Skip to content

Conversation

@krithalith
Copy link
Contributor

@krithalith krithalith commented Jan 7, 2026

As titled. The following flavors are added:

  • grouped_conv2d_fwd_dynamic_op (f16, bf16)
  • grouped_conv3d_fwd_dynamic_op (f16, bf16)
  • grouped_conv3d_fwd_convscale (f8f8, f8bf8, bf8f8, bf8bf8)
  • grouped_conv3d_fwd_convinvscale (f8)
  • grouped_conv3d_fwd_convscale_add (f8)
  • grouped_conv3d_fwd_convscale_relu (f8)
  • grouped_conv3d_fwd_combconvscale (f8 -> f32)
  • grouped_conv3d_fwd_scaleadd_scaleadd_relu (f16, bf16)

As these are not critical right now, we will only have one generic instance active per instance list, to reduce compile and test time. However, all instances should work and can be activated when necessary. None of these flavors have ever had Large Tensor versions.

Added instance lists, profilers, tests, and examples. Verified that the single generic instance per list provides enough support to have instances for all the expected tests.

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

wj-laskowski and others added 4 commits January 7, 2026 15:26
As titled. Following variants are added:
- grouped_conv2d_fwd_dynamic_op
- grouped_conv3d_fwd_dynamic_op
- grouped_conv3d_fwd_bilinear
- grouped_conv3d_fwd_convscale
- grouped_conv3d_fwd_convinvscale
- grouped_conv3d_fwd_convscale_add
- grouped_conv3d_fwd_convscale_relu
- grouped_conv3d_fwd_scale
- grouped_conv3d_fwd_combconvscale
- grouped_conv3d_fwd_scaleadd_scaleadd_relu
…rary() cmakelists function so we don't build f8 on RDNA3.
…ated to the post-ces extra flavors, with scalarPerVector = 1. Then disable all but one generic instance per instance list to reduce compile time.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants