nanorq

nanorq is a compact, performant implementation of the raptorq fountain code capable of reaching multi-gigabit speeds on a single core.

API

• High-level API nanorq.h abstracts allocation, alignment, and I/O handling from the user.
• Low-level API nanorq_core.h provides more flexibility and can be used without any libc dependency¹, but requires more integration from the user.

Performance²

Use cases

• firmware deployment / software updates
• video streaming
• large data transfers across high latency links

Helper Modules

Tunable Sparse Network Coding (TSNC)

The high-level API includes experimental block-based recoding helpers that leverage the optimized primitives used for nanorq. These helpers are demos, not part of the RFC 6330 interoperability surface. Dense TSNC recovery is bounded to 256 MiB of coefficient and payload matrix memory and fails cleanly when that limit would be exceeded.

Some examples are provided:

• Blockchain gossip propagation: Accelerating block synchronization across peer-to-peer gossip networks.
• Multipath recoding: Re-encoding packets at intermediate network hops without full block decoding to maximize throughput.
• Real-time continuous streaming: Utilizing a sliding window mechanism to recover missing stream data with minimal overhead and latency.

nanorq objects cache mutable encoder/decoder state. A single object must not be used concurrently without external synchronization; separate objects may be used by separate threads.

Footnotes

1. When compiled with NDEBUG. Test suite requires stdlib. ↩

2. Build autodetects and optimizes for the host CPU architecture (using -march=native). ↩