llvm.org GIT mirror llvm / 1ce51ae
[X86][AVX1] Account for cost of extract/insert of 256-bit shifts by splat git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@303021 91177308-0d34-0410-b5e6-96231b3b80d8 Simon Pilgrim 3 years ago
4 changed file(s) with 50 addition(s) and 50 deletion(s). Raw diff Collapse all Expand all
453453
454454 static const CostTblEntry SSE2UniformShiftCostTable[] = {
455455 // Uniform splats are cheaper for the following instructions.
456 { ISD::SHL, MVT::v16i16, 2 }, // psllw.
457 { ISD::SHL, MVT::v8i32, 2 }, // pslld
458 { ISD::SHL, MVT::v4i64, 2 }, // psllq.
459
460 { ISD::SRL, MVT::v16i16, 2 }, // psrlw.
461 { ISD::SRL, MVT::v8i32, 2 }, // psrld.
462 { ISD::SRL, MVT::v4i64, 2 }, // psrlq.
463
464 { ISD::SRA, MVT::v16i16, 2 }, // psraw.
465 { ISD::SRA, MVT::v8i32, 2 }, // psrad.
466 { ISD::SRA, MVT::v2i64, 4 }, // 2 x psrad + shuffle.
467 { ISD::SRA, MVT::v4i64, 8 }, // 2 x psrad + shuffle.
456 { ISD::SHL, MVT::v16i16, 2+2 }, // 2*psllw + split.
457 { ISD::SHL, MVT::v8i32, 2+2 }, // 2*pslld + split.
458 { ISD::SHL, MVT::v4i64, 2+2 }, // 2*psllq + split.
459
460 { ISD::SRL, MVT::v16i16, 2+2 }, // 2*psrlw + split.
461 { ISD::SRL, MVT::v8i32, 2+2 }, // 2*psrld + split.
462 { ISD::SRL, MVT::v4i64, 2+2 }, // 2*psrlq + split.
463
464 { ISD::SRA, MVT::v16i16, 2+2 }, // 2*psraw + split.
465 { ISD::SRA, MVT::v8i32, 2+2 }, // 2*psrad + split.
466 { ISD::SRA, MVT::v2i64, 4 }, // 2*psrad + shuffle.
467 { ISD::SRA, MVT::v4i64, 8+2 }, // 2*(2*psrad + shuffle) + split.
468468 };
469469
470470 if (ST->hasSSE2() &&
190190 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v4i64':
191191 ; SSE2: Found an estimated cost of 8 for instruction: %shift
192192 ; SSE41: Found an estimated cost of 8 for instruction: %shift
193 ; AVX: Found an estimated cost of 8 for instruction: %shift
194 ; AVX2: Found an estimated cost of 8 for instruction: %shift
193 ; AVX: Found an estimated cost of 10 for instruction: %shift
194 ; AVX2: Found an estimated cost of 10 for instruction: %shift
195195 ; AVX512: Found an estimated cost of 1 for instruction: %shift
196196 ; XOP: Found an estimated cost of 6 for instruction: %shift
197197 %insert = insertelement <4 x i64> undef, i64 %b, i32 0
204204 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i64':
205205 ; SSE2: Found an estimated cost of 16 for instruction: %shift
206206 ; SSE41: Found an estimated cost of 16 for instruction: %shift
207 ; AVX: Found an estimated cost of 16 for instruction: %shift
208 ; AVX2: Found an estimated cost of 16 for instruction: %shift
207 ; AVX: Found an estimated cost of 20 for instruction: %shift
208 ; AVX2: Found an estimated cost of 20 for instruction: %shift
209209 ; AVX512: Found an estimated cost of 1 for instruction: %shift
210210 ; XOP: Found an estimated cost of 12 for instruction: %shift
211211 %insert = insertelement <8 x i64> undef, i64 %b, i32 0
232232 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i32':
233233 ; SSE2: Found an estimated cost of 2 for instruction: %shift
234234 ; SSE41: Found an estimated cost of 2 for instruction: %shift
235 ; AVX: Found an estimated cost of 2 for instruction: %shift
235 ; AVX: Found an estimated cost of 4 for instruction: %shift
236236 ; AVX2: Found an estimated cost of 1 for instruction: %shift
237237 ; AVX512: Found an estimated cost of 1 for instruction: %shift
238238 ; XOPAVX: Found an estimated cost of 6 for instruction: %shift
247247 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i32':
248248 ; SSE2: Found an estimated cost of 4 for instruction: %shift
249249 ; SSE41: Found an estimated cost of 4 for instruction: %shift
250 ; AVX: Found an estimated cost of 4 for instruction: %shift
250 ; AVX: Found an estimated cost of 8 for instruction: %shift
251251 ; AVX2: Found an estimated cost of 2 for instruction: %shift
252252 ; AVX512: Found an estimated cost of 1 for instruction: %shift
253253 ; XOPAVX: Found an estimated cost of 12 for instruction: %shift
276276 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i16':
277277 ; SSE2: Found an estimated cost of 2 for instruction: %shift
278278 ; SSE41: Found an estimated cost of 2 for instruction: %shift
279 ; AVX: Found an estimated cost of 2 for instruction: %shift
