Skip to content

Conversation

@yincs-intellif
Copy link
Contributor

During NPU schedule, we usually encounter multiple out graphs and need do loop partition on each out then compute at other block to the different out block. But current "compute_at" primitive can't handle it.

This PR add "merge" schedule primitive to merge different loops, when cut same tiles on different out blocks by "split" primitive, we can merge these out block under an outer loop, then other blocks can "compute_at" this loop normally.

cc @Hzfengsy @wrongtest-intellif

@tvm-bot
Copy link
Collaborator

tvm-bot commented Mar 24, 2023

Thanks for contributing to TVM! Please refer to the contributing guidelines https://tvm.apache.org/docs/contribute/ for useful information and tips. Please request code reviews from Reviewers by @-ing them in a comment.

Generated by tvm-bot

@yincs-intellif yincs-intellif force-pushed the main branch 5 times, most recently from 7b9124b to 2c1bb3d Compare March 27, 2023 01:19
@yincs-intellif
Copy link
Contributor Author

@tvm-bot rerun

1 similar comment
@yincs-intellif
Copy link
Contributor Author

@tvm-bot rerun

@yincs-intellif
Copy link
Contributor Author

yincs-intellif commented Mar 29, 2023

@Hzfengsy Hi Siyuan, please help review this PR. Thanks very much!

@yincs-intellif
Copy link
Contributor Author

@tvm-bot rerun

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.

4 participants