CUDA: mul_mat_vec_q tiling, refactor mul mat logic #8029
Job | Run time |
---|---|
21m 49s | |
25m 48s | |
21m 30s | |
9m 20s | |
9m 52s | |
7m 57s | |
9m 4s | |
8m 31s | |
9m 1s | |
7m 43s | |
6m 37s | |
2h 17m 12s |
Job | Run time |
---|---|
21m 49s | |
25m 48s | |
21m 30s | |
9m 20s | |
9m 52s | |
7m 57s | |
9m 4s | |
8m 31s | |
9m 1s | |
7m 43s | |
6m 37s | |
2h 17m 12s |