279 ; AVX: Found an estimated cost of 4 for instruction: %shift
280280 ; AVX2: Found an estimated cost of 1 for instruction: %shift
281281 ; AVX512: Found an estimated cost of 1 for instruction: %shift
282282 ; XOPAVX: Found an estimated cost of 6 for instruction: %shift
291291 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v32i16':
292292 ; SSE2: Found an estimated cost of 4 for instruction: %shift
293293 ; SSE41: Found an estimated cost of 4 for instruction: %shift
294 ; AVX: Found an estimated cost of 4 for instruction: %shift
294 ; AVX: Found an estimated cost of 8 for instruction: %shift
295295 ; AVX2: Found an estimated cost of 2 for instruction: %shift
296296 ; AVX512F: Found an estimated cost of 2 for instruction: %shift
297297 ; AVX512BW: Found an estimated cost of 1 for instruction: %shift
521521 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v4i64':
522522 ; SSE2: Found an estimated cost of 8 for instruction: %shift
523523 ; SSE41: Found an estimated cost of 8 for instruction: %shift
524 ; AVX: Found an estimated cost of 8 for instruction: %shift
524 ; AVX: Found an estimated cost of 10 for instruction: %shift
525525 ; AVX2: Found an estimated cost of 4 for instruction: %shift
526526 ; AVX512: Found an estimated cost of 1 for instruction: %shift
527527 ; XOPAVX: Found an estimated cost of 6 for instruction: %shift
534534 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i64':
535535 ; SSE2: Found an estimated cost of 16 for instruction: %shift
536536 ; SSE41: Found an estimated cost of 16 for instruction: %shift
537 ; AVX: Found an estimated cost of 16 for instruction: %shift
537 ; AVX: Found an estimated cost of 20 for instruction: %shift
538538 ; AVX2: Found an estimated cost of 8 for instruction: %shift
539539 ; AVX512: Found an estimated cost of 1 for instruction: %shift
540540 ; XOPAVX: Found an estimated cost of 12 for instruction: %shift
559559 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i32':
560560 ; SSE2: Found an estimated cost of 2 for instruction: %shift
561561 ; SSE41: Found an estimated cost of 2 for instruction: %shift
562 ; AVX: Found an estimated cost of 2 for instruction: %shift
562 ; AVX: Found an estimated cost of 4 for instruction: %shift
563563 ; AVX2: Found an estimated cost of 1 for instruction: %shift
564564 ; AVX512: Found an estimated cost of 1 for instruction: %shift
565565 ; XOPAVX: Found an estimated cost of 6 for instruction: %shift
572572 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i32':
573573 ; SSE2: Found an estimated cost of 4 for instruction: %shift
574574 ; SSE41: Found an estimated cost of 4 for instruction: %shift
575 ; AVX: Found an estimated cost of 4 for instruction: %shift
575 ; AVX: Found an estimated cost of 8 for instruction: %shift
576576 ; AVX2: Found an estimated cost of 2 for instruction: %shift
577577 ; AVX512: Found an estimated cost of 1 for instruction: %shift
578578 ; XOPAVX: Found an estimated cost of 12 for instruction: %shift
597597 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i16':
598598 ; SSE2: Found an estimated cost of 2 for instruction: %shift
599599 ; SSE41: Found an estimated cost of 2 for instruction: %shift
600 ; AVX: Found an estimated cost of 2 for instruction: %shift
600 ; AVX: Found an estimated cost of 4 for instruction: %shift
601601 ; AVX2: Found an estimated cost of 1 for instruction: %shift
602602 ; AVX512: Found an estimated cost of 1 for instruction: %shift
603603 ; XOPAVX: Found an estimated cost of 6 for instruction: %shift
610610 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i16':
611611 ; SSE2: Found an estimated cost of 4 for instruction: %shift
612612 ; SSE41: Found an estimated cost of 4 for instruction: %shift
613 ; AVX: Found an estimated cost of 4 for instruction: %shift
613 ; AVX: Found an estimated cost of 8 for instruction: %shift
614614 ; AVX2: Found an estimated cost of 2 for instruction: %shift
615615 ; AVX512F: Found an estimated cost of 2 for instruction: %shift
616616 ; AVX512BW: Found an estimated cost of 1 for instruction: %shift
192192 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v4i64':
193193 ; SSE2: Found an estimated cost of 2 for instruction: %shift
194194 ; SSE41: Found an estimated cost of 2 for instruction: %shift
