Skip to content

tests : fix test-thread-safety when compiling with multiple backends #16699

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Oct 22, 2025
Merged

tests : fix test-thread-safety when compiling with multiple backends #16699

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 6 additions & 4 deletions tests/test-thread-safety.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
// - Creates n_parallel (--parallel) contexts per model
// - Runs inference in parallel on each context

#include <array>
#include <thread>
#include <vector>
#include <atomic>
Expand Down Expand Up @@ -38,13 +39,14 @@ int main(int argc, char ** argv) {
cparams.n_seq_max = 1;

int dev_count = ggml_backend_dev_count();
int gpu_dev_count = 0;
std::vector<std::array<ggml_backend_dev_t, 2>> gpus;
for (int i = 0; i < dev_count; ++i) {
auto * dev = ggml_backend_dev_get(i);
if (dev && ggml_backend_dev_type(dev) == GGML_BACKEND_DEVICE_TYPE_GPU) {
gpu_dev_count++;
gpus.push_back({dev, nullptr});
}
}
const int gpu_dev_count = (int)gpus.size();
const int num_models = gpu_dev_count + 1 + 1; // GPUs + 1 CPU model + 1 layer split
//const int num_models = std::max(1, gpu_dev_count);
const int num_contexts = std::max(1, params.n_parallel);
Expand All @@ -58,12 +60,12 @@ int main(int argc, char ** argv) {

if (m < gpu_dev_count) {
mparams.split_mode = LLAMA_SPLIT_MODE_NONE;
mparams.main_gpu = m;
mparams.devices = gpus[m].data();
} else if (m == gpu_dev_count) {
mparams.split_mode = LLAMA_SPLIT_MODE_NONE;
mparams.main_gpu = -1; // CPU model
} else {
mparams.split_mode = LLAMA_SPLIT_MODE_LAYER;;
mparams.split_mode = LLAMA_SPLIT_MODE_LAYER;
}

llama_model * model = llama_model_load_from_file(params.model.path.c_str(), mparams);
Expand Down
Loading