[Mlir-commits] [mlir] [MLIR][GPU] subgroup_mma fp64 extension (PR #165873)

[Giacomo Castiglioni]

castigli wrote:

This PR is causing the CI to fail. I think the issue is that the new test  should be guarded by the flag `mlir_run_cuda_sm80_tests` and to run on the cuda runner with the `T4` gpu.
The fix is to move the file under `/mlir/test/Integration/GPU/CUDA/TensorCore/sm80`.
 What is the the procedure for a fix like this? Revert and re-land?

https://github.com/llvm/llvm-project/pull/165873