Abstract—The well known Floyd-Warshall (FW) algorithm solves the
all-pairs shortest path problem on directed graphs. In this work, we
parallelize the standard FW and two cache-friendly versions using three
different parallel programming environments, namely OpenMP, Cilk and
Threading Building Blocks. We experimented with multiple alternative
parallel versions, in order to gain insight on the execution behavior
of the parallelized algorithms on modern multicore platforms, and on
the programmability of the aforementioned environments. We were able
to significantly accelerate FW performance utilizing the full capacity
provided by the multicore architectures used.