happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

add server memory debug logging

This commit is contained in:
Ruben Ortlam 2026-04-02 10:07:04 +02:00
parent 4af1a283a6
commit 7e10ec8ff2
1 changed files with 15 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
tools/server/server-models.cpp
View File

@ -179,7 +179,7 @@ server_models::server_models(
LOG_WRN("using original argv[0] as fallback: %s\n", argv[0]);
}
const uint64_t memory_margin = base_params.models_memory_margin * 1024 * 1024;
const uint64_t memory_margin = (uint64_t)base_params.models_memory_margin * 1024 * 1024;
if (memory_margin > 0) {
const size_t n_devs = ggml_backend_dev_count();
@ -188,7 +188,11 @@ server_models::server_models(
size_t free, total;
ggml_backend_dev_memory(dev, &free, &total);
if (total > 0) {
memory_per_device[dev] = (free > memory_margin) ? free - memory_margin : 0;
const uint64_t available = (free > memory_margin) ? free - memory_margin : 0;
memory_per_device[dev] = available;
SRV_DBG("device %s: available memory after margin=%lu MB\n",
ggml_backend_dev_name(dev),
(unsigned long)(available / (1024 * 1024)));
}
}
}
@ -527,7 +531,15 @@ uint64_t server_models::get_memory_exceeded(const model_memory_map& new_model_me
uint64_t memory_exceeded = 0;
for (const auto& [key, limit] : memory_per_device) {
if (get(new_model_memory_per_device, key) + get(total_memory_per_device, key) > limit) {
const uint64_t total_memory = get(total_memory_per_device, key);
const uint64_t new_memory = get(new_model_memory_per_device, key);
SRV_DBG("device %s: total=%lu MB, new=%lu MB, limit=%lu MB\n",
ggml_backend_dev_name(key),
(unsigned long)(total_memory / (1024 * 1024)),
(unsigned long)(new_memory / (1024 * 1024)),
(unsigned long)(limit / (1024 * 1024)));
if (total_memory + new_memory > limit) {
memory_exceeded++;
}
}