The simulation of incompressible fluids is one of the important problem classes in computational fluid dynamics. We consider a simulation algorithm for the convection in binary fluid mixtures, a problem where a quite ...
详细信息
ISBN:
(纸本)0769500595
The simulation of incompressible fluids is one of the important problem classes in computational fluid dynamics. We consider a simulation algorithm for the convection in binary fluid mixtures, a problem where a quite simple model describes a very complex behavior: In a parallel implementation on an IBM SP2, we investigate several implementation strategies involving different data layouts and communication organizations.
作者：
Rosato, VPucello, NENEA
HPCN Project Ente Nuove Tecnol Energia & Ambiente I-00100 Rome Italy
A code for Me simulation of X-ray diffraction pattern of a powder has been implemented on a massively parallel SIMD platform developed in the frame of the PQE2000 Project. The code allows the evaluation of the diffrac...
详细信息
ISBN:
(纸本)0769500595
A code for Me simulation of X-ray diffraction pattern of a powder has been implemented on a massively parallel SIMD platform developed in the frame of the PQE2000 Project. The code allows the evaluation of the diffraction pattern of atomic-scale models of both perfectly ordered and disordered structures. The code has been wed to investigate the structures resulting from the non-equilibrium alloying process of an immiscible metallic couple (Ag-Cu).
We show that the latest version of massively parallelprocessing associative string processing architecture (System-V) is applicable for fast Monte Carlo simulation if an effective on-processor random number generator ...
详细信息
ISBN:
(纸本)0769500595
We show that the latest version of massively parallelprocessing associative string processing architecture (System-V) is applicable for fast Monte Carlo simulation if an effective on-processor random number generator is implemented. Our lagged Fibonacci generator can produce 10(8) random numbers on a processor string of 12K PE-s. The time dependent Monte Carlo algorithm of the one-dimensional non-equilibrium kinetic Ising model perfoms 80 faster than the corresponding serial algorithm on a 300 MHz UltraSparc.
This paper considers the application of bio-inspired systems in the design of a novel and highly parallel image processing tool to detect edges within conventional grey-scale images. The aim of the work is to implemen...
详细信息
ISBN:
(纸本)0769500595
This paper considers the application of bio-inspired systems in the design of a novel and highly parallel image processing tool to detect edges within conventional grey-scale images. The aim of the work is to implement a new image processing architecture through evolvable hardware that is able to adapt according to the particular images encountered The simulation of such a system through the toe of evolutionary algorithms and genetic programming is demonstrated for rite conventional image processing operation of edge detection. Results al-e presented for this system and evaluated with respect to a conventional Sobel edge detector.
A set of tasks has to be scheduled on the parallel identical processors of the clusters of a two-levels distributed memory multiprocessor, subject to precedence constraints and small intra-cluster communication delays...
详细信息
ISBN:
(纸本)0769500595
A set of tasks has to be scheduled on the parallel identical processors of the clusters of a two-levels distributed memory multiprocessor, subject to precedence constraints and small intra-cluster communication delays. The architecture model includes networks of shared memory multiprocessors. In this paper, we present a new critical-path like algorithm that finds an optimal solution to this new problem in polynomial time, if task duplication is allowed and the number of available processors is not limited. The solution found is an earliest schedule that spreads the tasks between the clusters and the processors.
The bubble algorithm evaluated in this paper assures message deadlock freedom in k-ary n-cube network without wing virtual channels. This algorithm is based both on a dimension order routing (DOR) and on a restricted ...
详细信息
ISBN:
(纸本)0769500595
The bubble algorithm evaluated in this paper assures message deadlock freedom in k-ary n-cube network without wing virtual channels. This algorithm is based both on a dimension order routing (DOR) and on a restricted injection policy extended to the dimension changes. An exhaustive comparison between the bubble mechanism and the classical deterministic virtual channels solution is presented here. For that purpose, the message router of both proposals has been designed by using VHDL descriptions and the Synopsys VLSI CAD tool. Additionally :formal models of the routers, based on colored Petri nets, have been carried out together with simulation techniques in order to assure the validation of the results and shorten the design cycle. The performance evaluation of n-dimension tori highlights the benefits of the bubble algorithm as both the temporal delay and the necessary silicon area of the message router ape reduced.
