, i32 12, i32 13>
1533 %bitcast64 = bitcast <16 x i16> %shuffle16 to <4 x double>
1534 ret <4 x double> %bitcast64
1535 }
1537 define <4 x i64> @concat_v4i64_0167(<4 x i64> %a0, <4 x i64> %a1) {
1538 ; ALL-LABEL: concat_v4i64_0167:
1539 ; ALL: # %bb.0:
1540 ; ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
1541 ; ALL-NEXT: retq
1542 %a0lo = shufflevector <4 x i64> %a0, <4 x i64> %a1, <2 x i32> <i32 0, i32 1>
1543 %a1hi = shufflevector <4 x i64> %a0, <4 x i64> %a1, <2 x i32> <i32 6, i32 7>
1544 %shuffle64 = shufflevector <2 x i64> %a0lo, <2 x i64> %a1hi, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
1545 ret <4 x i64> %shuffle64
1546 }
1548 define <4 x i64> @concat_v4i64_0145_bc(<4 x i64> %a0, <4 x i64> %a1) {
1549 ; ALL-LABEL: concat_v4i64_0145_bc:
1550 ; ALL: # %bb.0:
1551 ; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
1552 ; ALL-NEXT: retq
1553 %a0lo = shufflevector <4 x i64> %a0, <4 x i64> %a1, <2 x i32> <i32 0, i32 1>
1554 %a1lo = shufflevector <4 x i64> %a0, <4 x i64> %a1, <2 x i32> <i32 4, i32 5>
1555 %bc0lo = bitcast <2 x i64> %a0lo to <4 x i32>
1556 %bc1lo = bitcast <2 x i64> %a1lo to <4 x i32>
1557 %shuffle32 = shufflevector <4 x i32> %bc0lo, <4 x i32> %bc1lo, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
1558 %shuffle64 = bitcast <8 x i32> %shuffle32 to <4 x i64>
1559 ret <4 x i64> %shuffle64
1560 }
1562 define <4 x i64> @insert_dup_mem_v4i64(i64* %ptr) {
1563 ; ALL-LABEL: insert_dup_mem_v4i64:
1564 ; ALL: # %bb.0:
1565 ; ALL-NEXT: vbroadcastsd (%rdi), %ymm0
1566 ; ALL-NEXT: retq
1567 %tmp = load i64, i64* %ptr, align 1
1568 %tmp1 = insertelement <2 x i64> undef, i64 %tmp, i32 0
1569 %tmp2 = shufflevector <2 x i64> %tmp1, <2 x i64> undef, <4 x i32> zeroinitializer
1570 ret <4 x i64> %tmp2
1571 }
1573 define <4 x i64> @shuffle_v4i64_1234(<4 x i64> %a, <4 x i64> %b) {
1574 ; AVX1-LABEL: shuffle_v4i64_1234:
1575 ; AVX1: # %bb.0:
1576 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1]
1577 ; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[3],ymm1[2]
1578 ; AVX1-NEXT: retq
1579 ;
1580 ; AVX2-LABEL: shuffle_v4i64_1234:
1581 ; AVX2: # %bb.0:
1582 ; AVX2-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1]
1583 ; AVX2-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm0[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
1584 ; AVX2-NEXT: retq
1585 ;
1586 ; AVX512VL-LABEL: shuffle_v4i64_1234:
1587 ; AVX512VL: # %bb.0:
1588 ; AVX512VL-NEXT: valignq {{.*#+}} ymm0 = ymm0[1,2,3],ymm1[0]
1589 ; AVX512VL-NEXT: retq
1590 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 2, i32 3, i32 4>
1591 ret <4 x i64> %shuffle
1592 }
1594 define <4 x i64> @shuffle_v4i64_1230(<4 x i64> %a) {
1595 ; AVX1-LABEL: shuffle_v4i64_1230:
1596 ; AVX1: # %bb.0:
1597 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
1598 ; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[3],ymm1[2]
1599 ; AVX1-NEXT: retq
1600 ;
1601 ; AVX2-LABEL: shuffle_v4i64_1230:
1602 ; AVX2: # %bb.0:
1603 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,2,3,0]
1604 ; AVX2-NEXT: retq
1605 ;
1606 ; AVX512VL-LABEL: shuffle_v4i64_1230:
1607 ; AVX512VL: # %bb.0:
1608 ; AVX512VL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,2,3,0]
1609 ; AVX512VL-NEXT: retq
1610 %shuffle = shufflevector <4 x i64> %a, <4 x i64> undef, <4 x i32> <i32 1, i32 2, i32 3, i32 0>
1611 ret <4 x i64> %shuffle
1612 }
1614 define <4 x i64> @shuffle_v4i64_z0z3(<4 x i64> %a, <4 x i64> %b) {
1615 ; AVX1-LABEL: shuffle_v4i64_z0z3:
1616 ; AVX1: # %bb.0:
1617 ; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
