PARALLELIZED DIRECT EXECUTION SIMULATION OF MESSAGE-PASSING PARALLEL PROGRAMS

As massively parallel computers proliferate, there is growing interest in finding ways by which performance of massively parallel codes can be efficiently predicted. This problem arises in diverse contexts such as parallelizing compilers, parallel performance monitoring, and para llel algorithm development. In this paper, we describe one solution wh ere one directly executes the application code, but uses a discrete-ev ent simulator to model details of the presumed parallel machine, such as operating system and communication network behavior, Because this a pproach is computationally expensive, we are interested in Its own par allelization, specifically the parallelization of the discrete-event s imulator. We describe methods suitable for parallelized direct executi on simulation of message-passing parallel programs, and report on the performance of such a system, LAPSE (Large Application Parallel Simula tion Environment), we have built on the Intel Paragon. On all codes me asured to date, LAPSE predicts performance well, typically within 10% relative error. Depending on the nature of the application code, we ha ve observed low slowdowns (relative to natively executing code) and hi gh relative speedups using up to 64 processors.