[CK_TILE] Fix DP + 2 Tile Stream-K Validation Errors

[ecamartins]

Proposed changes

In CK Tile Stream-K, there may be multiple workgroups contributing to a given tile in the C tensor. When the accumulation strategy involves atomics, there may be round off error in cases where the accumulator type is not the same as the C type. To compute an error tolerance for test validation, the Stream-K Tile Partitioner has a function called estimate_num_wgs_per_tile to estimate the number of workgroups per tile. That said, this function only provides an estimate; it may underestimate in some cases. This underestimation was causing the error tolerance computed by calculate_rtol_atol to be too low for the DP + 2 Tile SK test cases. This led to a regression for some Stream-K tests. Note, that these validation failures due to round-off errors may not always be present due to the non-determinism that accompany atomics.

Thus, this change updates the estimate_num_wgs_per_tile function to explicitly return the value of 2 for DP+2TSK instances to ensure that we have a better error tolerance to avoid test failures due to round-off error.

We tested locally on gfx90a, gfx942, and gfx950; all tests pass. We found some unrelated issues on gfx908, so we have disabled our tests on gfx908 for now and will be creating a ticket to investigate the issues.

Checklist

Please put an x into the boxes that apply. You can also fill these out after creating the PR. If you're not sure, please don't hesitate to ask.

• I have added tests relevant to the introduced functionality, and the unit tests are passing locally
• I have added the test to REGRESSION_TESTS list defined at the top of CMakeLists.txt in tests/CMakeLists.txt, IF the test takes more than 30 seconds to run.
• I have added inline documentation which enables the maintainers with understanding the motivation
• I have removed the stale documentation which is no longer relevant after this pull request
• (If this change is user-facing) I have added release notes which provide the end users with a brief summary of the improvement from this pull request
• I have run clang-format on all changed files
• Any dependent changes have been merged