Repository navigation
hexagon: extend scalar DMA broadcasting support for binary ops - #30245
Merged
max-krasnyansky merged 9 commits intoOct 13, 2026
Merged
max-krasnyansky merged 9 commits into
max-krasnyansky merged 9 commits into
Conversation
github-actions
Bot
added
ggml
kurquhar
marked this pull request as ready for review
October 9, 2026 18:16
Member
|
Closing for a sec to undraft another PR |
kurquhar
force-pushed
the
kurquhar/hexagon-extended-scalar-dma-external
branch
from
October 13, 2026 01:18
3eda01b to
85ff470
Compare
Assisted-by: OpenCode
Assisted-by: OpenCode
Assisted-by: OpenCode
…sed them last time)
max-krasnyansky
force-pushed
the
kurquhar/hexagon-extended-scalar-dma-external
branch
from
October 13, 2026 03:05
85ff470 to
d86c81d
Compare
max-krasnyansky
approved these changes
Oct 13, 2026
Member
|
@lhez for another review and ack |
lhez
approved these changes
Oct 13, 2026
Member
|
@ggerganov @CISC we added a couple more tests here for binary ops that were not covered and cause issues in |
1 task done
edwardyoon
pushed a commit
to edwardyoon/focus-llama
that referenced
this pull request
Oct 24, 2026
…org#30245) * tests: cover weighted scalar broadcasts Assisted-by: OpenCode * ggml : support extended scalar DMA binary ops Assisted-by: OpenCode * hex-binary: remove all non-dma paths * hex-binary: add support for non-contig strides and cleanup dma preable a bit * hex-binary: cleanup the gating and simplify dma retry * hex-geglu-erf: add a gate for F32 only, we dont support F16 for this op * tests: cover weighted repeat broadcast Assisted-by: OpenCode * hex-binary: update REPEAT and COMPLEX kernels to use dma as well (missed them last time) * hex-add-id: add dma support for src2 (idx) --------- Co-authored-by: Max Krasnyansky <maxk@qti.qualcomm.com>
auRose94
pushed a commit
to auRose94/llama.cpp
that referenced
this pull request
Oct 28, 2026
…org#30245) * tests: cover weighted scalar broadcasts Assisted-by: OpenCode * ggml : support extended scalar DMA binary ops Assisted-by: OpenCode * hex-binary: remove all non-dma paths * hex-binary: add support for non-contig strides and cleanup dma preable a bit * hex-binary: cleanup the gating and simplify dma retry * hex-geglu-erf: add a gate for F32 only, we dont support F16 for this op * tests: cover weighted repeat broadcast Assisted-by: OpenCode * hex-binary: update REPEAT and COMPLEX kernels to use dma as well (missed them last time) * hex-add-id: add dma support for src2 (idx) --------- Co-authored-by: Max Krasnyansky <maxk@qti.qualcomm.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Overview
Extend scalar DMA broadcasting support on hexagon backend
Additional information
Hexagon does not currently support this op condition, but it is needed by stable-diffusion.cpp. There will be a future companion PR for the corresponding backend ops test support.
Specifically, FastWan has the following geometry for ADD:
src1 = [1,1,1,C] as bias WEIGHTS
src0 = [W,H,T,C]
Before this change:
After this change:
Verification:
Requirements
Assisted-by: OpenCode