[PATCH] D105263: [X86] AVX512FP16 instructions enabling 1/6

[Pengfei Wang via Phabricator via llvm-commits]

pengfei added inline comments.


================
Comment at: llvm/lib/Target/X86/X86InstrAVX512.td:4159
+defm VMOVSHZ : avx512_move_scalar<"vmovsh", X86Movsh, X86vzload16, f16x_info,
+                                  [HasFP16]>,
+                                  VEX_LIG, T_MAP5XS, EVEX_CD8<16, CD8VT1>;
----------------
pengfei wrote:
> LuoYuanke wrote:
> > Why there is no OptForSize for vmovsh?
> Good catch. I think we should add it here.
Sorry, I think we should not add `OptForSize` here.
This predicate is used to force to select blend instead of mov due to performance consideration.
E.g.: https://godbolt.org/z/W4v38K6va

Since we don't have a blendph instruction, I think we can always select it to movsh. Not sure if using pblendw is beneficial.
I'll change it back in next patch.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D105263/new/

https://reviews.llvm.org/D105263