llama.cpp

Fix potential int8_t overflow in non-SIMD vec_dot

#986

Merged