nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 27 Packages Projects Releases Wiki Activity

llama : Extend fallback, fix fileno for dio file, exclude case that mmap uses dio file (#18887)

Browse Source
This commit is contained in:
Julius Tischbein
2026-01-18 18:35:57 +02:00
committed by GitHub
parent 293a1565dc
commit 287a33017b
2 changed files with 17 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-mmap.cpp
+5 -1
View File
@@ -265,7 +265,8 @@ struct llama_file::impl {
continue; // Interrupted by signal, retry continue; // Interrupted by signal, retry
} }
// Fallback to std::fread in case the DMA controller cannot access the buffer // Fallback to std::fread in case the DMA controller cannot access the buffer
if (errno == EFAULT) { if (errno == EFAULT || errno == EINVAL) {
LLAMA_LOG_WARN("%s: Falling back to buffered IO due to %s\n", __func__, strerror(errno));
auto curr_off = tell(); auto curr_off = tell();
close(fd); close(fd);
fd = -1; fd = -1;
@@ -384,6 +385,9 @@ int llama_file::file_id() const {
#ifdef _WIN32 #ifdef _WIN32
return _fileno(pimpl->fp); return _fileno(pimpl->fp);
#else #else
if (pimpl->fd != -1) {
return pimpl->fd;
}
#if defined(fileno) #if defined(fileno)
return fileno(pimpl->fp); return fileno(pimpl->fp);
#else #else
src/llama-model-loader.cpp
+10 -4
View File
@@ -539,12 +539,18 @@ llama_model_loader::llama_model_loader(
files.emplace_back(new llama_file(fname.c_str(), "rb", use_direct_io)); files.emplace_back(new llama_file(fname.c_str(), "rb", use_direct_io));
contexts.emplace_back(ctx); contexts.emplace_back(ctx);
use_direct_io = use_direct_io && files.back()->has_direct_io(); if (use_mmap && use_direct_io) {
if (files.back()->has_direct_io()) {
// Disable mmap in case Direct I/O is enabled and available // Disable mmap, as DirectIO is available
if (use_direct_io && use_mmap) {
use_mmap = false; use_mmap = false;
LLAMA_LOG_WARN("%s: direct I/O is enabled, disabling mmap\n", __func__); LLAMA_LOG_WARN("%s: direct I/O is enabled, disabling mmap\n", __func__);
} else {
// Disable DirectIO and reopen file using std::fopen for mmap
use_direct_io = false;
files.pop_back();
files.emplace_back(new llama_file(fname.c_str(), "rb", false));
LLAMA_LOG_WARN("%s: direct I/O is not available, using mmap\n", __func__);
}
} }
// Save tensors data offset of the main file. // Save tensors data offset of the main file.