[Mlir-commits] [mlir] [mlir][amdgpu] Add tensor load store operations (PR #170918)

[Krzysztof Drewniak]

https://github.com/krzysz00 requested changes to this pull request.

Changes requested because of the size bit - I'll be getting on an LLVM rewrite.

I think the nontemporality also may need further discussion, especially given that gfx9 and co have an entirely different set of flags there and there's been a strong sentiment from the backend folks that they'd prefer we not set these flags by hand

https://github.com/llvm/llvm-project/pull/170918