Applied Parallel Computing Projects
Ocean Modeling using CUDA
This is an ongoing project where students work with the Regional Ocean Modeling System (ROMS) to port the software for use on GPGPUs. ROMS is mostly written in Fortran 90, which is still the language of choice for high performance computing applications.
We've successfully ported a portion of ROMS to CUDA-Fortran using the Portland Group PGI Fortran Accelerator Compiler. This work was published in 2011 at the IEEE/MTS OCEANS'11 Conference in Kona, Hawaii.
- OCEANS'11 Paper
Jason Mak, Paul Choboter, and Christopher Lupo. Numerical ocean modeling and simulation with CUDA. In OCEANS 2011, MTS/IEEE KONA - Oceans of Opportunity: International cooperation and partnership across the Pacific, 2011.
This work is continuing, with particular emphasis in using the OpenACC framework.
Cross Teaching Applied Parallel Computing
My colleague Zoë Wood and I have been working to develop innovative teaching approaches for parallel computing and it's applications. In particular, we've been exploring cross-teaching, where two courses merge for two or three weeks in a quarter, and students from each course teach each other how to parallelize an application. In Spring 2011, the application was Ray Tracing. The results from that experiment were quite positive, and the work was published at SIGCSE 2012, and presented again later at NVIDIA's GPU Technology Conference (GTC) in May of 2012.
- SIGCSE 12 Paper
Christopher Lupo, Zoëe Wood, and Christine Victorino. Cross teaching parallelism and ray tracing: A project-based approach to teaching applied parallel computing. In Proceedings of the 43rd ACM SIGCSE Technical Symposium on Computer Science Education, SIGCSE '12, New York, NY, USA, 2012. ACM. - GTC 2012 Session