The Spatial-Purpose Computing Group focuses on specializing computing hardware to specific tasks, and using spatial hardware to improve performance and reduce energy consumption.
This area includes the use of reconfigurable logic, such as FPGAs, to accelerate applications. In reconfigurable computing, sections of application code are converted to hardware circuitry to improve performance and/or lower energy costs. This specialized hardware is then used as needed by the remaining software code. FPGAs and other reconfigurable hardware provide a reusable platform to load in a changing set of circuits throughout runtime.
An alternate form of accelerator logic that one might use are general-purpose graphics processing units (GPGPUs). GPGPU-based acceleration is a rapidly growing area due to the fact that the programming model is more accessible than is available for most reconfigurable computing systems. We also investigate this area.
