[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:

> Ok so I'm going to take issue with jumping the constants in a bunch of tests to gfx8942. Unless we're planning to completely wipe gfx940/1 from the codebase - which seems extremely unusual and like bad practice ... gfx940 and 941 are real targets that did really exist and might need code compiled for them in the future - we should still allow compiling for there targets while updating tests and such to gfx942.
> 
> And as to the atomics emulation pass ... what's the harm in keeping some old workaround around?

I'll bring this issue up for discussion in the next compiler team meeting.

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