GPU-Based Acceleration In Linear Algebra, Bioinformatics, and Risk Analysis

The recent introduction of general-purpose GPUs has created an opportunity and a need to harness their processing power for a wide variety of scientific-computing applications. In this talk, we summarize our recent results on GPU-based computing in applications coming from linear algebra, bioinformatics, and risk analysis. Solving linear systems of equations [1], carrying out BLAST queries [2], and running Monte Carlo simulations [3] exhibit inherent parallelism, but have significant differences, which pose different challenges in their parallel GPU-based implementation. We review and contrast these differences and discuss how each implementation succeeded in gaining significant speedups in comparison to conventional hardware.

[1] P. D. Vouzis, N. V. Sahinidis, “GPU-BLAST: using graphics processors to accelerate protein sequence alignment,” Bioinformatics, Vol. 27, no. 2, pages 182-188, 2011.

[2] J. Elble, N. V. Sahinidis, P. D. Vouzis, “GPU computing with Kaczmarz’s and other iterative algorithms for linear systems,” Parallel Computing, Vol. 36, no. 5-6, 2010.

[3] Y. Zhang, P. D. Vouzis, N. V. Sahinidis, “GPU simulations for risk assessment in CO2 geologic sequestration,” Computers & Chemical Engineering, 2011 (accepted).