[llvm] [SPIRV] convergence anchor intrinsic does not have a parent token (PR #122230)

Steven Perron via llvm-commits llvm-commits at lists.llvm.org
Thu Jan 9 10:15:28 PST 2025


s-perron wrote:

> Is there an existing test case that even reaches here?

That is a good question. I thought there would be. If you have a valid use of `Intrinsic::experimental_convergence_anchor`, then it would execute that case in the old code, and would assert. That would be a test case that executes the old code, and shows the need for it to be fixed.

Your change would avoid the assert, and do something. However, I'm not convinced that what it is doing is the right thing.

Sorry for my ignorance, but I'm not sure what a reasonable and expected us of the anchor is. That is why I cannot reason about it very well. I'm assuming the existing test do not use the anchor intrinsic because it would have either failed the verifier (as you said) or asserted.

https://github.com/llvm/llvm-project/pull/122230


More information about the llvm-commits mailing list