Towards an optimal parallel decoding of turbo codes

High throughput decoding of turbo-codes can be achieved thanks to parallel decoding. However, for finite block sizes, the initialisation duration of each half-iteration reduces the activity of the processing units, especially for higher degrees of parallelism. To solve this issue, a new decoding scheduling is proposed, with a partial processing overlapping of two successive half iterations. Potential memory conflicts introduced by this new scheduling are solved by a constrained interleaver design. An example of application of the proposed technique shows that the complexity of the decoder is reduced by 25 % compared to a conventional approach.