Abstract :
The broadcast function is one of the most used collective communication functions of the message passing interface (MPI) library. Broadcasts are usually implemented with invariable algorithms, which fail to yield the best performance with all kinds of applications in all execution environments. This problem should be addressed, since the performance of the function has great influence on MPI-based applications. In this paper, we present, simulate, analytically model, verify and analyze RMBcast, a reconfigurable broadcast function, which presents variable structures/behaviors, in order to optimize flexibility and performance. Our results show that reconfiguration at the algorithm level yields flexibility and performance gains over traditional broadcast functions, which are implemented with invariable algorithms found in well-known implementations of the MPI standard (i.e. MPICH and LAM/MPI)
Keywords :
application program interfaces; message passing; software libraries; LAM/MPI; MPI standard; MPI-based applications; MPICH; RMBcast; collective communication functions; invariable algorithms; message passing interface; reconfigurable MPI broadcast function; Algorithm design and analysis; Analytical models; Art; Broadcasting; Delay; Hardware; Libraries; Message passing; Optimization; Performance gain;