A MIXED APPROACH TOWARDS IMPROVING SOFTWARE PERFORMANCE OF COMPUTE UNIFIED DEVICE ARCHITECTURE APPLICATIONS
Alexandru Pîrjan · 2016
One of the most important aspects when developing applications that leverage the powerful parallel processing power of graphics processing units (GPUs) that offer support for the Compute Unified Device Architecture (CUDA) is to divide the tasks that are to be processed appropriately into sections that can be processed either serially or in parallel. The paper analyses this important developing issue by proposing a mixed approach to programming efficient software applications. The article makes an in depth analysis of the key aspects that carry significant weight when deciding to parallelize a certain part of an application: the analysis phase of the application that is about to be parallelized; the amount of time involved to achieve the implementation; the feasibility of parallelizing the source code; situations when one should aim for central processing units optimization techniques that yield better performance on sequential source code rather than parallelizing the whole algorithm.