[PATCH] D82652: AMDGPU/GlobalISel: Select icmp intrinsic

Matt Arsenault via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Jun 26 07:40:03 PDT 2020


arsenm added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp:1019-1020
+
+  assert(MRI->getType(Dst).getSizeInBits() == STI.getWavefrontSize() &&
+         "Result size should match WavefrontSize");
+
----------------
foad wrote:
> arsenm wrote:
> > Should fail to select rather than assert
> I suggested this assert, based on my interpretation of your [[ https://reviews.llvm.org/D65088?id=211078#inline-582680 | comment ]] that there was no need to try to handle misuse of the intrinsics because they should really be checked by some kind of IR verification.
They shouldn't crash, but also shouldn't select. This matches the behavior of intrinsics on subtargets that don't really support the instruction


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D82652/new/

https://reviews.llvm.org/D82652





More information about the llvm-commits mailing list