Selecting Error Correcting Codes to Minimize Power in Memory Checker Circuits

The approach proposed in this paper reduces power consumption in single-error correcting, double error-detecting checker circuits that perform memory ECC. Power is minimized with little or no impact on area and delay, using the degrees of freedom in selecting the parity check matrix of the error correcting code. The non-linear power optimization problem is solved using two methods, genetic algorithms and simulated annealing. Both the methods are applied to two SEC-DED codes: standard Hamming codes and odd-column-weight Hsiao codes. Experiments on actual memory traces of Spec and MediaBench benchmarks indicate that considering power along with area and delay when selecting the parity check matrix can result in power reductions of up to 27% for Hsiao codes and up to 41% for Hamming codes. Experiments are also performed to motivate the choice of parameters of the non-linear optimization algorithms, using sensitivity analysis of the low-power solutions to the choice of the different parameters of each algorithm.