Several different libraries have been developed at the Department of Applied Mathematics, including Shallows and GreenBox.
Project 1: Develop a library for GPGPU / Stream programming in Python.
Many algorithms used in the industry relies heavily on efficient implementations of linear algebra routines. We have previously supervised master student studying direct solvers using the GPU in MATLAB, and one studying direct, and iterative methods.
Project 1: Develop a library that accelerates sparse linear algebra on the GPU / Cell BE.
We have a many-GPU machine consisting of (as of 2009-08-16) four NVIDIA G92 GPUs, and one NVIDIA Tesla GPU.
Programming such a system is far from trivial, and poses many interresting questions. We have previously supervised one student that explored the use of GPU clusters for numerical simulation.
Project 1: Explore cluster/multi-GPU systems to simulate physical phenomonon.
Project 2: Explore cluster/multi-GPU systems for visualization of out-of-core datasets.
Published September 16, 2008