CUDA: mul_mat_vec_q tiling, refactor mul mat logic #8498
Job | Run time |
---|---|
6m 39s | |
1m 13s | |
5m 19s | |
25s | |
17s | |
6m 31s | |
5m 31s | |
1s | |
15s | |
5m 8s | |
2m 0s | |
2m 2s | |
1s | |
7m 23s | |
3m 30s | |
4m 47s | |
2m 2s | |
3m 54s | |
21m 40s | |
2m 17s | |
9m 6s | |
3m 12s | |
9m 28s | |
7m 52s | |
12m 33s | |
7m 19s | |
2m 20s | |
5m 13s | |
14m 6s | |
4m 21s | |
3m 55s | |
3m 23s | |
0s | |
2h 43m 43s |