llama.cpp
ggml-hexagon: gelu optimization
#18151
Merged

ggml-hexagon: gelu optimization #18151

joeldushouyu
joeldushouyu
joeldushouyu joeldushouyu changed the title Hexagon gelu optimization ggml-hexagon: gelu optimization 23 days ago
github-actions github-actions added ggml
chraac
chraac commented on 2025-12-18
max-krasnyansky
joeldushouyu
max-krasnyansky
joeldushouyu
joeldushouyu feat: working gelu with src0 put on vtcm
59c88691
joeldushouyu feat: gelu ping-pong for both in and out
423a7f80
joeldushouyu fix: fixu compile error
1c5ffc4b
joeldushouyu break: distinguish dma ddr->vtcm and vtcm->ddr operation
22161873
joeldushouyu fix: fix dma queue size
3a607f37
joeldushouyu break: update dma api to either pop src or dst ptr
29a46c32
joeldushouyu fix: fix activation vtcm allocation issue for src1 when swapperd
4032576e
joeldushouyu refactor: ping-pong gelu logic to avoid unnecessary if else
e9f13ba5
max-krasnyansky dma: improved queue interface and prefetch handling
dccbcb2c
max-krasnyansky
joeldushouyu joeldushouyu force pushed from f590092f to dccbcb2c 18 days ago
joeldushouyu
joeldushouyu joeldushouyu marked this pull request as ready for review 18 days ago
joeldushouyu joeldushouyu requested a review from lhez lhez 18 days ago
joeldushouyu joeldushouyu requested a review from max-krasnyansky max-krasnyansky 18 days ago
max-krasnyansky gelu: fix N+2 block prefetch
4ad1507d
max-krasnyansky
joeldushouyu
max-krasnyansky
max-krasnyansky approved these changes on 2025-12-22
max-krasnyansky max-krasnyansky merged bf6bc3c1 into master 18 days ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone