Estudo comparativo da programação paralela de algoritmos em CPU com Go e GPU com CUDA

Abstract

Multicore processors, in other words, with more than one core, have become common usage. However, the study of parallel programming is not much explored. There are many problems that can be solved more efficiently exploiting parallelism and concurrency. This paper presents a study of parallel programming on multicore architectures. Two different approaches are analyzed and compared. The use of new programming language developed by Google, Go, which brings an approach to parallel programming on multicore CPUs, and the CUDA parallel programming plataform which uses the graphics cards from NVIDIA GPUs for parallel processing. Among the computational problems that benefit from parallelism, there are also genetic algorithms.