[Mlir-commits] [mlir] [mlir][gpu] Add innermost-first policy when mapping loops to GPU IDs (PR #160634)

[Georgios Pinitas]

GeorgeARM wrote:

> Why is this is necessary if you can reorder the loop?

HW mapping is rather cheap compared to loop reordering in a way; especially if we unsettle the IR and cause any issues with analysis that we might have in place.

https://github.com/llvm/llvm-project/pull/160634