Fig. 1

Multiprocessing performance for PyIR and multithreaded IgBLAST (version 1.14). a One million synthetic immunoglobulin sequences were used to time PyIR (dark grey, ♦) against multithreaded IgBLAST (version 1.14) (grey, ■) as a function of the number of processes. Idealized timings are shown as a black dashed line. Average timings were measured over the three trial runs for 1 million sequences and computed separately for both IgBLAST and PyIR. Standard deviations appear as error bars for both methods. X and Y axes are in log₂ space. b The speedup of PyIR relative to multithreaded IgBLAST (version 1.14) as a function of the number of simultaneous processes. Timings were done on a workstation equipped with 4 Opteron 6278 hyper-threaded 8-core processors for a total of 64 CPU threads using the average timings from (a). The X and Y axes are in log₂ space. c One billion synthetic immunoglobulin sequences were used to determine the speedup PyIR achieved over multithreaded IgBLAST (version 1.14) as a function of the number of sequences. Idealized speedups are shown as a black dashed line. Timings were done on a workstation equipped 4 Xeon Platinum 8280 hyperthreaded 28-core processors for a total of 224 CPU threads. X and Y axes are in log₁₀ space