pyreedsolomon

A high performance Python interface to the linux kernel Reed Solomon libraries

See more on https://github.com/mugpahug/pyreedsolomon

Author: Edwin G.W. Peters

Requirements

Userspace interface to the kernel Reed Solomon libraries and numpy Source and build instructions on https://github.com/CyberLeo/reed-solomon

configure with --prefix=/usr/ or add the location of librs.so is in your library path (LD_LIBRARY_PATH)

Installation

pip install pyreedsolomon

Usage

This library provides fast methods where the user needs to allocate the buffers and handle padding as well as easy-to-use methods that have a slightly lower performance

import pyreedsolomon
import numpy as np

rs_dr = pyreedsolomon.Reed_Solomon(8,223,255,0x11D,0,1,32)

data = np.random.randint(0,256,150).astype(np.uint8)

data_enc = rs_dr.encode(data)

# create a few errors
err_idx = [23,53,12,97,102]


data_enc[err_idx] = 255

data_dec, n_errors = rs_dr.decode(data_enc)

verify = np.all(data_dec==data)

print(f"Decoding succes: {verify}. errors corrected {n_errors}")

For more examples including how to use the fast encoder and decoder that omits variable checking, check tests/bench.py

Performance

Performance comparison of pyreedsolomon with different input data types and comparison to unireedsolomon and reedsolo Input data types supported:

• numpy.ndarray
• list
• bytes
• bytearray

Test run on Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz

RS (973, 935, 10) -- encoding and decoding 1000 runs

RS (255, 223, 8) -- encoding and decoding 10000 runs

RS (15, 11, 4) -- encoding and decoding 100000 runs

Test run on Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz

RS (973, 935, 10) -- encoding and decoding 1000 runs

RS (255, 223, 8) -- encoding and decoding 10000 runs

RS (15, 11, 4) -- encoding and decoding 100000 runs