[Mlir-commits] [mlir] [mlir][SCF] Unify tileUsingFor and tileReductionUsingFor implementation (PR #120115)

[llvmlistbot at llvm.org]

MaheshRavishankar wrote:

@Groverkss  I am looking at this again, and looking through this again with fresh eyes, I am not sure this was a good direction to go. It seems like two separate algorithms mashed together with a switch case at every step. Further the `SCFTilingResult` has fields now that are irrelevant in certain cases. For example `MergeReduction` and `initialValue` are not relevant for the case where you are just doing tiling. 

I think it is better to split these back into two separate implementations cause they generate two really different code structures. I am happy to walk this back a bit while maintaining as much code-reuse as we can.

https://github.com/llvm/llvm-project/pull/120115