Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[DispatchCreation] Move concat decomposition to Dispatch creation and only decompose non-outer dim concats #19148

Draft
wants to merge 3 commits into
base: main
Choose a base branch
from

Conversation

MaheshRavishankar
Copy link
Contributor

Move the pass to decompose concats just before dispatch region formation, and only decompose concats that are not along the outer-most dimension. This could allow folding these insert-slices into their producers.

Outer dim concats are not decomposed here. They are left till flow conversion. Decomposing those into flow.tensor.update allows stream allocation to do the concat in-place.

Fixes #19092

Depends on #19126 and llvm/llvm-project#116004

@MaheshRavishankar MaheshRavishankar force-pushed the concat_to_flow_move_pass branch 3 times, most recently from 78d1bdc to b9ebc4a Compare November 14, 2024 02:38
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.

tensor.cat of dynamic length tensors always results in slow memcopys
1 participant