[Mlir-commits] [mlir] [mlir][Target] Support Fatbin target for static nvptxcompiler (PR #118044)

[Mehdi Amini]

joker-eph wrote:

> My concern is downstream projects that use custom extracted Cuda Toolkit (not extracted libnvfatbin_static.a) will directly cause the original build to fail after this MR if there is no MLIR_ENABLE_NVPTXCOMPILER_NVFATBIN.

I don't see this as a problem? Seems like a legitimate break and they should update their flow.

> Of course, this does not preserve consistency when MLIR_ENABLE_NVPTXCOMPILER_NVFATBIN=OFF.

That makes it something we should not support then IMO.

https://github.com/llvm/llvm-project/pull/118044