[GitHub] [tvm] Wheest edited a comment on pull request #6137: Better grouped convolution for CPU targets

[GitBox <gi...@apache.org>]

Wheest edited a comment on pull request #6137:
URL: https://github.com/apache/tvm/pull/6137#issuecomment-749136716


   Hello there, updating this pull request to be up-to-date with the latest `main` branch.
   
   In terms of things remaining to do:
   
   - [x] [Consider using compute_at and vectorize data load](https://github.com/apache/tvm/pull/6137#discussion_r463852066) - did not get an improvement.
   - [x] [We should support asymmetic padding like other compute / schedule.](https://github.com/apache/tvm/pull/6137#pullrequestreview-459474121) - this is implemented in GSPC, however requires extending `get_workload` for Conv2D generally.  I began working on this in `505c127`, but have reverted it, and will have this as it's own pull request in the comings days.
   - [ ] [Pack in alter_op_layout for kernel](https://github.com/apache/tvm/pull/6137#discussion_r463844394):  have been working on this, but have an issue.   My data is being passed to my `group_conv2d_NCHWc.x86` in the `conv2d_NCHWc` format (5D input data), rather than the GSPC format (6D input data).  Despite my changes to the x86 `_alter_conv2d_layout`.  [See this branch](https://github.com/Wheest/incubator-tvm/tree/wheest/gspc-dev-alter-op).  Some guidance or pointers would be appreciated @FrozenGene.
   
   In the interests of more transparent development, [here's part of my test suite](https://github.com/Wheest/tvm-grouped-conv-test).


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org