Acelerando uma aplicação de simulação computacional para o processo de ablação por radiofrequência usando GPU

Abstract

Computer simulation is a tool of great importance in the scientific research field nowadays. The use of this tool depends on an efficient computational representation to make its use viable. Thus, in this work we study and analyze the RAFEM application which was developed aiming to simulate the RFA clinical procedure. An existing problem in this program is the high execution time involving a single simulation. Thus, the present work aims to achieve a reduction in the total waiting time for computing the numerical results of the application so that its use is feasible in a clinical environment. As a methodology, it was adopted an iterative development process aimed at parallel applications, in order to coordinate this task and generate parallel versions of the code for reducing the total execution time. Using this process allied to the use of GPU-accelerated code libraries, it was possible to generate different parallel versions of the application in an iterative way, adding improvements to each new version. The results obtained with the last version generated showed that it was possible to reduce the computation time by up to 27 times, maintaining the numerical results generated equivalent to those produced by the original version, thus making the use of the RAFEM application more feasible by reducing waiting time for the results.