tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-24 08:49:15 -04:00
Code Issues 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
2 changed files with 4 additions and 4 deletions
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) {