Parallel Computation: Models and Methods

Focusing throughout on models of computation and methods of problem solving, this text shows how a parallel algorithm can be designed for a given computational problem to run on a parallel computer, and then how it can be analyzed to determine its goodness. The book covers in detail the main three models of computation used to design parallel algorithms (namely combinational circuits, shared memory machines, and interconnection networks). The algorithms are grouped by method of design (prefix computation, divide and conquer, and pointer based methods) and described for interconnection networks varying from the simplest (linear array) to the most advanced (star). The chapter on parallel synergy shows that the concept of superlinear speedup is mathematically possible, and describes some real-life examples where such speedup can be achieved.