Parallel for Loop and Parallel Reduction: An SMP Comparison of four Languages

AbstractParallel for loop, a typical example of task parallelism assigns different iterations of the loop to different tasks. In combination with parallel reduction, it is a handy, yet powerful construct for the programmer. We give a joint evaluation of these constructs, under four recent parallel processing languages / libraries on a Symmetric Multi-Processing (SMP) system. The problem used for benchmarking is the evaluation of Pi (π) using trapezoidal rule on the derivative of arc tangent. OpenMP, one of the de-facto standards for shared memory multiprocessing, is found to be the best, on a quad core 64-bit processor on Linux platform