llama.cpp

66645a52 - SYCL: Disable mul_mat kernels for noncontiguous tensor b (#13308)