V. Getov, S. Flynn-Hummel, and S. Mintchev. High-performance Parallel Programming in Java: Exploiting Native Libraries. In ACM 1998 Workshop on Java for High-performance Network Computing, Feb. 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....is against Java s philosophy. It also prevents future optimizations arising from advances in data structures and algorithms. The implementations of the BLAS and other aggregate operations, and various elemental operations can be very sophisticated (and built upon existing libraries using JNI [4, 8, 19]) for applications requiring high performance. Alternatively, a very simple implementation is also possible, leading to a small Array package implementation and allowing it to be downloaded more easily and to fit on clients with much less memory than high performance compute servers. 11 To ....

V. Getov, S. Flynn-Hummel, and S. Mintchev. High-performance parallel programming in Java: Exploiting native libraries. In ACM

....para computaci on de alto rendimiento [5] Java provee un ambiente de ejecuci on orientado a objetos, port atil y seguro, para el desarrollo de aplicaciones. Estudios recientes muestran que Java tiene el potencial para lograr el rendimiento de lenguajes de programaci on m as tradicionales [6] [7], 8] suma (Scienti c Ubiquitous Metacomputing Architecture) 9] es un sistema de metacomputaci on que ejecuta bytecode de Java. Est a implementado como un conjunto de clases corba, siguiendo un modelo de metacomputaci on de tres niveles. Los agentes de ejecuci on de suma, es decir, las ....

V. Getov, S. Flynn-Hummel, and S. Mintchev, \High-performance parallel programming in Java: Exploiting native libraries," in ACM

....code generated by the Manta compiler. 6. RELATED WORK Java is increasingly recognized as a suitable platform for highperformance computing. The driving force in high performance Java is the Java Grande Forum (www.javagrande.org) Many research projects investigate parallel programming in Java [7, 8, 12, 22]. Efficient communication mechanisms are a vital building block for high performance Java [15] One approach to sharing objects between parallel processes is to use either object caching [2, 25] or object replication [16] However, neither replication nor caching of objects provides the ....

V. Getov, S. Flynn-Hummel, and S. Mintchev. High-performance Parallel Programming in Java: Exploiting Native Libraries. In ACM

....traffic by up to 85 via event filtering, with consequent additional savings in the processing requirements for events received by clients. Even higher savings are experienced when using event differencing. 6. Related Work There has been a considerable work on high performance messaging in Java[21][22] 24] 25] Some of these systems are native code libraries with Java interfaces [21] 24] while pure Java systems have performance limits, especially concerning roundtrip latencies[23] 25] Jini[12] s distributed event specification does not rely on RMI, but most current implementations of ....

....requirements for events received by clients. Even higher savings are experienced when using event differencing. 6. Related Work There has been a considerable work on high performance messaging in Java[21] 22] 24] 25] Some of these systems are native code libraries with Java interfaces [21][24] while pure Java systems have performance limits, especially concerning roundtrip latencies[23] 25] Jini[12] s distributed event specification does not rely on RMI, but most current implementations of this specification are based on unicast RMI, which, as we demonstrated, has performance ....

V. Getov, S. Flynn-Hummel, and S. Mintchev. "High-Performance parallel programming in Java: Exploiting native libraries ". ACM

....increasing interest in using Java as a language for high performance computing [1] Java provides a portable, secure, clean object oriented environment for application development. Recent results have shown that Java has the potential to attain the performance of traditional scientific languages [2, 3, 4]. On the other hand, the access to distributed high performance computing facilities through a metacomputing system has recently gained considerable acceptance [5] A metacomputing system allows uniform access to heterogeneous resources, including high performance computers. This is achieved by ....

V. Getov, S. Flynn-Hummel, and S. Mintchev. High-performance parallel programming in Java: Exploiting native libraries. In ACM

....Ubiquitous Metacomputing Architecture) Her99] is a metacomputing system for execution of JAVA bytecode, with additional support for scientific computing development. JAVA was selected because of its portability and the increasing interest in this language for high performance computing [Get98][Bak98] JAVA provides a portable, secure and clean object oriented environment for application development. The main goal of SUMA is to extend the Java Virtual Machine model to provide seamless access to distributed high performance resources. SUMA project focuses on the development of the ....

V. Getov, S. Flynn-Hummel and S. Mintchev. "High Performance Parallel Programming in JAVA: Exploiting Native Libraries". In ACM

....computing in Java has been recognized by many authors. In particular, 4, 5, 29] describe projects to develop such libraries entirely in Java. In addition, there are approaches in which access to existing numerical libraries (typically coded in C and Fortran) is provided to Java applications [3, 9, 16]. All of these approaches shift the burden of delivering high performance entirely to the libraries. 13 JDK116 HPCJ HPCJ SE BOUNDS FMA FORTRAN 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 17.5 26.5 3.6 111.0 205.0 227.0 Code version Performance of MATMUL on RS 6000 590 (Mflops) ....

