[clang] [llvm] [IR] Mark convergence intrins as has-side-effect (PR #134844)
    Sameer Sahasrabuddhe via llvm-commits 
    llvm-commits at lists.llvm.org
       
    Wed Apr  9 20:56:45 PDT 2025
    
    
  
ssahasra wrote:
Thanks for the example! I think #133684 is a correct incremental step forward, but maybe just going the whole way bottom-up is better. It is also correct to remove the verifier check. It's not a well-formedness error or even a semantic error to pass a convergence control token to a non-convergent call. It is also not UB. The only outcome is that the user will not get the convergence they expected, and this is not even observable since there are no convergent operations that will be affected by this unused token.
@nhaehnle does this look okay to you?
https://github.com/llvm/llvm-project/pull/134844
    
    
More information about the llvm-commits
mailing list