Inproceedings,
A Novel Multi-GPU Neural Simulator
Proceedings of 3rd International Conference on Bioinformatics and Computational Biology 2011, New Orleans, LA, (March 2011)
Abstract
Between the biophysical and behavioral studies of the brain
lies computational neuroscience. The goal of which, among
other things, is to help bridge the gap in our knowledge and
provide alternative or com- plimentary theories to other
neurological studies. As more information is provided and
more complex theo- ries are developed, the size and
computational cost of neural models continues to
increase. This is an obvious impediment to the field and
something that develop- ers are constantly attempting to
overcome. Presented here is a unique simulator design aimed at
leveraging advances in hardware for the simulation of
biologically realistic neural models. This proof-of-concept
design offers an example of a high-performance environment
that utilizes multiple general purpose graphical pro- cessing
units in a novel configuration. The result is a scalable
system that offers the promise of both perfor- mance and
biophysical faithfulness.
