CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-04 09:32:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

metal : disable concurrency optimization

Browse Source
This commit is contained in:
Georgi Gerganov
2023-09-18 18:00:01 +03:00
parent f015b26689
commit 86c90e34f5
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
llama.cpp
View File

@@ -6605,8 +6605,8 @@ struct llama_context * llama_new_context_with_model(
llama_free(ctx);
return NULL;
}
ggml_metal_graph_find_concurrency(ctx->ctx_metal, gf, false);
ggml_allocr_set_parse_seq(ctx->alloc, ggml_metal_get_concur_list(ctx->ctx_metal), ggml_metal_if_optimized(ctx->ctx_metal));
//ggml_metal_graph_find_concurrency(ctx->ctx_metal, gf, false);
//ggml_allocr_set_parse_seq(ctx->alloc, ggml_metal_get_concur_list(ctx->ctx_metal), ggml_metal_if_optimized(ctx->ctx_metal));
}
#endif
// measure memory requirements for the graph
@@ -6621,7 +6621,7 @@ struct llama_context * llama_new_context_with_model(
ctx->alloc = ggml_allocr_new(ctx->buf_alloc.data, ctx->buf_alloc.size, tensor_alignment);
#ifdef GGML_USE_METAL
if (ctx->ctx_metal) {
ggml_allocr_set_parse_seq(ctx->alloc, ggml_metal_get_concur_list(ctx->ctx_metal), ggml_metal_if_optimized(ctx->ctx_metal));
//ggml_allocr_set_parse_seq(ctx->alloc, ggml_metal_get_concur_list(ctx->ctx_metal), ggml_metal_if_optimized(ctx->ctx_metal));
}
#endif
#ifdef GGML_USE_CUBLAS