[PATCH] D64659: AMDGPU: Decompose all values to 32-bit pieces for calling conventions

Fri Jul 12 12:38:22 PDT 2019

arsenm marked an inline comment as done.
arsenm added inline comments.

================
Comment at: lib/Target/AMDGPU/AMDGPUCallingConv.td:141
     VGPR24, VGPR25, VGPR26, VGPR27, VGPR28, VGPR29, VGPR30, VGPR31]>>,
-  CCIfType<[i64, f64, v2i32, v2f32, v3i32, v3f32, v4i32, v4f32, v5i32, v5f32, v8i32, v8f32, v16i32, v16f32, v2i64, v2f64, v4i16, v4f16], CCCustom<"allocateVGPRTuple">>,
   CCIfType<[i32, f32, v2i16, v2f16, i16, f16, i1], CCAssignToStack<4, 4>>,
----------------
rampitec wrote:
> Should you also list v32i32?
This removes the need to list the tuple types. I think this can also be removed for the stack passed arguments, but I'll look at that separately.

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D64659/new/

https://reviews.llvm.org/D64659