summaryrefslogtreecommitdiff
path: root/ggml/src/ggml.c
diff options
context:
space:
mode:
authorKawrakow <48489457+ikawrakow@users.noreply.github.com>2024-09-14 19:47:26 +0300
committerGitHub <noreply@github.com>2024-09-14 19:47:26 +0300
commit76be98fdec39fed876803d54e83cb4d6f2b50f90 (patch)
tree95337447eea7060b9a7de98fd58a08ce23a9dcdd /ggml/src/ggml.c
parent064b99365c4426b83b09a518b29cc1ffc0250d04 (diff)
Improve Q5_0 performance (#55)
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 060d10c6..08b292b7 100644
--- a/ggml/src/ggml.c
+++ b/ggml/src/ggml.c
@@ -767,7 +767,11 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = {
.from_float = quantize_row_q5_0,
.from_float_ref = (ggml_from_float_t) quantize_row_q5_0_ref,
.vec_dot = ggml_vec_dot_q5_0_q8_0,
+#if GGML_USE_IQK_MULMAT && defined __AVX2__
+ .vec_dot_type = GGML_TYPE_Q8_1,
+#else
.vec_dot_type = GGML_TYPE_Q8_0,
+#endif
.nrows = 1,
},
[GGML_TYPE_Q5_1] = {