1 Introduction

Parallel machines with an extremely large number of processors are now being designed and built. For example, the BlueGene/L (BG/L) machine built by IBM has 64,000 dual-processor nodes with 360 teraflops of peak performance. Another more radical design from IBM, code-named Cyclops (BlueGene/C), had over one million floating point units, fed by 8 million instructions streams supported by individual thread units, targeting 1 petaflops of peak performance.

It is important that one can study the programming issues and performance of parallel applications on such machines even before the machine is built. Thus, we have developed a parallel simulator - BigSim, to facilitate this research.

Since our research was initiated by the BlueGene/C project, in previous editions of this manual we also called our simulator as Blue Gene Simulator. Our simulator is capable of simulating a broad class of "Massively Parallel Processors-In-Memory", or MPPIM machines.