Computational Grids are widely emerging and dramatically influencing scientific computing practice. In this paper, we presents the use of a problem solving environment - the GridPSi to automatically generate Matlab code and carry out simplified Grid computing, in which the previous work of GridPSi is briefly introduced, the implementation of a Matlab script generator for automatic Matlab 3D finite element code generation is described, and running a simulation example on the Grid through UNICORE Grid middleware is illustrated. The paper demonstrates that the proposed PSE can not only tailor both the innovative Grid technologies and legacy computational codes into use easily, but also greatly reduce user's work.

As the power of computational grids increases, there is a corresponding need for better usability in PSE community. In this paper, we describe our experiences on design of project GridPSi, a grid-enabled PSE for physical simulation. GridPSi provides a rich suite of functions that facilitate the composition of computational scripts to automate physical simulations by design domain-specific templates for a verity of research fields. The generic functionality provided by it includes problem modeling, mesh file conversion, computational script generation, solution scheme specification and access to grid resources. GridPSi offers a simple interface that is intuitive to users in the human-PSE interactions and ways of thinking. Furthermore, we developed higher-level graphical components named PDE wizards that build upon the core functions to guide users through problem modeling process with minimum effort. By using Unicore as grid middleware, we are lowering the barriers to entry for users wishing to exploit grid technologies

The authors discuss high-performance methods for solving first-order linear recurrences on a vector computer, investigate automatic transformations, and develop the compiling techniques for first-order linear recurrence problems. The results show that the improved vector code generated by the vectorizing compiler run at the rate of 150 MFLOPS (million floating operations per second), in the case of the moderate loop lengths (>1000), and at the rate of over 200 MFLOPS, in the case of long loop lengths (>10000), on the HITAC S-820 supercomputer. Overall performance improvements of 69% in the 14 Lawrence Livermore Loops and 25% in the 24 Lawrence Livermore Loops, as measured by the harmonic mean, are attained.>