From 43e65a672a98d931998559785b58f1e980e87f54 Mon Sep 17 00:00:00 2001 From: Kawrakow Date: Sun, 8 Dec 2024 15:27:13 +0100 Subject: 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 --- ggml/include/ggml.h | 1 + 1 file changed, 1 insertion(+) (limited to 'ggml/include/ggml.h') 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, -- cgit v1.2.3