LDPC/LDGM FOR LINUX

LDPC/LDGM is a cool library that helps with forward error correction. It’s way faster than the older Reed-Solomon codecs, making it a great choice for handling data errors.

Why Choose LDPC/LDGM?

Traditional small block Forward Error Correction (FEC) codes like Reed-Solomon can hit some snags, especially when they work with the Asynchronous Layered Coding (ALC) multicast protocol. That’s why we created this large block FEC codec using Low-Density Parity Check (LDPC) codes. This new codec can handle source blocks that are tens or even hundreds of megabytes long. The idea builds on some smart theories from Gallager back in the 1960s.

What’s New in This Release?

Here are some exciting updates:

• The codec now supports any size of symbols! In past versions, symbols had to be multiples of 4. Just keep in mind that sometimes the packet size still needs to be a power of two, especially if you pack multiple symbols together.


• Performance has improved! We’re seeing a 6.3% boost during encoding and a 3.8% boost during decoding when using LDPC-Staircase codes with a big 20,000 symbol object at a 2/3 code rate.


• The source code is clearer now. Check out src/ldpc_profile.h, which gathers all the codec compilation controls—you can tweak things there if needed!


• A new feature called SPARSE_MATRIX_OPT_SMALL_INDEX slashes memory needs for sparse matrices by reducing their size by about one-sixth on 32-bit systems.


• This version has been tested on 64-bit architectures, specifically Pentium D running Linux.


• It aligns with the IETF Internet-Draft: draft-ietf-rmt-bb-fec-ldpc-03.txt, found in the doc directory, and is compatible with release 2.0 (not with earlier releases).


• No big changes were made to the API, but we added an LDPCFecSession::GetMaxN() method that’s super handy when using SPARSE_MATRIX_OPT_SMALL_INDEX mode.