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

Revert "Support using mmap when applying LoRA (#2095)" (#2206)

Browse source 
Has perf regression when mlock is used.

This reverts commit 2347463201.
This commit is contained in:
Howard Su 2023-07-13 21:58:25 +08:00 • committed by GitHub
parent ff5d58faec
commit 32c5411631
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
5 changed files with 9 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

6
llama-util.h
View file

@ -175,13 +175,13 @@ struct llama_mmap {
llama_mmap(struct llama_file * file, size_t prefetch = (size_t) -1 /* -1 = max value */, bool numa = false) {
size = file->size;
int fd = fileno(file->fp);
int flags = MAP_PRIVATE;
int flags = MAP_SHARED;
// prefetch/readahead impairs performance on NUMA systems
if (numa) { prefetch = 0; }
#ifdef __linux__
if (prefetch) { flags |= MAP_POPULATE; }
#endif
addr = mmap(NULL, file->size, PROT_READ | PROT_WRITE, flags, fd, 0);
addr = mmap(NULL, file->size, PROT_READ, flags, fd, 0);
if (addr == MAP_FAILED) {
throw std::runtime_error(format("mmap failed: %s", strerror(errno)));
}
@ -223,7 +223,7 @@ struct llama_mmap {
throw std::runtime_error(format("CreateFileMappingA failed: %s", llama_format_win_err(error).c_str()));
}
addr = MapViewOfFile(hMapping, FILE_MAP_COPY, 0, 0, 0);
addr = MapViewOfFile(hMapping, FILE_MAP_READ, 0, 0, 0);
error = GetLastError();
CloseHandle(hMapping);