RELDEC matlab implementation

Folders

• LDPC_Matlab contains all matlab implementations of decoders and other utils required
• LDPC_Cpp contains cpp implementation of layered and flooding decoders for WLAN LDPC

To run reldec:

• We need a dataset of LLRs and correspoding codewords from which these LLRs are generated. This LC_dataset.mat can be found in LDPC_Matlab/Datasets/ . It can be generated by:

  1. run setup from ./LDPC_Matlab
  2. cd Extra
  3. run LC_generate after changing required params in the same file

• We also need the clusters and the variable nodes belongignto the cluster, and this is generated into another mat file: LDPC_init.mat. To generate this, we need to change parameters in ./cluster_create.py file and run it (requirements: scipy, numpy, networkx, matlab.engine). ./cluster_create.py requires ./utils.py for internal utility.

These 2 files are already part of ./LDPC_Matlab/Datasets/ so there is no need to generate these unless you require some parameter changes (cluster size, LLR dataset size, etc).

You can then run RELDEC from LDPC_Matlab directory. This will generate the Q table after training.

To run matlab decoders

You can navigate to ./LDPC_Matlab and check for files

• ldpc_layered.m
• ldpc_reldecode.m (calls RELDEC inside to generate Q table)
• ldpc_flooding.m

These files contain all the required parameters to run an FER simulation for a given range of SNRs. Generally these files are run using a batch job over the cluster. This provides the option of using parallel computation (each FER simulation over a given SNR is independent of the other). The parfor utility provided by Matlab is used in this case.