Skip to content

Conversation

@Simonsays095
Copy link
Contributor

Recent changes alter a user-facing API in gemmstone (the kernel selector's select function) which other teams use. Since the change in API is unnecessary, this PR reverts that change and instead adds a new API:

  • select: Earlier API, unchanged. Returns the lowest-scoring (best) matching kernel from the catalog.
  • collect_kernels: Returns a vector of entries, sorted by score. Also includes additional information in the form of EvaluateAuxOutput and score, for later use.

This also is the first step to simplifying gen_desc_t, as the set_entry function sets both aux_params_ and entry, guaranteeing consistency in the state.

@Simonsays095 Simonsays095 requested a review from a team as a code owner January 9, 2026 19:23
@github-actions github-actions bot added the platform:gpu-intel Codeowner: @oneapi-src/onednn-gpu-intel label Jan 9, 2026
@Simonsays095
Copy link
Contributor Author

make test
disable test_device_cpu
disable build_cpu_runtime_omp
disable build_cpu_runtime_sycl
disable build_cpu_runtime_tbb
disable benchdnn_all
enable benchdnn_matmul
enable benchdnn_ip
enable benchdnn_rn

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

platform:gpu-intel Codeowner: @oneapi-src/onednn-gpu-intel

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants