[PATCH] D27811: [CostModel][X86] Add support for broadcast shuffle costs

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: lib/Target/X86/X86TargetTransformInfo.cpp:683
+      { TTI::SK_Broadcast, MVT::v8i32,  2 }, // vperm2f128 + vpermilps
+      { TTI::SK_Broadcast, MVT::v16i16, 3 }, // vpshuflw + vpshufd + vinsertf128
+      { TTI::SK_Broadcast, MVT::v32i8,  2 }, // vpshufb + vinsertf128
----------------
delena wrote:
> theoretically, v16i16 should not be more expensive than v32i8, vpshufb should work for v8i16 as it works for v16i8. (I don't know how it is implemented).
It does currently have the vpshuflw + vpshufd + vinsertf128 pattern - I'll look improving this in shuffle combining.


Repository:
  rL LLVM

https://reviews.llvm.org/D27811