Home | History | Annotate | Download | only in X86

Lines Matching full:ymm0

11 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
16 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
27 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
33 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
44 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
50 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
61 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
67 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
76 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
78 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,3,4,4,4,7]
79 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
85 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
94 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
95 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
96 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,4,4]
102 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
111 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
112 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
113 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,2,0,0,4,4,4,4]
119 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
128 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
129 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
130 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,0,0,0,4,4,4,4]
137 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
146 ; ALL-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
157 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
163 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
174 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
180 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
189 ; ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
201 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
202 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
208 ; AVX2-NEXT: vbroadcastsd %xmm0, %ymm0
209 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
218 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
219 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
228 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,0],ymm0[2,3],ymm1[4,4],ymm0[6,7]
237 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[3,2],ymm1[5,4],ymm0[7,6]
246 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
255 ; ALL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
264 ; ALL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
275 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
283 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
284 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
297 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
303 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
306 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
317 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
318 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
324 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
325 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
335 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
336 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
342 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
343 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
352 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
361 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
370 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
379 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
388 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
397 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
406 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
415 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
424 ; ALL-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
433 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
442 ; ALL-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
451 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
460 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
469 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,5,4,4,4]
478 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,6,4,4]
487 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,4,7,4]
496 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,4,4,4,4]
505 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,4,4,6]
514 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,4,7,4]
523 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7]
532 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,6,7,4,4]
541 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,2,6,6,4,4]
550 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,6,7]
559 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7]
568 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,7]
577 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,6]
586 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,u,6,u,4,4]
595 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,u,u,6,6,u,u]
604 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,u,u]
613 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,u,u,6,7]
622 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,u,u,3,5,4,u,u]
631 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[u,u,u,3,u,u,6,6]
640 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
641 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm2[0,0],ymm0[4,7],ymm2[4,4]
645 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
651 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
654 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
663 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
665 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,1,1,4,5,5,5]
666 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
669 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
677 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
678 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
688 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
694 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
703 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
705 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
711 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
720 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
721 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
727 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
736 ; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
737 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
746 ; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
747 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
756 ; ALL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
757 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
766 ; ALL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
767 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
778 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
779 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
781 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
789 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
790 ; AVX2-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
799 ; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
800 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
809 ; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
810 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
819 ; ALL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
828 ; ALL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
838 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
844 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
853 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
855 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
861 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
870 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
877 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
887 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
892 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
903 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
909 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
920 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
926 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
937 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
943 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
952 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
954 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,3,4,4,4,7]
955 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
961 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
970 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
971 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
972 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,4,4]
978 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
987 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
988 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
989 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,2,0,0,4,4,4,4]
995 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1004 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
1005 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
1006 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,0,0,0,4,4,4,4]
1013 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1022 ; AVX1-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
1027 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
1038 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
1044 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1055 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
1061 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1070 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
1075 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
1087 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1088 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1094 ; AVX2-NEXT: vpbroadcastq %xmm0, %ymm0
1095 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1104 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
1105 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
1111 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
1112 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1121 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,0],ymm0[2,3],ymm1[4,4],ymm0[6,7]
1127 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
1136 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[3,2],ymm1[5,4],ymm0[7,6]
1141 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
1142 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
1151 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
1156 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,1,0,4,5,5,4]
1158 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
1167 ; AVX1-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
1172 ; AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
1181 ; AVX1-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
1186 ; AVX2-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
1197 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1204 ; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
1205 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1218 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
1224 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
1227 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1238 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
1239 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1244 ; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
1245 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1255 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1256 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1262 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
1263 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1272 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
1277 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
1286 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
1291 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
1300 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
1305 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
1314 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
1319 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
1328 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
1333 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
1342 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
1347 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
1356 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1361 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1370 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
1375 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
1384 ; AVX1-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
1389 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
1398 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
1403 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
1412 ; AVX1-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
1417 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
1426 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
1431 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
1440 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
1445 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
1454 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,5,4,4,4]
1460 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1469 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,6,4,4]
1475 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1484 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,4,7,4]
1490 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1499 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,4,4,4,4]
1505 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1514 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,4,4,6]
1520 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1529 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,4,7,4]
1535 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1544 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7]
1550 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1559 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,6,7,4,4]
1565 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1574 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,2,6,6,4,4]
1580 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1589 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,6,7]
1595 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1604 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7]
1610 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1619 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,7]
1625 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1634 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,6]
1640 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1649 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,u,6,u,4,4]
1655 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1664 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,u,u,6,6,u,u]
1670 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1679 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,u,u]
1685 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1694 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,u,u,6,7]
1700 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1709 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,u,u,3,5,4,u,u]
1715 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1724 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[u,u,u,3,u,u,6,6]
1730 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1741 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
1743 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1744 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
1751 ; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,1,3,2]
1752 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
1762 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1768 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1777 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
1779 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1785 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1794 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
1795 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1801 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1810 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
1811 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1816 ; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
1817 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1826 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
1827 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1832 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
1833 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1842 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
1843 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1848 ; AVX2-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
1849 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1858 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
1859 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1864 ; AVX2-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
1865 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1874 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
1875 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1880 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
1881 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1890 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
1891 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1896 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
1897 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1907 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
1912 ; AVX2-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,2,3],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19]
1922 ; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
1923 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,2],ymm1[2,0],ymm0[5,6],ymm1[6,4]
1928 ; AVX2-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,ymm0[20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero
1937 ; AVX1-NEXT: vunpcklps {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
1942 ; AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
1952 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1958 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1967 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
1969 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1975 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1984 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
1991 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
2000 ; ALL-NEXT: vbroadcastss (%rdi), %ymm0
2012 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
2017 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
2031 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2],ymm1[0,2],ymm0[4,6],ymm1[4,6]
2032 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0,3,1,6,4,7,5]
2037 ; AVX2-NEXT: vpsllq $32, %ymm0, %ymm0
2047 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm1[1,3],ymm0[5,7],ymm1[5,7]
2048 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
2053 ; AVX2-NEXT: vpsrlq $32, %ymm0, %ymm0
2062 ; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
2063 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,2],ymm0[1,2],ymm1[4,6],ymm0[5,6]
2068 ; AVX2-NEXT: vpalignr {{.*#+}} ymm0 = ymm1[12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10,11],ymm1[28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26,27]
2077 ; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm0[3,0],ymm1[4,4],ymm0[7,4]
2078 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,2],ymm1[2,0],ymm0[5,6],ymm1[6,4]
2083 ; AVX2-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0,1,2,3],ymm0[20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16,17,18,19]
2092 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[3,0],ymm0[4,4],ymm1[7,4]
2093 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,2],ymm0[2,0],ymm1[5,6],ymm0[6,4]
2098 ; AVX2-NEXT: vpalignr {{.*#+}} ymm0 = ymm1[4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1,2,3],ymm1[20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17,18,19]
2107 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
2108 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2],ymm1[1,2],ymm0[4,6],ymm1[5,6]
2113 ; AVX2-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
2122 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,0,1,2,7,4,5,6]
2127 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,0,1,2,7,4,5,6]
2136 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,2,3,0,5,6,7,4]
2141 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,2,3,0,5,6,7,4]
2203 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
2208 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
2219 ; ALL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
2233 ; ALL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
2249 ; ALL-NEXT: vbroadcastss (%rdi), %ymm0