1618 ; AVX1-NEXT: vxorpd %xmm1, %xmm1, %xmm1
1619 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
1620 ; AVX1-NEXT: retq
1621 ;
1622 ; AVX2-SLOW-LABEL: shuffle_v4i64_z0z3:
1623 ; AVX2-SLOW: # %bb.0:
1624 ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
1625 ; AVX2-SLOW-NEXT: vxorps %xmm1, %xmm1, %xmm1
1626 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
1627 ; AVX2-SLOW-NEXT: retq
1628 ;
1629 ; AVX2-FAST-LABEL: shuffle_v4i64_z0z3:
1630 ; AVX2-FAST: # %bb.0:
1631 ; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,2,3,4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[24,25,26,27,28,29,30,31]
1632 ; AVX2-FAST-NEXT: retq
1633 ;
1634 ; AVX512VL-SLOW-LABEL: shuffle_v4i64_z0z3:
1635 ; AVX512VL-SLOW: # %bb.0:
1636 ; AVX512VL-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
1637 ; AVX512VL-SLOW-NEXT: vxorps %xmm1, %xmm1, %xmm1
1638 ; AVX512VL-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
1639 ; AVX512VL-SLOW-NEXT: retq
1640 ;
1641 ; AVX512VL-FAST-LABEL: shuffle_v4i64_z0z3:
1642 ; AVX512VL-FAST: # %bb.0:
1643 ; AVX512VL-FAST-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,2,3,4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[24,25,26,27,28,29,30,31]
1644 ; AVX512VL-FAST-NEXT: retq
1645 %1 = shufflevector <4 x i64> %a, <4 x i64> <i64 0, i64 undef, i64 undef, i64 undef>, <4 x i32> <i32 4, i32 0, i32 4, i32 3>
1646 ret <4 x i64> %1
1647 }
1649 define <4 x i64> @shuffle_v4i64_1z2z(<4 x i64> %a, <4 x i64> %b) {
1650 ; AVX1-LABEL: shuffle_v4i64_1z2z:
1651 ; AVX1: # %bb.0:
1652 ; AVX1-NEXT: vxorpd %xmm1, %xmm1, %xmm1
1653 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
1654 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
1655 ; AVX1-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm0[0]
1656 ; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
1657 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
1658 ; AVX1-NEXT: retq
1659 ;
1660 ; AVX2-SLOW-LABEL: shuffle_v4i64_1z2z:
1661 ; AVX2-SLOW: # %bb.0:
1662 ; AVX2-SLOW-NEXT: vxorps %xmm1, %xmm1, %xmm1
1663 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
1664 ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,0,2,0]
1665 ; AVX2-SLOW-NEXT: retq
1666 ;
1667 ; AVX2-FAST-LABEL: shuffle_v4i64_1z2z:
1668 ; AVX2-FAST: # %bb.0:
1669 ; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19,20,21,22,23],zero,zero,zero,zero,zero,zero,zero,zero
1670 ; AVX2-FAST-NEXT: retq
1671 ;
1672 ; AVX512VL-SLOW-LABEL: shuffle_v4i64_1z2z:
1673 ; AVX512VL-SLOW: # %bb.0:
1674 ; AVX512VL-SLOW-NEXT: vxorps %xmm1, %xmm1, %xmm1
1675 ; AVX512VL-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
1676 ; AVX512VL-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,0,2,0]
1677 ; AVX512VL-SLOW-NEXT: retq
1678 ;
1679 ; AVX512VL-FAST-LABEL: shuffle_v4i64_1z2z:
1680 ; AVX512VL-FAST: # %bb.0:
1681 ; AVX512VL-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19,20,21,22,23],zero,zero,zero,zero,zero,zero,zero,zero
1682 ; AVX512VL-FAST-NEXT: retq
1683 %1 = shufflevector <4 x i64> %a, <4 x i64> <i64 0, i64 undef, i64 undef, i64 undef>, <4 x i32> <i32 1, i32 4, i32 2, i32 4>
1684 ret <4 x i64> %1
1685 }
1687 define <4 x double> @add_v4f64_0246_1357(<4 x double> %a, <4 x double> %b) {
1688 ; AVX1-LABEL: add_v4f64_0246_1357:
1689 ; AVX1: # %bb.0: # %entry
1690 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
1691 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
1692 ; AVX1-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
1693 ; AVX1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
1694 ; AVX1-NEXT: vaddpd %ymm0, %ymm1, %ymm0
1695 ; AVX1-NEXT: retq
1696 ;
1697 ; AVX2-LABEL: add_v4f64_0246_1357:
1698 ; AVX2: # %bb.0: # %entry
1699 ; AVX2-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
1700 ; AVX2-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
1701 ; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
1702 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
1703 ; AVX2-NEXT: vaddpd %ymm0, %ymm2, %ymm0
1704 ; AVX2-NEXT: retq
1705 ;
1706 ; AVX512VL-SLOW-LABEL: add_v4f64_0246_1357:
1707 ; AVX512VL-SLOW: # %bb.0: # %entry
1708 ; AVX512VL-SLOW-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
1709 ; AVX512VL-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
1710 ; AVX512VL-SLOW-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
1711 ; AVX512VL-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
1712 ; AVX512VL-SLOW-NEXT: vaddpd %ymm0, %ymm2, %ymm0
1713 ; AVX512VL-SLOW-NEXT: retq
1714 ;
1715 ; AVX512VL-FAST-LABEL: add_v4f64_0246_1357:
1716 ; AVX512VL-FAST: # %bb.0: # %entry
1717 ; AVX512VL-FAST-NEXT: vmovapd {{.*#+}} ymm2 = [0,2,4,6]
1718 ; AVX512VL-FAST-NEXT: vpermi2pd %ymm1, %ymm0, %ymm2
1719 ; AVX512VL-FAST-NEXT: vmovapd {{.*#+}} ymm3 = [1,3,5,7]
1720 ; AVX512VL-FAST-NEXT: vpermi2pd %ymm1, %ymm0, %ymm3
1721 ; AVX512VL-FAST-NEXT: vaddpd %ymm3, %ymm2, %ymm0
1722 ; AVX512VL-FAST-NEXT: retq
1723 entry:
1724 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
1725 %shuffle1 = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
1726 %add = fadd <4 x double> %shuffle, %shuffle1
1727 ret <4 x double> %add
1728 }
1730 define <4 x double> @add_v4f64_4602_5713(<4 x double> %a, <4 x double> %b) {
1731 ; AVX1-LABEL: add_v4f64_4602_5713:
1732 ; AVX1: # %bb.0: # %entry
1733 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm0[2,3]
1734 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
1735 ; AVX1-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
1736 ; AVX1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
1737 ; AVX1-NEXT: vaddpd %ymm0, %ymm1, %ymm0
1738 ; AVX1-NEXT: retq
1739 ;
1740 ; AVX2-LABEL: add_v4f64_4602_5713:
1741 ; AVX2: # %bb.0: # %entry
1742 ; AVX2-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
1743 ; AVX2-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
1744 ; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
1745 ; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
1746 ; AVX2-NEXT: vaddpd %ymm0, %ymm2, %ymm0
1747 ; AVX2-NEXT: retq
1748 ;
1749 ; AVX512VL-SLOW-LABEL: add_v4f64_4602_5713:
1750 ; AVX512VL-SLOW: # %bb.0: # %entry
1751 ; AVX512VL-SLOW-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
1752 ; AVX512VL-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
1753 ; AVX512VL-SLOW-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
1754 ; AVX512VL-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
1755 ; AVX512VL-SLOW-NEXT: vaddpd %ymm0, %ymm2, %ymm0
1756 ; AVX512VL-SLOW-NEXT: retq
1757 ;
1758 ; AVX512VL-FAST-LABEL: add_v4f64_4602_5713:
1759 ; AVX512VL-FAST: # %bb.0: # %entry
1760 ; AVX512VL-FAST-NEXT: vmovapd {{.*#+}} ymm2 = [0,2,4,6]
1761 ; AVX512VL-FAST-NEXT: vpermi2pd %ymm0, %ymm1, %ymm2
1762 ; AVX512VL-FAST-NEXT: vmovapd {{.*#+}} ymm3 = [1,3,5,7]
1763 ; AVX512VL-FAST-NEXT: vpermi2pd %ymm0, %ymm1, %ymm3
1764 ; AVX512VL-FAST-NEXT: vaddpd %ymm3, %ymm2, %ymm0
1765 ; AVX512VL-FAST-NEXT: retq
1766 entry:
1767 %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 4, i32 6, i32 0, i32 2>
1768 %shuffle1 = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 5, i32 7, i32 1, i32 3>
1769 %add = fadd <4 x double> %shuffle, %shuffle1
