Recently, parallel computing has been applied to many systems. Functional programming is suitable for parallel programming because of referential transparency and is applied to symbol processing systems and parallel d...
详细信息
ISBN:
(纸本)0769514146
Recently, parallel computing has been applied to many systems. Functional programming is suitable for parallel programming because of referential transparency and is applied to symbol processing systems and parallel database systems. Programs with some functional programming can be regarded as graphs and are processed in terms of reduction of the corresponding graph. The paper proposes fault tolerance of functional programming based on graph reduction. The proposed method stores the received graph as a message log and an erroneous task is recovered by using the checkpoint and the stored graph. Computer simulations reveal that the time overhead of the proposed method is small. If the checkpoint interval is 30 seconds and the number of tasks is 3, for example, the time overhead is less than 10%.
In this paper we describe an instrumentation environment for the performance analysis and visualization of parallel applications written in JOMP, an OpenMP-like interface for Java. The environment includes two complem...
详细信息
In this paper we describe an instrumentation environment for the performance analysis and visualization of parallel applications written in JOMP, an OpenMP-like interface for Java. The environment includes two complementary approaches. The first one has been designed to provide a detailed analysis of the parallel behavior at the JOMP programming model level. At this level, the user is faced with parallel, work-sharing and synchronization constructs, which are the core of JOMP. The second mechanism has been designed to support an in-depth analysis of the threaded execution inside the Java virtual machine (JVM). At this level of analysis, the user is faced with the supporting threads layer monitors and conditional variables. The paper discusses the implementation of both mechanisms and evaluates the overhead incurred by them.
Cluster computing is becoming increasingly popular among users of parallel and distributed applications. However since few clusters are solely dedicated to run individual user jobs, it is necessary to coordinate those...
详细信息
ISBN:
(纸本)0769515282
Cluster computing is becoming increasingly popular among users of parallel and distributed applications. However since few clusters are solely dedicated to run individual user jobs, it is necessary to coordinate those jobs among independently administered clusters effectively and with less user interaction. For this purpose, we have developed an infrastructure for inter-cluster job coordination using Voyager mobile agents. Once a user submits a job request, which is converted into XML, a mobile agent searches for the most available cluster establishes a job execution environment, executes the job, and reports its results back to the client. Using Voyager's dynamic aggregation feature, we also plan to make such job coordination agents to evolve themselves as to be suitable to be deployed in dynamic cluster environments. This paper presents the design principles and the latest implementation status of our infrastructure as well as advanced services using evolvable agents.
Provides a fuzzy programming expected value model for a fuzzy parallel machine scheduling problem. A fuzzy simulation-based genetic algorithm is designed to solve the proposed model. Moreover, a numerical experiment i...
详细信息
Provides a fuzzy programming expected value model for a fuzzy parallel machine scheduling problem. A fuzzy simulation-based genetic algorithm is designed to solve the proposed model. Moreover, a numerical experiment is provided to illustrate the computational efficiency of the proposed approach.
We present two generic parallel skeletons for the tabu search method-a well known meta-heuristic for approximately solving combinatorial optimization problems. The first skeleton is based on independent runs while the...
详细信息
ISBN:
(纸本)0769511538
We present two generic parallel skeletons for the tabu search method-a well known meta-heuristic for approximately solving combinatorial optimization problems. The first skeleton is based on independent runs while the second in the classical master-slave model. Our starting point is the design and implementation of a sequential skeleton that is used later as basis for the two parallel skeletons. Both skeletons provide the user with the following: a permit to obtain parallel implementations of the tabu search method for concrete combinatorial optimization problems from existing sequential implementations; there is no need for the user to know either parallel programming or communication libraries; and the parallel implementation of tabu search for a concrete problem is obtained automatically from a sequential implementation of tabu search for the problem. The skeletons, however, require from the user a sequential instantiation of the tabu search method for the problem at hand. The skeletons are implemented in C++ using MPI as the communication library and offer genericity, flexibility, component reuse, robustness and time savings. We have instantiated the two skeletons for the 0-1 multidimensional knapsack problem, among others, for which we report computational results.
An embedded flash memory module has 1.2 V read capability and a 1.5 V program/erase capability. The flash cell is 2-transistor FN-NOR in a 0.181 /spl mu/m logic process. Design techniques improve observability and red...
详细信息
ISBN:
(纸本)0780366085
An embedded flash memory module has 1.2 V read capability and a 1.5 V program/erase capability. The flash cell is 2-transistor FN-NOR in a 0.181 /spl mu/m logic process. Design techniques improve observability and reduce test time.
A new two-stage method for programming flash memory cells in an analog record and playback system is reported. The algorithm has a coarse and a fine portion to speed up the recording process. This method is used to st...
详细信息
ISBN:
(纸本)0780364120
A new two-stage method for programming flash memory cells in an analog record and playback system is reported. The algorithm has a coarse and a fine portion to speed up the recording process. This method is used to store 16 minutes of audio signals, at a sampling rate of 4 kHz (or 8 minutes at 8 kHz) in a 4 Mb flash array with a resolution approaching 8-bit.
暂无评论