vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

parallelize fattn compilation test

Browse source
This commit is contained in:
slaren 2024-05-28 01:19:36 +02:00
parent f4003cfba1
commit 1ca802a3e0
6 changed files with 390 additions and 288 deletions
Download patch file Download diff file Expand all files Collapse all files

7
ggml-cuda/fattn-vec-f16-f16.cu Normal file
View file

@ -0,0 +1,7 @@
#include "common.cuh"
#include "fattn-common.cuh"
#include "fattn-vec-f16.cuh"
DECL_FATTN_VEC_F16_INST(64, 1, 4, (vec_dot_fattn_vec_KQ_f16<half, 64>), false, dequantize_1_f16<half>);
DECL_FATTN_VEC_F16_INST(128, 1, 4, (vec_dot_fattn_vec_KQ_f16<half, 128>), false, dequantize_1_f16<half>);
DECL_FATTN_VEC_F16_INST(256, 1, 4, (vec_dot_fattn_vec_KQ_f16<half, 256>), false, dequantize_1_f16<half>);