llama.cpp
ggml-cuda: assign four GDN state columns per warp
#30087
Merged

ggml-cuda: assign four GDN state columns per warp #30087

SongXiaoXi
SongXiaoXi ggml-cuda: assign two GDN state columns per warp
799870f9
SongXiaoXi SongXiaoXi requested a review 42 days ago
github-actions github-actions added ggml
github-actions github-actions added CUDA
SongXiaoXi
am17an
am17an commented on 2026-10-07
am17an am17an assigned am17an am17an 42 days ago
SongXiaoXi
am17an
SongXiaoXi
am17an
SongXiaoXi ggml-cuda: use 4 GDN state columns per warp at S_v=128
80af7592
am17an
am17an commented on 2026-10-07
SongXiaoXi ggml-cuda: default cols_per_warp=4
a82115fe
am17an
am17an approved these changes on 2026-10-08
am17an am17an added merge ready
SongXiaoXi SongXiaoXi changed the title ggml-cuda: assign two GDN state columns per warp ggml-cuda: assign four GDN state columns per warp 41 days ago
am17an
ServeurpersoCom ServeurpersoCom requested a review from ServeurpersoCom ServeurpersoCom 41 days ago
ServeurpersoCom
ServeurpersoCom
ServeurpersoCom approved these changes on 2026-10-08
SongXiaoXi ggml-cuda: address GDN review nits
d6c7e29e
SongXiaoXi SongXiaoXi force pushed to d6c7e29e 41 days ago
am17an
am17an approved these changes on 2026-10-08
am17an am17an merged 24e41838 into master 41 days ago
am17an

Login to write a write a comment.

Login via GitHub

Assignees
Labels
Milestone