[PATCH] D116363: [X86] Combine to vpdpbusd when operand is constant and small enough.

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: llvm/test/CodeGen/X86/dpbusd_const.ll:124
+; AVX512VNNI-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3,4,5,6,7]
+; AVX512VNNI-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [0,1,2,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512VNNI-NEXT:    vpxor %xmm2, %xmm2, %xmm2
----------------
LuoYuanke wrote:
> This looks not good, as it accesses more memory.
This is very similar to https://llvm.org/PR45808 - we're not making use of implicit zeroing of upper subvectors to make smaller constants


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D116363/new/

https://reviews.llvm.org/D116363