mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2025-06-28 20:25:20 +00:00
@ -7641,8 +7641,8 @@ UseGgmlGemm2:;
|
||||
// dot kernels can handle 1 row and col at a time, but mmla kernels can process 2 rows and cols
|
||||
int64_t num_rows_per_vec_dot = vec_dot_num_rows;
|
||||
|
||||
// TODO: currently the mmla kernels support only even numbered rows/cols.
|
||||
// this check can be removed once they are extended to support odd numbered rows/cols too
|
||||
// these checks are needed to avoid crossing dim1 boundaries
|
||||
// can be optimized, but the logic would become more complicated, so keeping it like this for simplicity
|
||||
if ((nr0 % 2 != 0) || (ne11 % 2 != 0) || ((ir0_end - ir0_start) % 2 != 0) || ((ir1_end - ir1_start) % 2 != 0)) {
|
||||
num_rows_per_vec_dot = 1;
|
||||
}
|
||||
|
Reference in New Issue
Block a user