Table 4 GPU decoding comparison at SNR = 0.161 with d = 8 on BIAWGNC targeting FER = 0.04 with Rcode = 1/10 codes
From: Quasi-cyclic multi-edge LDPC codes for long-distance quantum cryptography
Specification | This Work (2018) | Jouguet and Kunz-Jacques (2014)31 | |||
|---|---|---|---|---|---|
Code rate | 1/10 | 1/10 | |||
Block length (Bits) | 220 | 220 | |||
SNR | 0.161 | 0.161 | |||
LDPC code structure | Random Multi-Edge | q = 512 QC Multi-Edge | Random Multi-Edge | ||
Connections in parity-check matrix | 4,063,229 | 7932 | N/A | ||
Early termination | No | Yes | No | Yes | No |
Max iterations | 88 | 88 | 100 | 100 | 100 |
Average iterations | 88 | 78 | 100 | 78 | 100 |
FER | 0.04 | 0.04 | 0.0243 | 0.0243 | 0.04 |
Latency per iteration (ms) | 4.73 | 4.84 | 1.28 | 1.47 | 1.48 |
\(K_{{\mathrm{GPU}}}^{{\mathrm{raw}}}\) GPU raw throughput (Mb/s) | 2.52 | 2.78 | 8.21 | 9.17 | 7.1 |
\(K_{{\mathrm{GPU}}}^\prime\) GPU info. throughput (Kb/s) | 242 | 267 | 801 | 895 | 682 |
GPU model | NVIDIA GeForce GTX 1080 | AMD Radeon HD 7970 | |||
CMOS technology | 16 nm | 28 nm | |||
GPU cores | 2560 | 2048 | |||
GPU GFLOPS | 8228 | 3789 | |||
GPU memory bus width (Bits) | 256 | 384 | |||
GPU memory bandwidth (GB/s) | 320 | 264 | |||
