[Mlir-commits] [mlir] [mlir][amdgpu] Shared memory access optimization pass (PR #75627)

[llvmlistbot at llvm.org]

erman-gurses wrote:

> "existing precedent" isn't a reason IMO. It's not because a mistake slipped that it justified continuing a bad pattern... But if @erman-gurses wants to tackle this as a separate PR that's fine by me.

@joker-eph, yes, I would like to tackle this as a separate PR. 

https://github.com/llvm/llvm-project/pull/75627