195 ; AVX: Found an estimated cost of 2 for instruction: %shift
195 ; AVX: Found an estimated cost of 4 for instruction: %shift
196196 ; AVX2: Found an estimated cost of 1 for instruction: %shift
197197 ; AVX512: Found an estimated cost of 1 for instruction: %shift
198198 ; XOPAVX: Found an estimated cost of 6 for instruction: %shift
207207 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i64':
208208 ; SSE2: Found an estimated cost of 4 for instruction: %shift
209209 ; SSE41: Found an estimated cost of 4 for instruction: %shift
210 ; AVX: Found an estimated cost of 4 for instruction: %shift
210 ; AVX: Found an estimated cost of 8 for instruction: %shift
211211 ; AVX2: Found an estimated cost of 2 for instruction: %shift
212212 ; AVX512: Found an estimated cost of 1 for instruction: %shift
213213 ; XOPAVX: Found an estimated cost of 12 for instruction: %shift
236236 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i32':
237237 ; SSE2: Found an estimated cost of 2 for instruction: %shift
238238 ; SSE41: Found an estimated cost of 2 for instruction: %shift
239 ; AVX: Found an estimated cost of 2 for instruction: %shift
239 ; AVX: Found an estimated cost of 4 for instruction: %shift
240240 ; AVX2: Found an estimated cost of 1 for instruction: %shift
241241 ; AVX512: Found an estimated cost of 1 for instruction: %shift
242242 ; XOPAVX: Found an estimated cost of 6 for instruction: %shift
251251 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i32':
252252 ; SSE2: Found an estimated cost of 4 for instruction: %shift
253253 ; SSE41: Found an estimated cost of 4 for instruction: %shift
254 ; AVX: Found an estimated cost of 4 for instruction: %shift
254 ; AVX: Found an estimated cost of 8 for instruction: %shift
255255 ; AVX2: Found an estimated cost of 2 for instruction: %shift
256256 ; AVX512: Found an estimated cost of 1 for instruction: %shift
257257 ; XOPAVX: Found an estimated cost of 12 for instruction: %shift
280280 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i16':
281281 ; SSE2: Found an estimated cost of 2 for instruction: %shift
282282 ; SSE41: Found an estimated cost of 2 for instruction: %shift
283 ; AVX: Found an estimated cost of 2 for instruction: %shift
283 ; AVX: Found an estimated cost of 4 for instruction: %shift
284284 ; AVX2: Found an estimated cost of 1 for instruction: %shift
285285 ; AVX512: Found an estimated cost of 1 for instruction: %shift
286286 ; XOPAVX: Found an estimated cost of 6 for instruction: %shift
295295 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v32i16':
296296 ; SSE2: Found an estimated cost of 4 for instruction: %shift
297297 ; SSE41: Found an estimated cost of 4 for instruction: %shift
298 ; AVX: Found an estimated cost of 4 for instruction: %shift
298 ; AVX: Found an estimated cost of 8 for instruction: %shift
299299 ; AVX2: Found an estimated cost of 2 for instruction: %shift
300300 ; AVX512F: Found an estimated cost of 2 for instruction: %shift
301301 ; AVX512BW: Found an estimated cost of 1 for instruction: %shift
528528 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v4i64':
529529 ; SSE2: Found an estimated cost of 2 for instruction: %shift
530530 ; SSE41: Found an estimated cost of 2 for instruction: %shift
531 ; AVX: Found an estimated cost of 2 for instruction: %shift
531 ; AVX: Found an estimated cost of 4 for instruction: %shift
532532 ; AVX2: Found an estimated cost of 1 for instruction: %shift
533533 ; AVX512: Found an estimated cost of 1 for instruction: %shift
534534 ; XOPAVX: Found an estimated cost of 6 for instruction: %shift
541541 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i64':
542542 ; SSE2: Found an estimated cost of 4 for instruction: %shift
543543 ; SSE41: Found an estimated cost of 4 for instruction: %shift
544 ; AVX: Found an estimated cost of 4 for instruction: %shift
544 ; AVX: Found an estimated cost of 8 for instruction: %shift
545545 ; AVX2: Found an estimated cost of 2 for instruction: %shift
546546 ; AVX512: Found an estimated cost of 1 for instruction: %shift
547547 ; XOPAVX: Found an estimated cost of 12 for instruction: %shift
566566 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i32':
567567 ; SSE2: Found an estimated cost of 2 for instruction: %shift
