Mini-simpósio no MECOM-CILAMCE 2010

Introdução do nosso mini-simpósio no MECOM-CILAMCE 2010 -IX Congresso Argentino de Mecânica Computacional e XXXI Congresso Ibero-Latino-Americano de Mecânica Computacional. Esses congressos acontecerão entre 15 e 18 de novembro de 2010 em Buenos Aires. Prepare seu trabalho!

Title

High Performance Computing on Graphics Hardware (GPGPU)

Organizers

Euclides Mesquita, euclides@fem.unicamp.br
Josué Labaki, labaki@fem.unicamp.br
Luiz Otávio Saraiva Ferreira lotavio@fem.unicamp.br

Department of Computational Mechanics DMC
School of Mechanical Engineering -FEM
State University at Campinas – UNICAMP/Brazil.

Summary

After a long period of steady growth, desktop commodity computer architecture has reached its ceiling on computing performance. Further progress is no longer enabled by growth in core clock rates, but by growth in parallelism. Many articles have been presented in the last editions of CILAMCE and MECOM on the use of multi-cored computing applied to a variety of problems. Furthermore, there is a growing trend toward parallel computation on the recently created general-purpose graphics hardware (GPGPU) [1]. Due to its architecture, the GPU is specially well-suited to address problems that can be expressed as data-parallel computations with high arithmetic intensity (the ratio of arithmetic operations to memory operations) [2]. Recent works have been shown astonishingly fast computations. For instance, papers on the use of GPGPU for computational mechanics have been reporting performances up to 100 times faster than CPUs for solution of linear systems, and around 900 times faster for numerical solution of infinite oscillatory integrals [3]. The first edition of this mini-symposium on CILAMCE 2009 has shown that the GPU has been investigated by researches of a very broad range of subjects in computational modeling with considerable gains in performance. The idea of the present mini-symposium is to gather researchers working on HPC tasks using GPGPU systems, aiming to discuss the state of the art issues, exchange experiences and solutions for existing implementation strategies and problems.

Suggested readings

[1] OWENS, J. D. et al., “A Survey of General-Purpose Computation on Graphics Hardware”, Computer Graphics, 26, 1 (2007), pp.  80-113.

[2] NVIDIA. “NVIDIA CUDA – Compute Unified Device Architecture – Programming Guide”. NVIDIA Corporation, Santa Clara (2008).

[3] MESQUITA, E.; LABAKI, J.; FERREIRA, L. O. S. “An Implementation of the Longman’s Integration Method on Graphics Hardware”. Computer Modeling in Engineering and Sciences, 51, 2 (2009), pp. 143-168.