A Comparison of Multithreading Implementations

As parallel and distributed processing becomes more prevalent in computer science the need for an optimal multithreading implementation becomes more pertinent. Artificial intelligence is an area where the application of optimal multithreading is well suited. One aspect of artificial intelligence is based on the idea of searching a given problem space for an optimal solution. The size of this search space is often extensive and parallel methods are employed to decrease the search time. With the implementation of these parallel search methodologies multithreading becomes an important part of the solution. Research has shown that a multithreaded algorithm implemented using Cilk threads performs dramatically better than either a POSIX threads or Java threads implementation. In comparing these multithreaded implementations an algorithm that uses double recursion to compute the Fibonacci function was selected as the comparison algorithm as it is well suited to demonstrate, and take advantage..