[PATCH] D30209: AMDGPU: Fold omod into instructions

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: lib/Target/AMDGPU/AMDGPUCallingConv.td:38
   // 32*4 + 4 is the minimum for a fetch shader consumer with 32 inputs.
-  CCIfNotInReg<CCIfType<[f32, i32] , CCAssignToReg<[
+  CCIfNotInReg<CCIfType<[f32, i32, f16] , CCAssignToReg<[
     VGPR0, VGPR1, VGPR2, VGPR3, VGPR4, VGPR5, VGPR6, VGPR7,
----------------
mareko wrote:
> Why are the calling conventions being changed?
I needed a way to get an f16 input into a graphics shader. This would just assert on unhandled value type before. I can commit this separately


https://reviews.llvm.org/D30209