high performance computing on graphics processing units: hgpu.org

This work has as first goal develop an parallel algorithm considerably faster than it serial pair. The algorithm is based on recurrence plots idea, that is a square matrix where it is possible detect many complex behaviours (e.g. with the recurrence quantification analysis technique) of dynamical systems. In the results and conclusion part, we will present the time saved using CUDA with C programming language in comparison with the serial program and the time relation with the number of kernels acess for some rates of threads per block.