[Mlir-commits] [mlir] [MLIR][RemoveDeadValues] Fix affine.for induction variable incorrectly removed (PR #172612)

[Francisco Geiman Thiesen]

FranciscoThiesen wrote:

> There is #169816 that seems to be addressing the same problem. I put a blocker on it for a testing/understanding reason, but generally I prefer that approach. Specifically, here, liveness is set on arguments at the very end of the process and is never propagated. So we will incorrectly mark as dead a result of a side-effect free operation whose operand is such an argument.
> 
> Could you folks collaborate with @linuxlonelyeagle to find a good way to test these?

I agree that #169816 is a better approach. Let me close this down and we can just wait for that.

https://github.com/llvm/llvm-project/pull/172612