Comparison of CellSs and native programming with a Jacobi solver and triple-matrix-multiply on Cell/B.E
نویسندگان
چکیده
The Cell Superscalar framework (CellSs), from Barcelona Supercomputing Centre, offers a high-level portable programming model to port, parallelise and tune applications on Cell Broadband Engine. Via implementation of a Jacobi solver and a triple-matrix-multiply (TMM) kernel from a wavelet-based evaluation of Coulomb potentials in molecular systems, using the native programming API of the IBM Software Development Kit (CellSDK) and the latest version of CellSs, ease of programming and resulting performance are assessed. CellSs is found to be a convenient and effective vehicle for achieving performant parallelisations. The relative simplicity of the CellSs programming model and its efficient automatic implementation of task scheduling and internal data management were particularly helpful during the development of a novel TMM algorithm with more efficient memory usage, which was necessary to realise the application within the limited SPE memory of Cell/B.E..
منابع مشابه
CellSs: Making it easier to program the Cell Broadband Engine processor
With the appearance of new multicore processor architectures, there is a need for new programming paradigms, especially for heterogeneous devices such as the Cell Broadband Enginee (Cell/B.E.) processor. CellSs is a programming model that addresses the automatic exploitation of functional parallelism from a sequential application with annotations. The focus is on the flexibility and simplicity ...
متن کاملExtracting Dynamics Matrix of Alignment Process for a Gimbaled Inertial Navigation System Using Heuristic Dynamic Programming Method
In this paper, with the aim of estimating internal dynamics matrix of a gimbaled Inertial Navigation system (as a discrete Linear system), the discretetime Hamilton-Jacobi-Bellman (HJB) equation for optimal control has been extracted. Heuristic Dynamic Programming algorithm (HDP) for solving equation has been presented and then a neural network approximation for cost function and control input ...
متن کاملThe coefficients of differentiated expansions of double and triple Jacobi polynomials
Formulae expressing explicitly the coefficients of an expansion of double Jacobi polynomials which has been partially differentiated an arbitrary number of times with respect to its variables in terms of the coefficients of the original expansion are stated and proved. Extension to expansion of triple Jacobi polynomials is given. The results for the special cases of double and triple ultraspher...
متن کاملDistributed Disk-based Solution Techniques for Large Markov Models
Very large Markov chains often arise from stochastic models of complex real-life systems. In this paper we investigate disk-based techniques for solving such Markov chains on a distributed-memory parallel computer. We focus on two scalable numerical methods, namely the Jacobi and Conjugate Gradient Squared (CGS) algorithms. The critical bottleneck in these methods is the parallel sparse matrix-...
متن کاملJacobi Operational Matrix Approach for Solving Systems of Linear and Nonlinear Integro-Differential Equations
This paper aims to construct a general formulation for the shifted Jacobi operational matrices of integration and product. The main aim is to generalize the Jacobi integral and product operational matrices to the solving system of Fredholm and Volterra integro--differential equations which appear in various fields of science such as physics and engineering. The Operational matr...
متن کامل