[PATCH] D138840: Verifier: Disallow uses of intrinsic global variables

Artem Belevich via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Nov 28 14:52:35 PST 2022


tra added a comment.

> Verifier: Disallow uses of intrinsic global variables

Is this the correct description for this patch? The modified test itself does not seem to use any global variables. Perhaps I'm missing something.



================
Comment at: llvm/test/CodeGen/AMDGPU/cgp-addressing-modes-flat.ll:31
+; OPT-GFX8-NEXT:  entry:
+; OPT-GFX8-NEXT:    [[OUT_GEP:%.*]] = getelementptr i32, i32* [[OUT:%.*]], i64 999999
+; OPT-GFX8-NEXT:    [[IN_GEP:%.*]] = getelementptr i32, i32* [[IN:%.*]], i64 7
----------------
I'm just curious whether it's expected that some address calculations were not sunk into a conditional branch as it's done on GFX9/10?
I would assume that such a generic optimization would be GPU-agnostic. 



CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D138840/new/

https://reviews.llvm.org/D138840



More information about the llvm-commits mailing list