Skip to content

Pull requests: ROCm/llama.cpp

New pull request New
11 Open 34 Closed
11 Open 34 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

roofline: emit raw per-tensor facts for DDR byte accounting (single + fused)
#48 opened Jul 13, 2026 by roberteg16 • Draft
ci(gfx11): cache HIP builds with ggml-org/ccache-actio
#45 opened Jul 13, 2026 by roberteg16 Loading…
2 of 3 tasks
3
ggml-cuda: use mma-f16 flash-attn for D=256 on RDNA3.5
#40 opened Jul 8, 2026 by roberteg16 Loading…
ggml-cuda: one-time K-padded f16 dequantization for prefill GEMMs
#37 opened Jul 7, 2026 by roberteg16 • Draft
ggml-cuda: overlap the MoE shared expert on a separate stream
#36 opened Jul 6, 2026 by roberteg16 Loading…
14
ggml-cuda: pad GEMM weight rows to avoid cache-set aliasing
#34 opened Jul 3, 2026 by roberteg16 • Draft
4 tasks done
1
RDNA3.5 (gfx11 / gfx1151) MMQ prefill optimizations
#32 opened Jun 30, 2026 by liangliangchang • Draft
4 tasks done
5
RDNA 3.5 MMVQ/fattn optimizations for gfx1151
#31 opened Jun 26, 2026 by jeffli-xilinx • Draft
5 tasks
Gfx11
#29 opened Jun 26, 2026 by liangliangchang • Draft
Basic CI for hrx-v2
#27 opened Jun 18, 2026 by AaronStGeorge Collaborator Loading…
Updates for new hrx C and CMake APIs
#26 opened Jun 18, 2026 by AaronStGeorge Collaborator Loading…
ProTip! Add no:assignee to see everything that’s not assigned.