1770 ret <4 x double> %add
1771 }
1773 define <4 x i64> @add_v4i64_0246_1357(<4 x i64> %a, <4 x i64> %b) {
1774 ; AVX1-LABEL: add_v4i64_0246_1357:
1775 ; AVX1: # %bb.0: # %entry
1776 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
1777 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
1778 ; AVX1-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
1779 ; AVX1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
1780 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
1781 ; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3
1782 ; AVX1-NEXT: vpaddq %xmm2, %xmm3, %xmm2
1783 ; AVX1-NEXT: vpaddq %xmm0, %xmm1, %xmm0
1784 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1785 ; AVX1-NEXT: retq
1786 ;
1787 ; AVX2-LABEL: add_v4i64_0246_1357:
1788 ; AVX2: # %bb.0: # %entry
1789 ; AVX2-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
1790 ; AVX2-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,2,1,3]
1791 ; AVX2-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
1792 ; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
1793 ; AVX2-NEXT: vpaddq %ymm0, %ymm2, %ymm0
1794 ; AVX2-NEXT: retq
1795 ;
1796 ; AVX512VL-SLOW-LABEL: add_v4i64_0246_1357:
1797 ; AVX512VL-SLOW: # %bb.0: # %entry
1798 ; AVX512VL-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
1799 ; AVX512VL-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,2,1,3]
1800 ; AVX512VL-SLOW-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
1801 ; AVX512VL-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
1802 ; AVX512VL-SLOW-NEXT: vpaddq %ymm0, %ymm2, %ymm0
1803 ; AVX512VL-SLOW-NEXT: retq
1804 ;
1805 ; AVX512VL-FAST-LABEL: add_v4i64_0246_1357:
1806 ; AVX512VL-FAST: # %bb.0: # %entry
1807 ; AVX512VL-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [0,2,4,6]
1808 ; AVX512VL-FAST-NEXT: vpermi2q %ymm1, %ymm0, %ymm2
1809 ; AVX512VL-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [1,3,5,7]
1810 ; AVX512VL-FAST-NEXT: vpermi2q %ymm1, %ymm0, %ymm3
1811 ; AVX512VL-FAST-NEXT: vpaddq %ymm3, %ymm2, %ymm0
1812 ; AVX512VL-FAST-NEXT: retq
1813 entry:
1814 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
1815 %shuffle1 = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
1816 %add = add <4 x i64> %shuffle, %shuffle1
1817 ret <4 x i64> %add
1818 }
1820 define <4 x i64> @add_v4i64_4602_5713(<4 x i64> %a, <4 x i64> %b) {
1821 ; AVX1-LABEL: add_v4i64_4602_5713:
1822 ; AVX1: # %bb.0: # %entry
1823 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm0[2,3]
1824 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
1825 ; AVX1-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
1826 ; AVX1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
1827 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
1828 ; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3
1829 ; AVX1-NEXT: vpaddq %xmm2, %xmm3, %xmm2
1830 ; AVX1-NEXT: vpaddq %xmm0, %xmm1, %xmm0
1831 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1832 ; AVX1-NEXT: retq
1833 ;
1834 ; AVX2-LABEL: add_v4i64_4602_5713:
1835 ; AVX2: # %bb.0: # %entry
1836 ; AVX2-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
1837 ; AVX2-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,2,1,3]
1838 ; AVX2-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
1839 ; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
1840 ; AVX2-NEXT: vpaddq %ymm0, %ymm2, %ymm0
1841 ; AVX2-NEXT: retq
1842 ;
1843 ; AVX512VL-SLOW-LABEL: add_v4i64_4602_5713:
1844 ; AVX512VL-SLOW: # %bb.0: # %entry
1845 ; AVX512VL-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
