Out-of-core solution of linear systems on graphics processors
Maribel Castillo, Francisco D. Igual, Mercedes Marqués, Rafael Mayo, Enrique S. Quintana–Ort́ı, Gregorio Quintana‐Ortí, Rafael Rubio, Robert A. Geijn · International Journal of Parallel Emergent and Distributed Systems · 2009
We combine two high-level application programming interfaces to solve large-scale linear systems with the data stored on disk using current graphics processors (GPU). The result is a simple yet powerful tool that enables a fast development of object-oriented codes, implemented as Matlab M-scripts, for linear algebra operations. The approach enhances the programmability of the solutions in this problem domain while unleashing the high performance of GPU. Experimental results are reported from Octave, linked with the implementation of basic linear algebra subprograms by Nvidia; as an additional novelty, we evaluate the performance of our approach for the Nvidia new double-precision graphics hardware.