happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
llama.cpp/ggml
History
Gong-Mi e0f3a6b292 vulkan: optimize Mali and Adreno tuning & fix FP16/matrix stability 
- Add VK_VENDOR_ID_ARM (0x13B5) and VK_VENDOR_ID_QUALCOMM (0x5143).
    - Disable 'l_warptile' (Large) paths on ARM and Qualcomm to avoid register pressure/driver crashes.
    - Explicitly disable cooperative matrix (fp16-matrix) on mobile GPUs to prevent crashes.
    - Disable integer dot product (int8-matrix) on Qualcomm due to driver issues.
    - Keep general FP16 enabled for large model support.
    - Limit suballocation block size to 256MB to improve memory stability on Android.
 		2026-01-23 20:29:33 +08:00
..
cmake ggml: Skip backend library linking code when GGML_BACKEND_DL=ON (#15094) 2025-08-07 13:45:41 +02:00
include ggml : add ggml_build_forward_select (#18550) 2026-01-19 20:03:19 +02:00
src vulkan: optimize Mali and Adreno tuning & fix FP16/matrix stability 2026-01-23 20:29:33 +08:00
.gitignore vulkan : cmake integration (#8119) 2024-07-13 18:12:39 +02:00
CMakeLists.txt ggml : bump version to 0.9.5 (ggml/1410) 2025-12-31 18:54:43 +02:00