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

test pp_threads

Browse source
This commit is contained in:
netrunnereve 2023-08-05 22:39:44 -04:00
parent f514d1b306
commit 5f022185a1
1 changed files with 4 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

5
llama.cpp
View file

@ -1814,9 +1814,12 @@ static bool llama_eval_internal(
// fprintf(stderr, "graph build time: %.3f ms (%d nodes, %d leafs)\n", (ggml_time_us() - t_start_us)/1000.0, gf->n_nodes, gf->n_leafs);
int32_t pp_threads = 3;
// for big prompts, if BLAS is enabled, it is better to use only one thread
// otherwise, the threads are spin-lock waiting for the BLAS calls and are degrading the performance
n_threads = N >= 32 && ggml_cpu_has_blas() && !ggml_cpu_has_gpublas() ? 1 : n_threads;
pp_threads = N >= 32 && ggml_cpu_has_blas() && !ggml_cpu_has_gpublas() ? 1 : pp_threads;
n_threads = N > 1 ? pp_threads : n_threads;
struct ggml_tensor * res = gf->nodes[gf->n_nodes - 1];
struct ggml_tensor * embeddings = gf->nodes[gf->n_nodes - 2];