Using a Commodity GPU in an Undergraduate Parallel Computing Course
|
Abstract
A modern graphics processor unit (GPU) contains dozens of high performance floating point processors. In raw computational power, they far outstrip current CPUs. GPUs have become a popular parallel platform among researchers. After graduation, our students are more likely to have a GPU available than a traditional parallel machine, yet they rarely gain experience with SIMD machines. We have developed a short module suitable for an undergraduate parallel computing course with very limited prerequisites.