high performance computing on graphics processing units: hgpu.org

Next generation DNA sequencing technologies generate terabytes of image data in a typical run over several days. Compute power to process the increasing amount of image data is becoming a problem in next generation sequencing. We propose to use the compute power of Graphical Processing Units (GPUs) to address this problem. GPUs have an efficient parallel processor architectur tailored for tasks like image processing. We explore several options for essential processing tasks like cluster/bead segmentation and image registration. Using the scan algorithm we show how some sequential algorithms can be parallelized. We show that cluster/bead segmentation and image registration tasks can be executed about 4x faster on a late 2009 mid-level GPU than a comparable CPU.