Example throughput charts¶

Bar charts for examples/README.md: MiB/s, before vs pyuring.

Workloads are many small/medium files (thread pool or asyncio.gather+executor vs batched io_uring), where batching usually favors pyuring. Regenerate with:

pip install numpy

PYTHONPATH=. python3 scripts/gen_example_graphs.py

PyTorch-style (many shards)¶

pytorch

asyncio (many files)¶

asyncio

FastAPI (many on-disk reads per batch)¶

fastapi

NumPy shards (`numpy_bins`)¶

numpy_bins

Cached blob split into parts (`cached_reads`)¶

cached_reads

SQLite-export-style shard files (`sqlite_blobs`)¶

sqlite_blobs