[PATCH] D75138: [WIP][AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions

[Scott Linder via Phabricator via llvm-commits]

scott.linder marked an inline comment as done.
scott.linder added inline comments.


================
Comment at: llvm/test/CodeGen/AMDGPU/scratch-simple.ll:96
 
 ; GCN-LABEL: {{^}}hs_ir_uses_scratch_offset:
 ; GCN: s_mov_b32 s8, SCRATCH_RSRC_DWORD0
----------------
arsenm wrote:
> Can you add a comment elaborating on what this tests
>From discussion with @mareko my understanding is that Mesa GS and HS shaders have the preloaded scratch wave offset SGPR fixed at SGPR5, and the inreg implementation is used to reference it in the IR. So here, the shader snippet inserted after the SI_RETURN_TO_EPILOG wants to use the scratch wave offset, and the IR passes it along by padding out the inreg arguments until it gets to where the scratch wave offset is, and then using it in the return value. I'll add something to that effect in the test.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75138/new/

https://reviews.llvm.org/D75138