568568 ; SSE41: Found an estimated cost of 2 for instruction: %shift
569 ; AVX: Found an estimated cost of 2 for instruction: %shift
569 ; AVX: Found an estimated cost of 4 for instruction: %shift
570570 ; AVX2: Found an estimated cost of 1 for instruction: %shift
571571 ; AVX512: Found an estimated cost of 1 for instruction: %shift
572572 ; XOPAVX: Found an estimated cost of 6 for instruction: %shift
579579 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i32':
580580 ; SSE2: Found an estimated cost of 4 for instruction: %shift
581581 ; SSE41: Found an estimated cost of 4 for instruction: %shift
582 ; AVX: Found an estimated cost of 4 for instruction: %shift
582 ; AVX: Found an estimated cost of 8 for instruction: %shift
583583 ; AVX2: Found an estimated cost of 2 for instruction: %shift
584584 ; AVX512: Found an estimated cost of 1 for instruction: %shift
585585 ; XOPAVX: Found an estimated cost of 12 for instruction: %shift
604604 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i16':
605605 ; SSE2: Found an estimated cost of 2 for instruction: %shift
606606 ; SSE41: Found an estimated cost of 2 for instruction: %shift
607 ; AVX: Found an estimated cost of 2 for instruction: %shift
607 ; AVX: Found an estimated cost of 4 for instruction: %shift
608608 ; AVX2: Found an estimated cost of 1 for instruction: %shift
609609 ; AVX512: Found an estimated cost of 1 for instruction: %shift
610610 ; XOPAVX: Found an estimated cost of 6 for instruction: %shift
617617 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i16':
618618 ; SSE2: Found an estimated cost of 4 for instruction: %shift
619619 ; SSE41: Found an estimated cost of 4 for instruction: %shift
620 ; AVX: Found an estimated cost of 4 for instruction: %shift
620 ; AVX: Found an estimated cost of 8 for instruction: %shift
621621 ; AVX2: Found an estimated cost of 2 for instruction: %shift
622622 ; AVX512F: Found an estimated cost of 2 for instruction: %shift
623623 ; AVX512BW: Found an estimated cost of 1 for instruction: %shift
193193 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v4i64':
194194 ; SSE2: Found an estimated cost of 2 for instruction: %shift
195195 ; SSE41: Found an estimated cost of 2 for instruction: %shift
196 ; AVX: Found an estimated cost of 2 for instruction: %shift
196 ; AVX: Found an estimated cost of 4 for instruction: %shift
197197 ; AVX2: Found an estimated cost of 1 for instruction: %shift
198198 ; AVX512: Found an estimated cost of 1 for instruction: %shift
199199 ; XOPAVX: Found an estimated cost of 4 for instruction: %shift
208208 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i64':
209209 ; SSE2: Found an estimated cost of 4 for instruction: %shift
210210 ; SSE41: Found an estimated cost of 4 for instruction: %shift
211 ; AVX: Found an estimated cost of 4 for instruction: %shift
211 ; AVX: Found an estimated cost of 8 for instruction: %shift
212212 ; AVX2: Found an estimated cost of 2 for instruction: %shift
213213 ; AVX512: Found an estimated cost of 1 for instruction: %shift
214214 ; XOPAVX: Found an estimated cost of 8 for instruction: %shift
237237 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v8i32':
238238 ; SSE2: Found an estimated cost of 2 for instruction: %shift
239239 ; SSE41: Found an estimated cost of 2 for instruction: %shift
240 ; AVX: Found an estimated cost of 2 for instruction: %shift
240 ; AVX: Found an estimated cost of 4 for instruction: %shift
241241 ; AVX2: Found an estimated cost of 1 for instruction: %shift
242242 ; AVX512: Found an estimated cost of 1 for instruction: %shift
243243 ; XOPAVX: Found an estimated cost of 4 for instruction: %shift
252252 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i32':
253253 ; SSE2: Found an estimated cost of 4 for instruction: %shift
254254 ; SSE41: Found an estimated cost of 4 for instruction: %shift
255 ; AVX: Found an estimated cost of 4 for instruction: %shift
255 ; AVX: Found an estimated cost of 8 for instruction: %shift
256256 ; AVX2: Found an estimated cost of 2 for instruction: %shift
257257 ; AVX512: Found an estimated cost of 1 for instruction: %shift
258258 ; XOPAVX: Found an estimated cost of 8 for instruction: %shift
281281 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v16i16':
