[PATCH] D149071: [X86] LowerRotate: prefer unpack-based algorithm

[Noah Goldstein via Phabricator via llvm-commits]

goldstein.w.n added inline comments.


================
Comment at: llvm/test/CodeGen/X86/min-legal-vector-width.ll:1942
+; CHECK-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; CHECK-NEXT:    retq
   %b8 = sub <16 x i8> <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>, %b
----------------
Nekotekina wrote:
> Code analyser:
> SKX version (left): https://bit.ly/3NamuO5
> VBMI version (left): https://bit.ly/444ZUfB
> 
> Common (right): https://bit.ly/3n7q4xQ
> Unpack version looks better despite more instructions.
This seems potentially not worth it for ICX unless its in a tight inner loop.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D149071/new/

https://reviews.llvm.org/D149071