Cody Steven: Master’s Thesis Defense
Abstract:Graphics Processing Units (GPUs) can be used for more than just gaming. With the addition of CUDA, NVIDIA has developed a library that allows programmers to repurpose these GPUs for parallel calculations. Parallel computations are most efficient when used to solve large problems, an example of which is graph analysis. Graph algorithms have historically been computationally complex and performing these algorithms on large graphs takes an infeasible amount of time, making them prime targets for parallelization. Using these GPU-Optimized graph algorithms allows us to analyze larger graphs based on real world phenomena, such as the dynamics of a protein folding. Protein folding is a biological process that occurs constantly in every living organism and many diseases and medical conditions have been linked to proteins unable to fold. By combining results from Molecular Dynamic Simulations and these graph analytics our goal is to find a correlation between the two and gain greater insight into protein folding dynamics.