The next generation of super computing has arrived. GPUs are becoming more widely available, and with the advent of new technology allowing them to process non-graphical information in parallel many scientists are beginning to harness their massive computing power. Per dollar GPUs can do several times the floating point work of CPUs, and are perfect for offloading instructions that can be done in parallel. CUDA allows programmers to harness this power in NVidia graphics cards, and with more GPU enabled nodes popping up in clusters around the country it has become a new programming challenge to effectively use multiple GPUs on multiple nodes in a cluster. Check out this paper to see their results!
Computer Science Ethics Class
