GPU

We were developing a Personal Supercomputer, or The Supercomputer for Everybody, based on General Purpose Graphic Processing Units (GP GPU). Our platform is Tesla GP GPU from nVidia. The reason is obvious - we wanted to save all of the three - computing time on, money for and size of our desktop supercomputers. In order to do this we used Tesla GPU cards that is designed ground-up for parallel computing and which offers exclusive computing features.

State-of-the-art of GP GPU implementation in LAAL, CS Dept, SoE, VCU:

Interface for GP GPU applications.

Distance (proximity, affinity, similarity) matrix calculation for huge data sets.

PCA for large number of both samples and features based on SVD.

Support Vector Machines design for large data sets.

Support Vector Machines design for large data sets by using geometric approaches.

Local Linear and Nonlinear Support Vector Machines (LL SVM & LnL SVM) design.

Adaptive Local Hyperplane (ALH) design.

LAAL was developing the course for GP GPU use in scientific computation. To this end we had presented the very first research seminar on GP GPU applications in solving scientific problems at VCU, the Virginia's largest public research university.