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

metal : reduce command encoding overhead (#9698)

Browse source 
* metal : reduce command encoding overhead

ggml-ci

* metal : add comments
This commit is contained in:
Georgi Gerganov 2024-10-01 16:00:25 +03:00 • committed by GitHub
parent a90484c6d9
commit cad341d889
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
5 changed files with 2000 additions and 1912 deletions
Download patch file Download diff file Expand all files Collapse all files

6
src/llama.cpp
View file

@ -17025,12 +17025,6 @@ static void llama_graph_compute(
ggml_cgraph * gf,
int n_threads,
ggml_threadpool * threadpool) {
#ifdef GGML_USE_METAL
if (ggml_backend_is_metal(lctx.backend_metal)) {
ggml_backend_metal_set_n_cb(lctx.backend_metal, n_threads);
}
#endif
if (lctx.backend_cpu != nullptr) {
ggml_backend_cpu_set_n_threads(lctx.backend_cpu, n_threads);
ggml_backend_cpu_set_threadpool(lctx.backend_cpu, threadpool);