veloxloop 0.2.0

A modern, high-performance asyncio event loop implementation written from scratch in Rust

VeloxLoop

VeloxLoop — A modern, high-performance asyncio event loop implementation written from scratch in Rust using PyO3, tokio and io-uring crate.

Velox is Latin for "swift" or "rapid" — reflecting the goal of delivering significantly faster I/O and lower overhead than the default selector event loop while remaining fully compatible with the standard asyncio API.

⚡ Why VeloxLoop?

• Built in Rust for memory safety, zero-cost abstractions, and exceptional performance.
• Powered by the lightweight and modern io-uring and tokio crate for cross-platform epoll/kqueue/IOCP support.
• Completely independent design — no shared code or direct architectural overlap with existing projects (including RLoop, uvloop, or others).
• Focus on clean, efficient readiness handling, minimal overhead, and excellent cross-platform behavior.
• Early results show strong potential for outperforming current alternatives.

Status

Pre-alpha / Work in Progress
VeloxLoop is under active development. Basic socket I/O and timers are functional, but many asyncio features are still being implemented.
Not yet suitable for production — intended for experimentation, testing, and contributions.

Installation

VeloxLoop will be distributed as pre-built wheels for Linux, macOS, and Windows via PyPI.

pip install veloxloop

Missing Features / Roadmap

The following asyncio features are not yet implemented and are planned for future development:

Core Event Loop Features

• Async context managers - Full support for async with on loop methods

Network & Transport Layer

• Unix domain sockets - Support for AF_UNIX sockets
• Unix pipes - connect_read_pipe() and connect_write_pipe()

Subprocess Management

• subprocess_exec() - Execute shell commands asynchronously
• subprocess_shell() - Shell command execution

Signal Handling

• add_signal_handler() - Register signal callbacks (Unix)
• remove_signal_handler() - Unregister signal callbacks
• Signal integration - Proper Ctrl+C and signal handling

Advanced I/O

• File descriptor passing - Passing file descriptors between processes
• connect_accepted_socket() - Create transport from accepted socket

Platform-Specific Features

• Windows named pipes - Windows IPC support
• Windows IOCP optimizations - Better Windows performance
• macOS/BSD kqueue optimizations - Platform-specific tuning

Performance & Diagnostics

• Slow callback warnings - Debug mode performance monitoring
• Loop instrumentation - Detailed metrics and statistics
• Zero-copy operations - Minimize data copying where possible

Compatibility & Standards

• PEP 567 context variables - Full context support in all callbacks
• ProactorEventLoop equivalence - Windows proactor-style operations

Benchmarks

VeloxLoop includes a comprehensive benchmark suite to compare performance against asyncio and uvloop. The benchmarks measure:

• Raw socket performance - TCP echo using sock_recv/sock_sendall
• Stream performance - TCP echo using asyncio.StreamReader/StreamWriter
• Protocol performance - TCP echo using asyncio.Protocol

Each benchmark tests three message sizes (1KB, 10KB, 100KB) and measures throughput, latency, and percentiles.

Running Benchmarks

cd benchmarks
pip install -r requirements.txt
python run.py all

See benchmarks/README.md for detailed documentation.