We propose a new parallel, noncollective I/O strategy called Distant I/O that targets clustered computer systems in which disks are attached to compute nodes. Distant I/O allows one-sided access to remote secondary st...
详细信息
ISBN:
(纸本)0818685794
We propose a new parallel, noncollective I/O strategy called Distant I/O that targets clustered computer systems in which disks are attached to compute nodes. Distant I/O allows one-sided access to remote secondary storage without installing server processes or daemons on remote compute nodes. We implemented this model using Active Messages and demonstrated its performance advantages over the PIOFS parallel filesystem for an I/O-intensive parallel application on the IBM SP.
Heterogeneous distributedcomputing has traditionally been a problematic undertaking which increases in complexity as heterogeneity increases. the recent advent of Java has made heterogeneous computing a fairly straig...
详细信息
ISBN:
(纸本)0818685794
Heterogeneous distributedcomputing has traditionally been a problematic undertaking which increases in complexity as heterogeneity increases. the recent advent of Java has made heterogeneous computing a fairly straight forward task. Nevertheless, many researchers have not considered the use of Java in a mainstream parallel programming environment. the performance of Java just-in-time compilers currently approaches optimized C for many applications, making Java a serious contender for supercomputing application development. this paper presents DOGMA-a new Java based system which enables parallelcomputing on, heterogeneous computers. DOGMA provides support for dedicated clusters as well as idle workstations through the use of a web based browse-in feature or the DOGMA screen saver, which improves workstation utilization. DOGMA supports parallel programming in both a traditional message passing form and a novel object-oriented approach. this research provides a unified environment for developing high performance supercomputing applications on heterogeneous systems.
Multi-comparand associative processors are efficient in parallel processing of complex search problems that arise from many application areas including computational geometry, graph theory and list/matrix computations...
详细信息
ISBN:
(纸本)9780769534725
Multi-comparand associative processors are efficient in parallel processing of complex search problems that arise from many application areas including computational geometry, graph theory and list/matrix computations. In this paper we report new FPGA implementations of a multi-comparand multi-search associative processor the architecture of the processor working in a combined bit-serial/bit-parallel word-parallel mode and its functions arc, described. then, several implementations of associative processors in VHDL, using Xilinx Foundation ISE software and Digilent development boards with Xilinx FPGA devices are reported. Parameters of the implemented FPGA processors are presented and discussed.
A pancake graph was proposed as a topology for interconnection networks of parallel computers, and it has a merit that it can connect many nodes with small diameter and small degree. However the number of nodes in a p...
详细信息
ISBN:
(纸本)076952611X
A pancake graph was proposed as a topology for interconnection networks of parallel computers, and it has a merit that it can connect many nodes with small diameter and small degree. However the number of nodes in a pancake graph must be equal to the factorial of an integer, which causes a demerit that it lacks incremental expandability. Hence, in this paper, we propose a graph, an incomplete pancake graph, that is obtained by deleting some of sub graphs of a pancake graph, and we give algorithms for the problems of the simple routing, the Hamiltonian cycle, the Hamiltonian path, and the container in incomplete pancake graphs.
We propose to address collective communications in distributed components through collective interfaces. Collective interfaces handle data distribution, parallelism and synchronization, and they expose collective beha...
详细信息
ISBN:
(纸本)9780769528335
We propose to address collective communications in distributed components through collective interfaces. Collective interfaces handle data distribution, parallelism and synchronization, and they expose collective behaviors in the definition of components. We show, as an illustration, that collective interfaces allow the encoding of SPMD programming in a better structured and less error prone ways We verify the scalability and performance of collective interfaces in an experiment on up to 100 machines.
In this paper we describe hardware implementations of generators of combinatorial objects. For implementation several systolic algorithms were selected that generate combinatorial configurations in a linear array mode...
详细信息
ISBN:
(纸本)9780769534725
In this paper we describe hardware implementations of generators of combinatorial objects. For implementation several systolic algorithms were selected that generate combinatorial configurations in a linear array model. the algorithms generate such objects as combinations, combinations with repetitions, t-ary trees, partitions, and variations with repetitions. the generators were implemented in VHLD with Xilinx Foundation ISE, software and tested on Digilent development boards with Xilinx FPGAs. Implementation data obtained for various input parameters and FPGA devices are given.
the proceedings contain 17 papers. the topics discussed include: a selection theory and methodology for heterogeneous supercomputing;partitioning problems in heterogeneous computer systems;experiments with a task part...
ISBN:
(纸本)0818635312
the proceedings contain 17 papers. the topics discussed include: a selection theory and methodology for heterogeneous supercomputing;partitioning problems in heterogeneous computer systems;experiments with a task partitioning model for heterogeneous computing;heuristics for mapping parallel computations to heterogeneous parallel architectures;load distribution optimization in heterogeneous multiple processor systems;problem representations for an automatic mapping algorithm on heterogeneous processing environments;towards a virtual multicomputer;developing applications for a heterogeneous computing environment;heterogeneous by design: an environment for exploiting heterogeneity;a case study in metacomputing: distributed simulations of mixing in turbulent convection;and design of a heterogeneous parallel processing system for beam forming.
A detailed description of transport phenomena in rear-life complex geometries is important in several areas of reservoir and environmental engineering. this paper presents numerical issues related to the application o...
详细信息
A detailed description of transport phenomena in rear-life complex geometries is important in several areas of reservoir and environmental engineering. this paper presents numerical issues related to the application of high-performance computing in this field. the specific example treated comes from reservoir engineering. A distributedparallel environment is developed based upon domain decomposition. We discuss boththe iterative properties and the parallel performance on an IBM-SP2. It will be shown that the numerical approach leads to an attractive tool for treating real-life large-scale problems.
We investigate the relative computational power of parallel models with shared memory. Based on,feasibility considerations present in the literature, we split these models into "lightweight" and "heavyw...
详细信息
ISBN:
(纸本)9780769534725
We investigate the relative computational power of parallel models with shared memory. Based on,feasibility considerations present in the literature, we split these models into "lightweight" and "heavyweight," and then find that the heavyweight class is strictly more powerful than the lightweight class, as expected. On the other hand, we contradict the long held belief that the heavyweight models (namely, the Combining CRCW PRAM and the BSR) form a hierarchy, showing that they are identical in computational power with each other We thus introduce the BSR into the family of practically meaningful massively parallel models. this result also has significant implications in the area of real-time computations.
暂无评论