This page presents a Channel Coding Software Decoders "Hall of Fame". It allows to see at a glance what has been achieved, what can be expected from today software decoders, and easily compare their respective characteristics. For now, three wide code families are considered: the Turbo codes (LTE, LTE-Advanced, CCSDS, etc.), the Low-Density Parity-Check (LDPC) codes (5G, Wi-Fi, WiMAX, CCSDS, WRAN, DVB-S2, etc.), and the more recently introduced Polar codes (5G).
All the presented results, collected from the state-of-the-art research papers published in the field, consider a BPSK (Bit Phase-Shift Keying) modulation/demodulation and an AWGN (Additive White Gaussian Noise) channel.
This Hall of Fame strives to present results as fairly as possible: for example, early termination criteria are not taken into consideration while computing throughput, in order to compare raw performances using a consistent method. It remains possible, however, for typos/glitches/mistakes to have inadvertantly made it to the scoreboard. In that eventuality, do not hesitate to contact us. If you would like to have your decoder listed as well in the Hall of Fame: please send us the corresponding research paper references, and we will be delighted to add them.
In blue, the results simulated or reproducible with AFF3CT: our Open-source communication chain dedicated to the Forward Error Correction (FEC) simulations.
Last update: 2021-05-17.
Do you like the FEC Software Decoders Hall of Fame? Is it useful in your research works? If yes, you can thank us by citing the following journal article: A. Cassagne et al., “AFF3CT: A Fast Forward Error Correction Toolbox!,“ SoftwareX, 2019.
Maximum A Posteriori (MAP) - 8-state trellis
Work | Year | Platform | Implem. | Pre. | Inter | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
[1] | 2010 | Tesla C1060 | ML-MAP | 32 | 100 | 6144 | 1/3 | 5 | 76800 | 8.0 | 6.7 | 0.021 | 29851 |
[2] | 2011 | GTX 470 | ML-MAP | 32 | 100 | 6144 | 1/3 | 5 | 20827 | 29.5 | 24.6 | 0.045 | 8740 |
[3] | 2011 | i7-960 | ML-MAP | 16 | 1 | 1008 | 1/3 | 8 | 138 | 7.3 | 9.7 | 0.380 | 13402 |
[4] | 2012 | 9800 GX2 | ML-MAP | 32 | 1 | 6144 | 1/3 | 5 | 3072 | 2.0 | 1.7 | 0.0043 | 115882 |
[5] | 2012 | Tesla C2050 | L-MAP | 32 | 32 | 11918 | 1/3 | 5 | 108965 | 3.5 | 2.9 | 0.0057 | 85172 |
[6] | 2012 | X5670 | EML-MAP | 8 | 6 | 5824 | 1/3 | 3 | 157 | 222.6 | 111.3 | 0.396 | 854 |
[7] | 2013 | GTX 480 | EML-MAP | 32 | 1 | 6144 | 1/3 | 6 | 50 | 122.8 | 122.8 | 0.183 | 2036 |
[8] | 2013 | GTX 580 | ML-MAP | 32 | 1 | 6144 | 1/3 | 6 | 1660 | 3.7 | 3.7 | 0.0047 | 63946 |
[9] | 2013 | GTX 550 Ti | EML-MAP | 32 | 1 | 6144 | 1/3 | 6 | 72 | 85.3 | 85.3 | 0.247 | 1360 |
[10] | 2013 | GTX 680 | EML-MAP | 32 | 16 | 6144 | 1/3 | 6 | 2657 | 37.0 | 37.0 | 0.024 | 5270 |
[10] | 2013 | i7-3770K | EML-MAP | 16 | 4 | 6144 | 1/3 | 6 | 323 | 76.2 | 76.2 | 0.680 | 1011 |
[11] | 2014 | Tesla K20c | ML-MAP | 32 | 1 | 6144 | 1/3 | 5 | 1097 | 5.6 | 4.7 | 0.0026 | 47872 |
[12] | 2014 | GTX 580 | BR-SOVA | 8 | 4 | 6144 | 1/3 | 5 | 192 | 127.8 | 106.5 | 0.135 | 2291 |
[13] | 2016 | GTX 680 | EML-MAP | 32 | 1 | 6144 | 1/3 | 7 | 817 | 8.2 | 9.6 | 0.0062 | 20313 |
[14] | 2016 | 2xE5-2680v3 | EML-MAP | 16 | 192 | 6144 | 1/3 | 6 | 2657 | 443.7 | 443.7 | 0.924 | 541 |
[14] | 2016 | 2xE5-2680v3 | EML-MAP | 8 | 384 | 6144 | 1/3 | 6 | 3293 | 716.4 | 716.4 | 0.746 | 335 |
[15] | 2019 | 2xE5-2680v3 | EML-MAP | 8 | 24 | 6144 | 1/3 | 6 | 84 | 1735.0 | 1735.0 | 0.904 | 138 |
Fully-Parallel Turbo Decoder (FPTD) - 8-state trellis
Work | Year | Platform | Implem. | Pre. | Inter | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|
[13] | 2016 | GTX 680 | FPTD | 32 | 1 | 6144 | 1/3 | 36 | 403 | 18.7 | 10428 |