10/9/14

Algorithmic Overscalability.

Overscalability is quality of a parallel algorithm to scale speed faster than we add processors, when compared to best known sequential algorithm.

For example: when conditions are met (when data list is numerous enough) sorting nets can sort data much faster than classic sequential algorithms and meet requirements of overscalability.

Sorting nets can sort data in time of O(log2 n) using n processors.

Classic sequential algorithms can sort data in time of O(n · log n).

No comments:

Post a Comment