[PATCH] D77547: AMDGPU/GlobalISel: Select llvm.amdgcn.div.scale

[Austin Kerbow via Phabricator via llvm-commits]

kerbowa added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp:825
+
+  Register Src0 = ChooseDenom != 0 ? Numer : Denom;
+
----------------
foad wrote:
> This doesn't seem to match the comment in IntrinsicsAMDGPU.td which says "0 = first, 1 = second" (where I assume "first" means "numerator"...).
It looks like this is copying the SD lowering to the intermediate node for this intrinsic which doesn't match up to that comment in IntrinsicsAMDGPU.td. So the GlobalISel tests at least match the DAG version. Was the intrinsic not being used directly before? I think it needs to be changed in both places or the comment should be updated.




CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D77547/new/

https://reviews.llvm.org/D77547