282282 ; SSE2: Found an estimated cost of 2 for instruction: %shift
283283 ; SSE41: Found an estimated cost of 2 for instruction: %shift
284 ; AVX: Found an estimated cost of 2 for instruction: %shift
284 ; AVX: Found an estimated cost of 4 for instruction: %shift
285285 ; AVX2: Found an estimated cost of 1 for instruction: %shift
286286 ; AVX512: Found an estimated cost of 1 for instruction: %shift
287287 ; XOPAVX: Found an estimated cost of 4 for instruction: %shift
296296 ; CHECK: 'Cost Model Analysis' for function 'splatvar_shift_v32i16':
297297 ; SSE2: Found an estimated cost of 4 for instruction: %shift
298298 ; SSE41: Found an estimated cost of 4 for instruction: %shift
299 ; AVX: Found an estimated cost of 4 for instruction: %shift
299 ; AVX: Found an estimated cost of 8 for instruction: %shift
300300 ; AVX2: Found an estimated cost of 2 for instruction: %shift
301301 ; AVX512F: Found an estimated cost of 2 for instruction: %shift
302302 ; AVX512BW: Found an estimated cost of 1 for instruction: %shift
530530 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v4i64':
531531 ; SSE2: Found an estimated cost of 2 for instruction: %shift
532532 ; SSE41: Found an estimated cost of 2 for instruction: %shift
533 ; AVX: Found an estimated cost of 2 for instruction: %shift
533 ; AVX: Found an estimated cost of 4 for instruction: %shift
534534 ; AVX2: Found an estimated cost of 1 for instruction: %shift
535535 ; AVX512: Found an estimated cost of 1 for instruction: %shift
536536 ; XOPAVX: Found an estimated cost of 4 for instruction: %shift
543543 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i64':
544544 ; SSE2: Found an estimated cost of 4 for instruction: %shift
545545 ; SSE41: Found an estimated cost of 4 for instruction: %shift
546 ; AVX: Found an estimated cost of 4 for instruction: %shift
546 ; AVX: Found an estimated cost of 8 for instruction: %shift
547547 ; AVX2: Found an estimated cost of 2 for instruction: %shift
548548 ; AVX512: Found an estimated cost of 1 for instruction: %shift
549549 ; XOPAVX: Found an estimated cost of 8 for instruction: %shift
569569 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v8i32':
570570 ; SSE2: Found an estimated cost of 2 for instruction: %shift
571571 ; SSE41: Found an estimated cost of 2 for instruction: %shift
572 ; AVX: Found an estimated cost of 2 for instruction: %shift
572 ; AVX: Found an estimated cost of 4 for instruction: %shift
573573 ; AVX2: Found an estimated cost of 1 for instruction: %shift
574574 ; AVX512: Found an estimated cost of 1 for instruction: %shift
575575 ; XOPAVX: Found an estimated cost of 4 for instruction: %shift
582582 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i32':
583583 ; SSE2: Found an estimated cost of 4 for instruction: %shift
584584 ; SSE41: Found an estimated cost of 4 for instruction: %shift
585 ; AVX: Found an estimated cost of 4 for instruction: %shift
585 ; AVX: Found an estimated cost of 8 for instruction: %shift
586586 ; AVX2: Found an estimated cost of 2 for instruction: %shift
587587 ; AVX512: Found an estimated cost of 1 for instruction: %shift
588588 ; XOPAVX: Found an estimated cost of 8 for instruction: %shift
607607 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v16i16':
608608 ; SSE2: Found an estimated cost of 2 for instruction: %shift
609609 ; SSE41: Found an estimated cost of 2 for instruction: %shift
610 ; AVX: Found an estimated cost of 2 for instruction: %shift
610 ; AVX: Found an estimated cost of 4 for instruction: %shift
611611 ; AVX2: Found an estimated cost of 1 for instruction: %shift
612612 ; AVX512: Found an estimated cost of 1 for instruction: %shift
613613 ; XOPAVX: Found an estimated cost of 4 for instruction: %shift
620620 ; CHECK: 'Cost Model Analysis' for function 'splatconstant_shift_v32i16':
621621 ; SSE2: Found an estimated cost of 4 for instruction: %shift
622622 ; SSE41: Found an estimated cost of 4 for instruction: %shift
623 ; AVX: Found an estimated cost of 4 for instruction: %shift
623 ; AVX: Found an estimated cost of 8 for instruction: %shift
624624 ; AVX2: Found an estimated cost of 2 for instruction: %shift
625625 ; AVX512F: Found an estimated cost of 2 for instruction: %shift
626626 ; AVX512BW: Found an estimated cost of 1 for instruction: %shift