Fig. 3

Memory footprint for multithreaded execution of FGS, FGS+ and FGSrs on long reads (1328 bp). Total memory footprint (heap, stack and memory-mapped file I/O) measured using the Massif heap profiler of Valgrind [20] with –pages-as-heap option. Race conditions consistently halt the execution of FGS+ above 10 threads. FGS and FGSrs generate DNA sequences, protein translations and metadata, whereas FGS+ only generates protein translations because the software crashes when other output is generated. FGS and FGS+ report gene predictions out-of-order, where default in-order reporting was used for FGSrs