[llvm] [RISCV][VLOPT] Fix assertion failure across blocks (PR #124734)
Michael Maitland via llvm-commits
llvm-commits at lists.llvm.org
Tue Jan 28 07:04:09 PST 2025
michaelmaitland wrote:
I'm not sure I agree with this approach. I think there is a deeper problem that is going on: something is on the worklist that is not a candidate, and this is an invariant we are currently not permitting.
I think we should handle it in one of two ways:
1. Allow things on the worklist that are not isCandidate. When when we pop it off, we will check whether it isCandidate and decide whether to optimize or not.
2. Do a better job of not adding something that !isCandidate to the worklist.
I think (1) is probably easier, but I don't have a strong preference.
https://github.com/llvm/llvm-project/pull/124734
More information about the llvm-commits
mailing list