Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-26 06:10:29 +01:00
Code Issues Packages Projects Releases Wiki Activity

sycl : Reenabled mmvq path for the SYCL Nvidia Backend (#8372)

Browse Source 
* SYCL : Reenabled mmvq path for the SYCL Nvidia Backend

* Reduced verbosity of comment
This commit is contained in:
Alberto Cabrera Pérez 2024-07-09 15:03:15 +01:00 committed by GitHub
parent 9925ca4087
commit 5b0b8d8cfb
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-sycl.cpp
View File

@ -3658,6 +3658,10 @@ static void ggml_sycl_mul_mat(ggml_backend_sycl_context & ctx, const ggml_tensor
use_mul_mat_q = use_mul_mat_q && (src1->ne[1] <= MMQ_MAX_BATCH_SIZE);
#endif // SYCL_USE_XMX
// mmvq path is faster in the CUDA backend.
if (ctx.stream()->get_backend() == sycl::backend::ext_oneapi_cuda)
use_dequantize_mul_mat_vec = use_dequantize_mul_mat_vec && !use_mul_mat_vec_q;
if (!split && src0->type == GGML_TYPE_F16 && ggml_is_permuted(src0) && ggml_is_permuted(src1) && src1->ne[1] == 1) {
// KQ single-batch
ggml_sycl_mul_mat_vec_p021(ctx, src0, src1, dst);