[llvm-branch-commits] [mlir] [mlir] Implement indexed access op interfaces for memref, vector, gpu, nvgpu (PR #177014)

[Andrzej Warzyński via llvm-branch-commits]

banach-space wrote:

Thanks! For those of us not quite in the loop, could you provide some justification/rationale? Thanks!

> PR stack (managed by hand):
> 
> * [[mlir][memref] Define interfaces for ops that access memrefs at an index #177013](https://github.com/llvm/llvm-project/pull/177013) - defining interfaces
> * [[mlir] Implement indexed access op interfaces for memref, vector, gpu, nvgpu #177014](https://github.com/llvm/llvm-project/pull/177014) - implementing interfaces (<- you are here)
> * [[mlir] Add [may]updateStartingPosition to VectorTransferOpInterface #177015](https://github.com/llvm/llvm-project/pull/177015) - amending vector transfer interface
> * [[mlir][MemRef] Make fold-memref-alias-ops use memref interfaces #177016](https://github.com/llvm/llvm-project/pull/177016) - using interfaces in fold-memref-alias-ops, testing

IIIUC, this stack is to replace the current PR?

https://github.com/llvm/llvm-project/pull/177014