summaryrefslogtreecommitdiff
path: root/ggml/src/ggml.c
diff options
context:
space:
mode:
authorKawrakow <iwankawrakow@gmail.com>2025-06-12 08:16:12 +0300
committerGitHub <noreply@github.com>2025-06-12 08:16:12 +0300
commit4fc3cb4a4709e747984f7b8c9097d9b997d949b5 (patch)
tree451208c6ee091a7f4de1fe21e9b839b0ac73cff1 /ggml/src/ggml.c
parent3f54b4978672cc236bf38e707b24f036cb37f3f7 (diff)
iq3_s: much faster GEMM via repacking to q8_0_r8 (#518)
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'ggml/src/ggml.c')
-rw-r--r--ggml/src/ggml.c4
1 files changed, 4 insertions, 0 deletions
diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c
index e4e954a1..3953cd7d 100644
--- a/ggml/src/ggml.c
+++ b/ggml/src/ggml.c
@@ -1153,7 +1153,11 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = {
.from_float = quantize_row_iq3_s,
.from_float_ref = (ggml_from_float_t)quantize_row_iq3_s_ref,
.vec_dot = ggml_vec_dot_iq3_s_q8_K,
+#ifdef __AVX2__
+ .vec_dot_type = GGML_TYPE_Q8_2_X4,
+#else
.vec_dot_type = GGML_TYPE_Q8_K,
+#endif
.nrows = 1,
.row_meta_size = 0,
},