high performance computing on graphics processing units: hgpu.org

Massively parallel Graphics Processing Unit (GPU) hardware has become increasingly powerful, available and affordable. Software tools have also advanced to the point that programmers can write general purpose parallel programs that take advantage of the large number of compute cores available in the hardware. With literally hundreds of compute cores available on a single device, program performance can increase by orders of magnitude. We believe that introducing students to the concepts of parallel programming for massively parallel hardware is of increasing importance in an undergraduate computer science curriculum. Furthermore, we believe that students learn best when given projects that reflect real problems in computer science. This paper describes the experience of integrating two undergraduate computer science courses to enhance student learning in parallel computing concepts. In this cross teaching experience we structured the integration of the two courses such that students studying parallel computing worked with students studying advanced rendering for approximately 30% of the quarter long courses. Working in teams on a joint project, both groups of students were able to see the application of parallelization to an existing software project with both the benefits and complications exposed early in the curriculum of both courses. Motivating projects and performance gains are discussed, as well as student survey data on the effectiveness of the learning outcomes. Both performance and survey data indicate a positive gain from the cross teaching experience.