diff options
| author | Kawrakow <iwankawrakow@gmail.com> | 2024-12-08 15:27:13 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-12-08 15:27:13 +0100 |
| commit | 43e65a672a98d931998559785b58f1e980e87f54 (patch) | |
| tree | f0cefd9605710ff529d637e507af01327792d004 /ggml/include | |
| parent | fc701cedd146152fb270482a7eef5aba23b20575 (diff) | |
Faster IQ4_XS_R4 on Zen4 (#128)
* Faster iq4_xs_r4 on Zen4
The trick is to simply prepare the Q8 block sums for
blocks of 32 as floats. This brings PP-512 up to 254.6 t/s
from 224 t/s.
* Fix broken matrix x vector product on Zen4
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'ggml/include')
| -rw-r--r-- | ggml/include/ggml.h | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/ggml/include/ggml.h b/ggml/include/ggml.h index e17540e9..bb50a025 100644 --- a/ggml/include/ggml.h +++ b/ggml/include/ggml.h @@ -407,6 +407,7 @@ extern "C" { GGML_TYPE_IQ2_KS = 145, GGML_TYPE_IQ4_KSS = 146, GGML_TYPE_Q8_K16 = 147, + GGML_TYPE_Q8_K32 = 148, GGML_TYPE_Q4_0_R4 = 202, GGML_TYPE_Q5_0_R4 = 206, |