Granularity and Complexity in Parallel Systems

M. Maliţa (USA) and G. Ştefan (Romaina)


parallel computation, size, (algorithmic) complexity, granularity


Computational performances for parallel architecture could be defined only application domain dependent. Building efficient parallel machines supposes supra-linear computational performances. An efficient parallel system must correlate its granularity with the complexity of its elements. The supra-linear condition is introduced to define the conditions for a parallel system to have supra-linear performances. Results show that the basic rule governing large computational systems is that their complexity must grow slower than their size because computation is a simple process. The distinction between size and (algorithmic) complexity is defined.

Important Links:

Go Back