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

Fixed mmap prefetch for GPU offloading (#2529)

Browse source
This commit is contained in:
Johannes Gäßler 2023-08-07 10:09:40 +02:00 committed by GitHub
parent f6f9896ac3
commit 3d9a551816
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 4 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
llama-util.h
View file

@ -219,7 +219,7 @@ struct llama_mmap {
// prefetch/readahead impairs performance on NUMA systems
if (numa) { prefetch = 0; }
#ifdef __linux__
if (prefetch) { flags |= MAP_POPULATE; }
if (prefetch >= file->size) { flags |= MAP_POPULATE; }
#endif
addr = mmap(NULL, file->size, PROT_READ, flags, fd, 0);
if (addr == MAP_FAILED) {

6
llama.cpp
View file

@ -747,12 +747,12 @@ struct llama_model_loader {
void load_all_data(llama_progress_callback progress_callback, void * progress_callback_user_data, llama_mlock * lmlock) {
size_t data_size = 0;
size_t prefetch_size = 0;
size_t prefetch_size = file_loader->file.size;
size_t lock_size = 0;
for (const llama_load_tensor & lt : tensors_map.tensors) {
data_size += lt.size;
if (lt.ggml_tensor->backend == GGML_BACKEND_CPU) {
prefetch_size += lt.size;
if (lt.ggml_tensor->backend != GGML_BACKEND_CPU) {
prefetch_size -= lt.size;
}
}