[Mlir-commits] [mlir] [MLIR][LLVM] Fix inlining of a single block ending with unreachable (PR #122646)

Oleksandr Alex Zinenko llvmlistbot at llvm.org
Mon Jan 13 09:04:17 PST 2025


ftynse wrote:

I'm onboard with the current proposition (`allowSingleBlockOptimization`), looks like the least intrusive change IMO. We can always revise if we get consensus for a bigger change. I'm rather undecided on having to run region simplification as a cleanup. Same for canonicalization, btw. On one hand, I don't want to duplicate simplification code everywhere when we have dedicated logic. On another hand, I don't think it's reasonable to run canonicalizer/CSE/DCE/CFG-simplify after every single pass. We can circulate the idea of removing this optimization completely on the forum, but beware of it devolving into that discussion.

https://github.com/llvm/llvm-project/pull/122646


More information about the Mlir-commits mailing list