llama.cpp
vulkan: disable large matmul tile on Samsung GPUs with 32KB shared memory
#28531
Merged

Loading