This survey considers all ASIC implementations of LDPC decoders that have been published since January 2014 on IEEEXplore, as well all of their benchmarkers (most of which were published before 2014).

Results highlighted in yellow have been inferred from the numbers that are provided in the papers, which are shown without highlighting. The method used to infer the value provided in each highlighted cell can be seen in its underlying formula.

Wherever possible, data throughputs are obtained according to throughput(data,min) = throughput(encoded,min)*rate(lowest). throughput(data,max) = throughput(encoded,max)*rate(highest). If the minimum encoded throughput is not available, then throughput(data,min) = throughput(encoded,max)*rate(lowest) is used.

Cell AG1 allows the target technology scale to be specified. All throughputs are scaled according to throughput(new)=throughput(old)*scale(old)/scale(new). All latencies are scaled according to latency(new)=latency(old)*scale(new)/scale(old). All areas are scaled according to area(new)=area(old)*scale(new)^2/scale(old)^2. Powers are not scaled.

The efficiencies are obtained according to areaeff(min) = throughput(data,min)/area, areaeff(max) = throughput(data,max)/area, energyeff(min) = throughput(data,min)/power(max), energyeff(max) = throughput(data,max)/power(max). Note that power(max) is used for both energy efficiencies because power(min) is unavailable for most decoders.

Scatter plots are provided for min and max throughput, area efficiency and energy efficiency, all as functions of the number of PCMs supported. In general, it can be seen that increasing the number of supported PCMs by 10 times reduces each of the achievable throughput, area efficiency and energy efficiency by 10 times.


This database is made available under the Open Database License: http://opendatacommons.org/licenses/odbl/1.0/. Any rights in individual contents of the database are licensed under the Database Contents License: http://opendatacommons.org/licenses/dbcl/1.0/