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

[Ivan Chikish via Phabricator via llvm-commits]

Nekotekina added a comment.

Added some code analyser results.



================
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
----------------
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.


================
Comment at: llvm/test/CodeGen/X86/vector-fshl-rot-128.ll:379
+; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm0
+; AVX2-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    retq
----------------
Even less instructions than on the left


================
Comment at: llvm/test/CodeGen/X86/vector-fshl-rot-128.ll:1549
+; AVX512VLBW-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX512VLBW-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
 ; AVX512VLBW-NEXT:    retq
----------------
Left: https://bit.ly/41G6Mia
Right: https://bit.ly/43Wymcm


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

https://reviews.llvm.org/D149071