CK Tile Group GEMM gfx1250

[aris134]

Description

Extend the present CK tile grouped GEMM (F16/F8) implementation for compatibility with gfx1250.

NOTE: I realized this PR also depends on the newer CK module revision provided for the grouped GEMM work. Since the upcoming MXFP8 PR will have the same dependency, I plan to split the CK module update into a separate prerequisite PR against gfx1250, then retarget/rebase this PR on top of that branch. This should avoid duplicating the CK update across both feature PRs.

Fixes #16286

Type of change

• Documentation change (change only to the documentation, either a fix or a new content)
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Infra/Build change
• Code refactoring

Changes

Please list the changes introduced in this PR:

• Change A
• Change B

Checklist:

• I have read and followed the contributing guidelines
• The functionality is complete
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

[ipanfilo]

It is CUDA path

[aris134]

Reverted in 752e0d3

[ipanfilo]

Why does it need template over reguler if-else or switch-case?

[aris134]

The template is needed because the arch selection affects CK kernel template instantiation, not just runtime control flow. GPUArch must be a compile-time value so if constexpr can prune unsupported tile/kernel combinations for a given architecture. In this case, it prevents the MFMA configs from being instantiated for gfx1250.