[PATCH] D113778: [AMDGPU] Use shift for b64 mov

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec added a comment.

In D113778#3127656 <https://reviews.llvm.org/D113778#3127656>, @foad wrote:

> Interesting. I see you're doing this when expanding V_MOV_B64_PSEUDO, but I don't really understand when we use V_MOV_B64_PSEUDO in the first place. copyPhysReg() does not generate it, instead it copies the logic from here to emit V_PK_MOV_B32. So does that mean you need to add your V_LSHLREV_B64_e64 code to copyPhysReg too?

Pseudo was created to deal with 64 bit immediates and fold these. It is not needed that late.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D113778/new/

https://reviews.llvm.org/D113778