[PATCH] D125319: [AMDGPU] gfx11 BUF Instructions

Wed May 11 07:53:13 PDT 2022

foad added a comment.
Herald added a subscriber: kosarev.

In D125319#3506104 <https://reviews.llvm.org/D125319#3506104>, @Joe_Nash wrote:

> In D125319#3506059 <https://reviews.llvm.org/D125319#3506059>, @foad wrote:
>
>>> Includes a small change in SMInstructions.td to correct encoded bits.
>>
>> Sounds like that should be a separate patch. Why doesn't it affect any MC tests?
>
> It has to do with SIRegisterInfo.td:214. We now have pregfx11 and postgfx11 version of M0 and SGPR_NULL encoding instead of a common one. So SIRegisterInfo, BUFInstructions and SMInstructions need to be changed in the same commit. Test output doesn't change because the same value is encoded, just referring to a new def name.

OK. From the commit message I thought you meant it was an unrelated bug fix.

================
Comment at: llvm/lib/Target/AMDGPU/SIRegisterInfo.td:214

-defm M0 : SIRegLoHi16 <"m0", 124>;
-defm SGPR_NULL : SIRegLoHi16 <"null", 125>;
+// encoding changes between gen, see Utils/AMDGPUBaseInfo.cpp MAP_REG2REG
+defm M0_gfxpre11 : SIRegLoHi16 <"m0", 124>;
----------------
Nit: comments should generally read and be punctuated like full sentences.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D125319/new/

https://reviews.llvm.org/D125319