Motion planning is a fundamental problem in a number of application areas, including robotics, automation, and virtual reality! This paper describes a parallel implementation of a motion planning algorithm particularl...
详细信息
ISBN:
(纸本)0769500595
Motion planning is a fundamental problem in a number of application areas, including robotics, automation, and virtual reality! This paper describes a parallel implementation of a motion planning algorithm particularly suited for complex systems characterized by many degrees of freedom. nle implementation is based on the concurrent exploration of the search space by a randomized planner replicated on each node of the parallel architecture. All processing elements compete to obtain a solution over the entire search space in an OR-parallel fashion, Reported results refer to a low-cost cluster of PCs and an SGI Onyx2 parallel machine. The experiments emphasize the effectiveness of the approach for complex, high-dimensionality planning problems.. We believe that the approach may, be useful in other complex search problems, especially when the parallel architecture exhibits relatively high communication latency.
The GRISSLi coupling interface is a runtime library designed for the coupled computation of grid-based multidisciplinary applications. Established simulation programs, which are deleted to a single discipline, can be ...
详细信息
ISBN:
(纸本)0769500595
The GRISSLi coupling interface is a runtime library designed for the coupled computation of grid-based multidisciplinary applications. Established simulation programs, which are deleted to a single discipline, can be interconnected to a complex simulation system via the GRISSLi coupling interface without recoding the participating programs. We introduce a data model to establish the links between the (dynamic) grids and the attached coupling values in a single application program and to establish the mappings between grids and coupling values among different application programs. The user-specified consistency model is realized ruing asynchronous communication based on the MPI message passing standard Efficient communication is achieved by pre-computing an optimized communication schedule and overlapping communication with computation. Flexible coupling algorithms with adaptive grids and branches are supported in GRISSLi. We have developed a prototype implementation, which provides language bindings in ANSI C and Fortran, and runs currently on IBM RS/6000 SP Gray T3E, SGI Origin/Octane and HP-V Class. We verify the functionality and evaluate the performance of our library rising an industry relevant pilot application, the steel strip production process.
Although there exist many approaches for classification of computer architectures, no system is able to distinguish parallel computers adequately yet. The first part of this paper is a brief survey of the main charact...
详细信息
ISBN:
(纸本)0769500595
Although there exist many approaches for classification of computer architectures, no system is able to distinguish parallel computers adequately yet. The first part of this paper is a brief survey of the main characteristics of various parallel architecture concepts and several approaches of classification. We realized that the idea of defining a sharp classification system is questionable in the world of parallel computing. But we also recognized important structural similarities of computing systems. So we introduce a modeling system, called rho (recursive hierarchical objects), which allows to describe the structure Of various computing systems hierarchically. However we do not aim at a new classification system but at a scheme for modeling parallel architectures.
This at-tide aims at introducing SYD, a tool for the generation of time-stamps synchronized with the Universal Time Coordinate: the implementation includes an NTP client and is written in Java(1). Possible application...
详细信息
ISBN:
(纸本)0769500595
This at-tide aims at introducing SYD, a tool for the generation of time-stamps synchronized with the Universal Time Coordinate: the implementation includes an NTP client and is written in Java(1). Possible applications include (but are not limited to) the measurement of the performance of distributed applications, their debugging, and Me implementation of real-time requirements in distributed systems. The features that characterize SYD are its portability across platforms, the low cost in terms of network overhead, and the absence of interference with other applications running on the same host (SYD does not synchronize the system clock!);In this paper, we prove that the choice of Java does impact on the performance of the tool, when compared with an equivalent application written inC. SYD is available on the Web as public domain software.
暂无评论