llama.cpp
hexagon: Q4_0 and MXFP4 repack fixes
#20527
Merged
Overview
Commits
7
Changes
View On GitHub

hexagon: Q4_0 and MXFP4 repack fixes #20527

max-krasnyansky merged 7 commits into ggml-org:master from qualcomm:hexagon-repack-fixes
max-krasnyansky
max-krasnyansky hexagon: fix tail corruption with rows sizes not multiple of 256
4c33e4dd
max-krasnyansky hexagon: use different stride for repacking partial blocks
40fab2c2
max-krasnyansky hex-mm: update repack and kernels to avoid shuffles for full 256-elem…
47f5b864
max-krasnyansky hex-mm: update rmpy x8 for better optimizations
e445dbf1
max-krasnyansky hex-mm: tighten supported MUL_MAT checks to avoid spurios failures
27243fc3
max-krasnyansky hex-mm: use vzero to init accumulators
a769f7c2
max-krasnyansky max-krasnyansky requested a review from lhez lhez 20 days ago
github-actions github-actions added ggml
lhez
lhez approved these changes on 2026-03-14
max-krasnyansky hex-mm: properly call partial rmpy_x8
efea8560
max-krasnyansky max-krasnyansky merged 609ea500 into master 19 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
ggml
Milestone
No milestone
Terms Of Use Privacy Policy Refund Policy Impressum