[PATCH] D109082: [AMDGPU] Use S_BITCMP1_* to replace AND in optimizeCompareInstr

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec added a comment.

In D109082#2979040 <https://reviews.llvm.org/D109082#2979040>, @foad wrote:

>> I.e. to my view relying on an ISel is a layering violation.
>
> In my view ISel is responsible for selection of instructions, and anything which changes those instructions to different ones later on is a bit of a hack :)
>
> But I take your point about moveToVALU. That does sound a bit awkward.

Everything about VALU to SALU relationship is awkward. We basically have two separate engines tied together and that does not make compiler life easy. For example x86 has finally killed their x87 coprocessor because it is awkward to deal with.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D109082/new/

https://reviews.llvm.org/D109082