happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Fixed typo in q4_1 kernel

This commit is contained in:
iacopPBK 2026-03-30 19:32:12 +02:00
parent 0bcddd2164
commit 5d7df5df44
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-cuda/mmq.cuh
View File

@ -492,7 +492,7 @@ static __device__ __forceinline__ void vec_dot_q4_1_q8_1_dp4a(
const int i = i0 + threadIdx.x;
const int kyqs = QI8_1 * ((k01/2) / (QI8_1/2)) + (k01/2) % (QI8_1/2);
int u[2*VDR_Q4_0_Q8_1_MMQ];
int u[2*VDR_Q4_1_Q8_1_MMQ];
constexpr int max_cpy = ggml_cuda_get_max_cpy_bytes();
constexpr int mcpy_int = max_cpy / sizeof(int);