1846 ; AVX512VL-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,2,1,3]
1847 ; AVX512VL-SLOW-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
1848 ; AVX512VL-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
1849 ; AVX512VL-SLOW-NEXT: vpaddq %ymm0, %ymm2, %ymm0
1850 ; AVX512VL-SLOW-NEXT: retq
1851 ;
1852 ; AVX512VL-FAST-LABEL: add_v4i64_4602_5713:
1853 ; AVX512VL-FAST: # %bb.0: # %entry
1854 ; AVX512VL-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [0,2,4,6]
1855 ; AVX512VL-FAST-NEXT: vpermi2q %ymm0, %ymm1, %ymm2
1856 ; AVX512VL-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [1,3,5,7]
1857 ; AVX512VL-FAST-NEXT: vpermi2q %ymm0, %ymm1, %ymm3
1858 ; AVX512VL-FAST-NEXT: vpaddq %ymm3, %ymm2, %ymm0
1859 ; AVX512VL-FAST-NEXT: retq
1860 entry:
1861 %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 4, i32 6, i32 0, i32 2>
1862 %shuffle1 = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 5, i32 7, i32 1, i32 3>
1863 %add = add <4 x i64> %shuffle, %shuffle1
1864 ret <4 x i64> %add
1865 }
1867 define <4 x double> @shuffle_v4f64_0zzz_optsize(<4 x double> %a) optsize {
1868 ; ALL-LABEL: shuffle_v4f64_0zzz_optsize:
1869 ; ALL: # %bb.0:
1870 ; ALL-NEXT: vxorpd %xmm1, %xmm1, %xmm1
1871 ; ALL-NEXT: vmovsd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
1872 ; ALL-NEXT: retq
1873 %b = shufflevector <4 x double> %a, <4 x double> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
1874 ret <4 x double> %b
1875 }
1877 define <4 x i64> @shuffle_v4i64_0zzz_optsize(<4 x i64> %a) optsize {
1878 ; ALL-LABEL: shuffle_v4i64_0zzz_optsize:
1879 ; ALL: # %bb.0:
1880 ; ALL-NEXT: vxorpd %xmm1, %xmm1, %xmm1
1881 ; ALL-NEXT: vmovsd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
1882 ; ALL-NEXT: retq
1883 %b = shufflevector <4 x i64> %a, <4 x i64> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
1884 ret <4 x i64> %b
1885 }
1887 define <8 x float> @shuffle_v8f32_0zzzzzzz_optsize(<8 x float> %a) optsize {
1888 ; ALL-LABEL: shuffle_v8f32_0zzzzzzz_optsize:
1889 ; ALL: # %bb.0:
1890 ; ALL-NEXT: vxorps %xmm1, %xmm1, %xmm1
1891 ; ALL-NEXT: vmovss {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
1892 ; ALL-NEXT: retq
1893 %b = shufflevector <8 x float> %a, <8 x float> zeroinitializer, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
1894 ret <8 x float> %b
1895 }
1897 define <8 x i32> @shuffle_v8i32_0zzzzzzz_optsize(<8 x i32> %a) optsize {
1898 ; ALL-LABEL: shuffle_v8i32_0zzzzzzz_optsize:
1899 ; ALL: # %bb.0:
1900 ; ALL-NEXT: vxorps %xmm1, %xmm1, %xmm1
1901 ; ALL-NEXT: vmovss {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
1902 ; ALL-NEXT: retq
1903 %b = shufflevector <8 x i32> %a, <8 x i32> zeroinitializer, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
1904 ret <8 x i32> %b
1905 }
1907 define <4 x i64> @unpckh_v4i64(<4 x i64> %x, <4 x i64> %y) {
1908 ; ALL-LABEL: unpckh_v4i64:
1909 ; ALL: # %bb.0:
1910 ; ALL-NEXT: vextractf128 $1, %ymm1, %xmm1
1911 ; ALL-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
1912 ; ALL-NEXT: retq
1913 %unpckh = shufflevector <4 x i64> %x, <4 x i64> %y, <4 x i32> <i32 1, i32 7, i32 undef, i32 undef>
1914 ret <4 x i64> %unpckh
1915 }
1917 define <4 x double> @unpckh_v4f64(<4 x double> %x, <4 x double> %y) {
1918 ; ALL-LABEL: unpckh_v4f64:
1919 ; ALL: # %bb.0:
1920 ; ALL-NEXT: vextractf128 $1, %ymm1, %xmm1
1921 ; ALL-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
1922 ; ALL-NEXT: retq
1923 %unpckh = shufflevector <4 x double> %x, <4 x double> %y, <4 x i32> <i32 1, i32 7, i32 undef, i32 undef>
1924 ret <4 x double> %unpckh
1925 }