summaryrefslogtreecommitdiff
path: root/ggml/src/ggml.c
diff options
context:
space:
mode:
Diffstat (limited to 'ggml/src/ggml.c')
-rw-r--r--ggml/src/ggml.c25
1 files changed, 15 insertions, 10 deletions
diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c
index 69b1b46d..cc056f89 100644
--- a/ggml/src/ggml.c
+++ b/ggml/src/ggml.c
@@ -1596,10 +1596,10 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = {
.from_float = quantize_row_iq2_kt,
.from_float_ref = (ggml_from_float_t)quantize_row_iq2_kt_ref,
.vec_dot = vec_dot_iq2_kt_q8_k,
-#ifdef __ARM_NEON
- .vec_dot_type = GGML_TYPE_F16,
+#if defined __AVX2__
+ .vec_dot_type = GGML_TYPE_Q8_2_X4,
#else
- .vec_dot_type = GGML_TYPE_F32,
+ .vec_dot_type = GGML_TYPE_Q8_0_X4,
#endif
.nrows = 1,
.row_meta_size = 4,
@@ -1613,11 +1613,16 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = {
.from_float = quantize_row_iq3_kt,
.from_float_ref = (ggml_from_float_t)quantize_row_iq3_kt_ref,
.vec_dot = vec_dot_iq3_kt_q8_k,
-#ifdef __ARM_NEON
- .vec_dot_type = GGML_TYPE_F16,
+#if defined __AVX2__
+ .vec_dot_type = GGML_TYPE_Q8_2_X4,
#else
- .vec_dot_type = GGML_TYPE_F32,
+ .vec_dot_type = GGML_TYPE_Q8_0_X4,
#endif
+//#ifdef __ARM_NEON
+// .vec_dot_type = GGML_TYPE_F16,
+//#else
+// .vec_dot_type = GGML_TYPE_F32,
+//#endif
.nrows = 1,
.row_meta_size = 4,
},
@@ -1630,13 +1635,13 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = {
.from_float = quantize_row_iq4_kt,
.from_float_ref = (ggml_from_float_t)quantize_row_iq4_kt_ref,
.vec_dot = vec_dot_iq4_kt_q8_k,
-#ifdef __ARM_NEON
- .vec_dot_type = GGML_TYPE_F16,
+#if defined __AVX2__
+ .vec_dot_type = GGML_TYPE_Q8_2_X4,
#else
- .vec_dot_type = GGML_TYPE_F32,
+ .vec_dot_type = GGML_TYPE_Q8_0_X4,
#endif
.nrows = 1,
- .row_meta_size = 8,
+ .row_meta_size = 4,
},
[GGML_TYPE_IQ3_K] = {
.type_name = "iq3_k",