Fig. 8

The parallel speedup of the Tree and HashMap algorithms. The speedup is compared with the theoretically optimal speedup for \(p=0.92\) given by Amdahl’s law. By profiling, we measured that approximately 8% of the code is serial and cannot be accelerated, although this number differs somewhat depending on sequence size and parameters. For short sequences the speedup is small, but longer sequences have close to optimal speedup. The benchmarking machine has 32 cores, and thus the result for 64 threads illustrates that there are some benefits to scheduling more threads than cores. The HashMap exhibits better parallelisation speedup overall