Repository navigation
hexagon: add FP32 GELU_ERF and GEGLU_ERF support - #29631
Merged
max-krasnyansky merged 3 commits intoSep 29, 2026
Merged
Conversation
github-actions
Bot
added
ggml
CISC
approved these changes
Sep 29, 2026
Member
|
Just retriggering CI. |
Member
|
Ah, nvm, rebase please. |
max-krasnyansky
approved these changes
Sep 29, 2026
5 tasks done
1 task
edwardyoon
pushed a commit
to edwardyoon/focus-llama
that referenced
this pull request
Oct 7, 2026
* hexagon: add FP32 GELU_ERF and GEGLU_ERF support * hex-erf: reduce register pressure in kernels (cherry picked from commit e904318)
2 of 5 tasks
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
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Add FP32 Hexagon HTP support for GELU_ERF and GEGLU_ERF. GELU_ERF is required by the Qwen3 0.6B ASR model's CLIP graph. GEGLU_ERF is added alongside it for completeness.
Additional information
The implementation uses an HVX vectorized erf approximation with boundary handling for numerical stability. Inputs outside [-10, 10] use GELU's saturated behavior to avoid problematic intermediate values.
Requirements