Skip to main content

Examples

Each example under examples/ is a self-contained workload with the same structure: run_bench.py (or run_bench.sh) builds the plugin and runs the graph; verify.py / verify.sh checks correctness (examples/README.md). The example code is not on this site — the links go to the repository.

Capability matrix

ExamplePlugin languageDAG featuresVerifyExternal deps
matrix-computeRust (nalgebra, rustfft)linear chain, shared initializationverify.shnone
matrix-compute-CC (FFTW, OpenBLAS)same topology, C pluginmake validationlibfftw3f, libopenblas
matrix-compute-pythonPython (NumPy)same topology, GIL / free-threaded demouv, NumPy
stream-analyticsRustconditional branches, grouped barriers, $dep ordering, prioritiesverify.pynone
mapreduceCfan-out / fan-in, variadic barrierverify.shnone (GCC)
gpu-vectoraddCUDA C++use_workers GPU-thread pinning, host↔device copiesinlineCUDA GPU
agent-tuning— (harness)4-arm verifier-gated knob searchper-workload gateOptuna, claude CLI
scheduler-pluginRustpluggable TaskScheduler (FIFO)none

Running them

All commands run from the repository root with the tomii package installed (examples/README.md):

python examples/matrix-compute/run_bench.py # Rust plugin
python examples/matrix-compute-C/run_bench.py # C plugin
bash examples/matrix-compute-python/run_bench.sh # Python plugin
python examples/stream-analytics/run_bench.py # conditional branching
python examples/mapreduce/run_bench.py # Map→Reduce wordcount
python examples/gpu-vectoradd/run_bench.py # CUDA (GPU required)

Where each one is covered in this guide

Comparator benchmarks against Taskflow and TBB live under bench/, not examples/ — see the benchmarks page for methodology and measured numbers.