[Mlir-commits] [mlir] [MLIR] Fuse locations of hoisted / merged constants (PR #74670)

[Mehdi Amini]

joker-eph wrote:

I looked at the LLVM revision, but I don't find the "parent op location" equivalent in MLIR here?
(I'm not sure I totally understood it though, their tests are... cryptic to me).

I actually may start to understand somehow: they attach the fused location of the *insertion point* somehow (for PGO reasons apparently), and you're trying to translate it to the "parent op" because you think that in the entry point it'll translate to the closest to this? 

It'll probably help to look at a dedicated revision with dedicated examples crafted to showcase this.

https://github.com/llvm/llvm-project/pull/74670