diff options
| author | Howard Su <howard0su@gmail.com> | 2023-07-13 21:58:25 +0800 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2023-07-13 21:58:25 +0800 |
| commit | 32c54116318929c90fd7ae814cf9b5232cd44c36 (patch) | |
| tree | 3b9126e3fb387ef1aa53d7461f9a41e1ce2965ed /llama-util.h | |
| parent | ff5d58faecf1f02b05bd015bdfc6a394cf2bc9ba (diff) | |
Revert "Support using mmap when applying LoRA (#2095)" (#2206)
Has perf regression when mlock is used.
This reverts commit 2347463201a9f4159ae95b737e1544dd300569c8.
Diffstat (limited to 'llama-util.h')
| -rw-r--r-- | llama-util.h | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/llama-util.h b/llama-util.h index 43b6f05..042ebe4 100644 --- a/llama-util.h +++ b/llama-util.h @@ -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); |