[PATCH] D112733: [AMDGPU] Fix cvt_f32_ubyte combine with shl

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/test/CodeGen/AMDGPU/cvt_f32_ubyte_vector.ll:4-6
+%Vec = type { [4 x i8] }
+
+define amdgpu_kernel void @cvt_f32_ubyte0_vector() local_unnamed_addr {
----------------
vangthao wrote:
> arsenm wrote:
> > You can just put this in the existing test
> When I put it with the existing test, I am getting:
> 
> > LLVM ERROR: Cannot select: t86: ch = store<(store (s8) into i32* undef + 3), trunc to i8> t97, t51, undef:i64, undef:i64
> 
> This is coming from the first check with -mcpu=tahiti.
> 
Just change the flat pointers to addrspace(1)


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D112733/new/

https://reviews.llvm.org/D112733