Pat McCormick, Research Computer Scientist, at Los Alamos National Lab talked to us today about his research at Los Alamos on heterogeneous co-processing architectures and programming models with data analysis and visualization.
Like many computational researchers, Pat faces a familiar challenge, as data complexity increases, the task of analyzing data is becoming harder and requires special tools and techniques. “Our research challenge is dealing with massive amounts of data, not only from the high performance computing aspect but how to analyze the data from simulations,” states McCormick.
In his research, McCormick is using a language called SCOUT that uses NVIDIA GPUs to help speed data analysis. “When using GPUs, rendering and data analysis goes directly through the GPU. The combination of using SCOUT and processing directly on the GPU saves us the time of writing separate programs as is traditionally required. This speeds up the process of data analysis to almost real time—saving time in both processing and analysis. ”