diff options
| author | Iwan Kawrakow <iwan.kawrakow@gmail.com> | 2024-06-08 09:02:23 +0300 |
|---|---|---|
| committer | Iwan Kawrakow <iwan.kawrakow@gmail.com> | 2024-06-22 12:02:50 +0300 |
| commit | cd3d8ae0e719b47fb0ef63b0f7b9e1dacbab7de1 (patch) | |
| tree | 5ea228ac8f54d743889a2df699af65b396d9048e /ggml.c | |
| parent | 299c7f6e89d2d8c4162be06463b82a07540d5691 (diff) | |
iqk_mul_mat: use block_q8_1_x4 also for AVX2
Here the performance gain is more significant. E.g., for q4_1,
PP-512 becomes 168 t/s up from 137 t/s.
Now the performance gap to q4_0 is so significant that I
wonder if I should change to using Q8_1 also for the
qX_0 legacy quants.
Diffstat (limited to 'ggml.c')
0 files changed, 0 insertions, 0 deletions