[PATCH] D100633: [AMDGPU] Remove error check for indirect calls and add missing queue-ptr
Matt Arsenault via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Thu Apr 22 18:17:28 PDT 2021
arsenm added inline comments.
================
Comment at: llvm/test/CodeGen/AMDGPU/simple-indirect-call.ll:42-44
+ %fptr = alloca void()*, addrspace(5)
+ %fptr.cast = addrspacecast void()* addrspace(5)* %fptr to void()**
+ store void()* @indirect, void()** %fptr.cast
----------------
madhur13490 wrote:
> arsenm wrote:
> > madhur13490 wrote:
> > > arsenm wrote:
> > > > Why bother with the alloca/addrspacecast? This ends up getting optimized to LDS anyway
> > > specifically denote typical sequence of instructions generated by frontend
> > What the frontend initially emits is less important unless we for some reason would break it. This should use the minimum IR to exercise the feature
> But we have many tests which covers minimum IR. We don't have any which covers typical pattern generated by frontend.
But this isn't what you are testing. The IR should be the minimum set required to reproduce the issue you are attempting to fix
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D100633/new/
https://reviews.llvm.org/D100633
More information about the llvm-commits
mailing list