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

gguf : avoid too many file size calls (#19919)

Browse Source
This commit is contained in:
Georgi Gerganov
2026-02-26 12:46:32 +02:00
committed by GitHub
parent bd72300591
commit 1ca3d1de15
1 changed files with 47 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/gguf.cpp
+47 -30
View File
@@ -228,13 +228,41 @@ struct gguf_context {
}; };
struct gguf_reader { struct gguf_reader {
FILE * file; gguf_reader(FILE * file) : file(file) {
// read the remaining bytes once and update on each read
nbytes_remain = file_remain(file);
}
gguf_reader(FILE * file) : file(file) {} // helper for remaining bytes in a file
static uint64_t file_remain(FILE * file) {
const int64_t cur = gguf_ftell(file);
if (cur < 0) {
return 0;
}
if (gguf_fseek(file, 0, SEEK_END) != 0) {
gguf_fseek(file, cur, SEEK_SET);
return 0;
}
const int64_t end = gguf_ftell(file);
if (end < 0) {
gguf_fseek(file, cur, SEEK_SET);
return 0;
}
gguf_fseek(file, cur, SEEK_SET);
return static_cast<uint64_t>(end - cur);
}
template <typename T> template <typename T>
bool read(T & dst) const { bool read(T & dst) const {
return fread(&dst, 1, sizeof(dst), file) == sizeof(dst); const size_t size = sizeof(dst);
if (nbytes_remain < size) {
return false;
}
const size_t nread = fread(&dst, 1, size, file);
nbytes_remain -= nread;
return nread == size;
} }
template <typename T> template <typename T>
@@ -242,20 +270,19 @@ struct gguf_reader {
if (n > GGUF_MAX_ARRAY_ELEMENTS) { if (n > GGUF_MAX_ARRAY_ELEMENTS) {
return false; return false;
} }
const uint64_t nbytes = nbytes_remain();
if constexpr (std::is_same<T, std::string>::value) { if constexpr (std::is_same<T, std::string>::value) {
// strings are prefixed with their length, so we need to account for that // strings are prefixed with their length, so we need to account for that
if (n > SIZE_MAX / sizeof(uint64_t)) { if (n > SIZE_MAX / sizeof(uint64_t)) {
return false; return false;
} }
if (nbytes < n * sizeof(uint64_t)) { if (nbytes_remain < n * sizeof(uint64_t)) {
return false; return false;
} }
} else { } else {
if (n > SIZE_MAX / sizeof(T)) { if (n > SIZE_MAX / sizeof(T)) {
return false; return false;
} }
if (nbytes < n * sizeof(T)) { if (nbytes_remain < n * sizeof(T)) {
return false; return false;
} }
} }
@@ -312,39 +339,29 @@ struct gguf_reader {
GGML_LOG_ERROR("%s: string length %" PRIu64 " exceeds maximum %" PRIu64 "\n", __func__, size, (uint64_t) GGUF_MAX_STRING_LENGTH); GGML_LOG_ERROR("%s: string length %" PRIu64 " exceeds maximum %" PRIu64 "\n", __func__, size, (uint64_t) GGUF_MAX_STRING_LENGTH);
return false; return false;
} }
const uint64_t nbytes = nbytes_remain(); if (size > nbytes_remain) {
if (size > nbytes) { GGML_LOG_ERROR("%s: string length %" PRIu64 " exceeds remaining file size %" PRIu64 " bytes\n", __func__, size, nbytes_remain);
GGML_LOG_ERROR("%s: string length %" PRIu64 " exceeds remaining file size %" PRIu64 " bytes\n", __func__, size, nbytes);
return false; return false;
} }
dst.resize(static_cast<size_t>(size)); dst.resize(static_cast<size_t>(size));
return fread(dst.data(), 1, dst.length(), file) == dst.length(); const size_t nread = fread(dst.data(), 1, size, file);
nbytes_remain -= nread;
return nread == size;
} }
bool read(void * dst, const size_t size) const { bool read(void * dst, const size_t size) const {
return fread(dst, 1, size, file) == size; if (size > nbytes_remain) {
return false;
}
const size_t nread = fread(dst, 1, size, file);
nbytes_remain -= nread;
return nread == size;
} }
// remaining bytes in the file private:
uint64_t nbytes_remain() const { FILE * file;
const int64_t cur = gguf_ftell(file);
if (cur < 0) {
return 0;
}
if (gguf_fseek(file, 0, SEEK_END) != 0) {
gguf_fseek(file, cur, SEEK_SET);
return 0; mutable uint64_t nbytes_remain;
}
const int64_t end = gguf_ftell(file);
if (end < 0) {
gguf_fseek(file, cur, SEEK_SET);
return 0;
}
gguf_fseek(file, cur, SEEK_SET);
return static_cast<uint64_t>(end - cur);
}
}; };
struct gguf_context * gguf_init_empty(void) { struct gguf_context * gguf_init_empty(void) {