[llvm-branch-commits] [mlir] [AMDGPU][MLIR] Replace gfx940 and gfx941 with gfx942 in MLIR (PR #125836)

[Fabian Ritter via llvm-branch-commits]

ritter-x2a wrote:

> Since this essentially breaks logic for gfx940 and gfx941, should we assert in code like `Chipset` that these are not used and silently miscompiled?

@kuhar as far as I can see that would be the first check for invalid targets in this part of the code base. I don't think we should treat gfx940/gfx941 differently than any other invalid target. We will only merge this PR together with the rest of the stack, which will remove gfx940\gfx941 code generation from the rest of LLVM as well.

https://github.com/llvm/llvm-project/pull/125836