[Mlir-commits] [mlir] [mlir][scf] Revisit affected loops during range folding (PR #230961)

[Ivan Butygin]

https://github.com/Hardcode84 commented:

> And the second is on a synthetic 1000-level nested-loop test, the greedy version takes ~39s vs. ~0.18s with the targeted worklist (measured as mlir-opt wall-clock time). This is an extreme stress case, but it suggests significant overhead from ancestor re-enqueuing.

Thanks for the experiment. Can you address the `walk` comment and try again, looking at `applyOpPatternsGreedily`,  `findCommonAncestor` can go quadratic if list on nested ops are passed.

https://github.com/llvm/llvm-project/pull/230961