High-throughput GPU layered decoder of quasi-cyclic multi-edge type low density parity check codes in...

Area of Science:

• Quantum Information Science
• Computer Engineering
• Applied Mathematics

Background:

• Decoding throughput is a critical bottleneck in continuous-variable quantum key distribution (CV-QKD) systems.
• Efficient decoding algorithms are essential for practical CV-QKD implementation.

Purpose of the Study:

• To propose a novel layered decoder architecture for quasi-cyclic multi-edge type LDPC (QC-MET-LDPC) codes.
• To accelerate the decoding process in CV-QKD systems using graphics processing units (GPUs).

Main Methods:

• Implementation of a layered decoder utilizing GPU parallel processing capabilities.
• Optimization of parity check matrix storage and merging of unrelated sub-matrices.
• Parallel decoding of multiple codewords on the GPU without early termination.

Main Results:

• Achieved average decoding speeds of up to 64.11 Mbits/s (0.1 code rate), 48.65 Mbits/s (0.05 code rate), and 39.51 Mbits/s (0.02 code rate).
• Demonstrated significant speed improvements by decoding 128 codewords simultaneously.
• Validated the effectiveness of the proposed GPU-based decoding approach.

Conclusions:

• The proposed layered decoder effectively addresses the decoding throughput bottleneck in CV-QKD.
• GPU acceleration provides a substantial performance enhancement for LDPC code decoding in QKD.
• This advancement paves the way for more practical and high-speed CV-QKD systems.