Lines Matching full:ymm0
10 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
15 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
26 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
32 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
43 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
49 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
60 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
66 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
75 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
77 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,3,4,4,4,7]
78 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
84 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
93 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
94 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
95 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,4,4]
101 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
110 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
111 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
112 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,2,0,0,4,4,4,4]
118 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
127 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
128 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
129 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,0,0,0,4,4,4,4]
136 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
145 ; ALL-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
156 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
162 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
173 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
179 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
188 ; ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
200 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
201 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
207 ; AVX2-NEXT: vbroadcastsd %xmm0, %ymm0
208 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
217 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
218 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
227 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,0],ymm0[2,3],ymm1[4,4],ymm0[6,7]
236 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[3,2],ymm1[5,4],ymm0[7,6]
245 ; ALL-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
254 ; ALL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
263 ; ALL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
274 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
282 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
283 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
296 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
302 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
305 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
316 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
317 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
323 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
324 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
334 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
335 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
341 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
342 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
351 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
360 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
369 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
378 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
387 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
396 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
405 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
414 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
423 ; ALL-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
432 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
441 ; ALL-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
450 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
459 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
468 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,5,4,4,4]
477 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,6,4,4]
486 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,4,7,4]
495 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,4,4,4,4]
504 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,4,4,6]
513 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,4,7,4]
522 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7]
531 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,6,7,4,4]
540 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,2,6,6,4,4]
549 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,6,7]
558 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7]
567 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,7]
576 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,6]
585 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,u,6,u,4,4]
594 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,u,u,6,6,u,u]
603 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,u,u]
612 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,u,u,6,7]
621 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,u,u,3,5,4,u,u]
630 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[u,u,u,3,u,u,6,6]
639 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
640 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm2[0,0],ymm0[4,7],ymm2[4,4]
644 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
650 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
653 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
662 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
664 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,1,1,4,5,5,5]
665 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
668 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
676 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
677 ; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
687 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
693 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
702 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
704 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
710 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
719 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
720 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
726 ; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
735 ; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
736 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
745 ; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
746 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
755 ; ALL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
756 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
765 ; ALL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
766 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
777 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
778 ymm0, %xmm2
780 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
788 ; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm0
789 ; AVX2-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
798 ; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
799 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
808 ; ALL-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
809 ; ALL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
818 ; ALL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
827 ; ALL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
837 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
842 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
853 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
859 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
870 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
876 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
887 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
893 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
902 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
904 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,3,4,4,4,7]
905 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
911 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
920 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
921 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
922 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,4,4]
928 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
937 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
938 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
939 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,2,0,0,4,4,4,4]
945 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
954 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
955 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
956 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,0,0,0,4,4,4,4]
963 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
972 ; AVX1-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
977 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
988 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
994 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1005 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
1011 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1020 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
1025 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
1037 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1038 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1044 ; AVX2-NEXT: vpbroadcastq %xmm0, %ymm0
1045 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1054 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
1055 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
1061 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
1062 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1071 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,0],ymm0[2,3],ymm1[4,4],ymm0[6,7]
1077 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
1086 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[3,2],ymm1[5,4],ymm0[7,6]
1091 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
1092 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
1101 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
1106 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,1,0,4,5,5,4]
1108 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
1117 ; AVX1-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
1122 ; AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
1131 ; AVX1-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
1136 ; AVX2-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
1147 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1154 ; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
1155 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1168 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
1174 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
1177 ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1188 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
1189 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1194 ; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
1195 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
1205 ; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
1206 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1212 ; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
1213 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
1222 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
1227 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
1236 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
1241 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
1250 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
1255 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
1264 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
1269 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
1278 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
1283 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
1292 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
1297 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
1306 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1311 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1320 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
1325 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
1334 ; AVX1-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
1339 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
1348 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
1353 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
1362 ; AVX1-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
1367 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
1376 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
1381 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
1390 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
1395 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
1404 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,5,4,4,4]
1410 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1419 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,6,4,4]
1425 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1434 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,4,7,4]
1440 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1449 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,4,4,4,4]
1455 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1464 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,4,4,6]
1470 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1479 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,4,7,4]
1485 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1494 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7]
1500 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1509 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,6,7,4,4]
1515 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1524 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,2,6,6,4,4]
1530 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1539 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,6,7]
1545 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1554 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7]
1560 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1569 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,7]
1575 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1584 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,5,4,6,6]
1590 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1599 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,u,6,u,4,4]
1605 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1614 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0ymm0[0,0,u,u,6,6,u,u]
1620 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1629 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,u,u]
1635 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1644 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,u,u,6,7]
1650 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1659 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,u,u,3,5,4,u,u]
1665 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1674 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[u,u,u,3,u,u,6,6]
1680 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1691 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
1693 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1694 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
1701 ; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,1,3,2]
1702 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
1712 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1718 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1727 ; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
1729 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1735 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1744 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
1745 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1751 ; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
1760 ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
1761 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1766 ; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
1767 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1776 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
1777 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1782 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
1783 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1792 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
1793 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1798 ; AVX2-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
1799 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1808 ; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
1809 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1814 ; AVX2-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
1815 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1824 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
1825 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1830 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
1831 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1840 ; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
1841 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1846 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
1847 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
1857 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
1862 ; 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]
1872 ; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
1873 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,2],ymm1[2,0],ymm0[5,6],ymm1[6,4]
1878 ; 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
1887 ; AVX1-NEXT: vunpcklps {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
1892 ; AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
1901 ; ALL-NEXT: vbroadcastss (%rdi), %ymm0
1913 ; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
1918 ; AVX2-NEXT: vbroadcastss %xmm0, %ymm0
1932 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2],ymm1[0,2],ymm0[4,6],ymm1[4,6]
1933 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0,3,1,6,4,7,5]
1938 ; AVX2-NEXT: vpsllq $32, %ymm0, %ymm0
1948 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm1[1,3],ymm0[5,7],ymm1[5,7]
1949 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
1954 ; AVX2-NEXT: vpsrlq $32, %ymm0, %ymm0
1963 ; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
1964 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,2],ymm0[1,2],ymm1[4,6],ymm0[5,6]
1969 ; 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]
1978 ; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm0[3,0],ymm1[4,4],ymm0[7,4]
1979 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,2],ymm1[2,0],ymm0[5,6],ymm1[6,4]
1984 ; 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]
1993 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[3,0],ymm0[4,4],ymm1[7,4]
1994 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,2],ymm0[2,0],ymm1[5,6],ymm0[6,4]
1999 ; 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]
2008 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
2009 ; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2],ymm1[1,2],ymm0[4,6],ymm1[5,6]
2014 ; 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]
2023 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,0,1,2,7,4,5,6]
2028 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,0,1,2,7,4,5,6]
2037 ; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[1,2,3,0,5,6,7,4]
2042 ; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,2,3,0,5,6,7,4]
