[PATCH] D25652: [AVX-512] Add support for commuting VPERMT2(B/W/D/Q/PS/PD) to/from VPERMI2(B/W/D/Q/PS/PD).

[Elena Demikhovsky via llvm-commits]

delena added a comment.

VPERMI2Q kills register with indices. I thought that VPERMT2Q is better for loop vectorizer, since we may want to repeat the same shuffle on multiple data.


https://reviews.llvm.org/D25652