V. Getov, S. Flynn-Hummel, and S. Mintchev. High-performance parallel programming in Java: Exploiting native libraries. In ACM

....is increasing interest in using Java as a language for high performance computing [14] Java provides a portable, secure, clean object oriented environment for application development. Recent results show that Java has also the potential to attain the performance of traditional scienti c languages [16, 12, 2]. suma (Scienti c Ubiquitous Metacomputing Architecture) 13] is a metacomputing system that executes Java bytecode. It is implemented as a set of corba classes, following a three tier metacomputing system model. suma execution agents (actual platforms where program execution takes place) may o er ....

V. Getov, S. Flynn-Hummel, and S. Mintchev. High-performance parallel programming in Java: Exploiting native libraries. In ACM

....rendered pixels to CPU 0 by individual messages. The speedups achieved by mpiJava and by CCJ are almost identical. 6 Related work The driving force in high performance Java is the Java Grande Forum (www.javagrande.org) There are also many other research projects for parallel programming in Java [1, 6, 7, 14, 16, 25]. Most of these systems, however, do not support collective communication. Taco [24] is a C template library that implements collective operations, however without exploiting MPI s concept of collective invocation by the participating processes. JavaNOW [27] implements some of MPI s collective ....

V. Getov, S. Flynn-Hummel, and S. Mintchev. High-performance Parallel Programming in Java: Exploiting Native Libraries. In ACM

....possibilities inherent in Java s use of threads and a relaxed memory model. There are a large number of parallel Java e orts that connect multiple Java virtual machines by utilizing Java s Remote Method Invocation facility (e.g. 1 5] or by grafting an existing message passing library (e.g. [6,7]) onto Java. In our work we view a cluster as executing a single Java virtual machine. The separate nodes of the cluster are hidden from the programmer and are simply resources for executing Java threads with true parallelism. The separate memories of the nodes are also hidden from the programmer ....

V. Getov, S. F. Hummel, S. Mintchev, High-performance parallel programming in Java: exploiting native libraries, Concurrency: Practice and Experience 10 (11-13) (1998) 863872.

....17: Speedup for the QR application 5. RELATED WORK Java is increasingly recognized as a suitable platform for highperformance computing. The driving force in high performance Java is the Java Grande Forum (www.javagrande.org) Many research projects investigate parallel programming in Java [1, 9, 10, 16, 19, 28]. Efficient communication mechanisms have been recognized as a vital building block for high performance Java [21] Most existing systems, however, do not support efficient group communication. In our previous work on parallel Java, we implemented several applications based on RMI and RepMI ....

V. Getov, S. Flynn-Hummel, and S. Mintchev. High-performance Parallel Programming in Java: Exploiting Native Libraries. In ACM 1998 workshop on Java for High-performance network computing, Feb. 1998.

.... by the Java Native Interface (JNI) 7] The JNI allows to develop the parts of the framework that are most critical to efficient application execution in ANSI C language and to introduce into them the desired level of architecture dependent optimization at the cost of increased development effort [8][9] The use of native code requires a different implementation of a service for each type of heterogeneous computational resource that need to deliver that service. This fact implies a development effort multiplied for each plug in including native code. However, if a version of the plug in for ....

V. Getov, S. Flynn-Hummel and S. Mintchev, High Performance Parallel Programming in Java: Exploiting Native Libraries, to appear on Concurrency: Practice and Experience, 1998.

....a Java language binding for MPI, version 1.1 [6, 8] Our current approach is a relatively direct transcription of standard MPI bindings, but Java object serialization introduces new possibilities for passing compound objects. Similar projects on Java MPI bindings are in progress elsewhere [26, 15]. 2.4 Integration of thread based single Java VM and multiVM data parallel Our language model is primarily aimed at distributed memory computers, including networks of workstations or PCs. Clearly the Java version of HPspmd also holds special promise in the domain of metacomputing targeting ....

Vladimir Getov, Susan Flynn-Hummel, and Sava Mintchev. Highperformance parallel programming in java: Exploiting native libraries. In ACM workshop on Java for High-performance Network Computing, 1998. To appear in Concurrency: Practice and Experience.

....two of the current authors on Java MPI bindings is reported in [3, 4] In those papers we compared various approaches to parallel programming in Java, including socket programming and MPI programming. A comparable approach to creating full Java MPI interfaces has been taken by Getov and Mintchev [17, 13]. In their work Java wrappers were automatically generated from the C MPI header. This eases the implementation work, but does not lead to a fully object oriented API. A subset of MPI is implemented in the DOGMA system for Java based parallel programming [16] MPI Software Technology, Inc have ....

Vladimir Getov, Susan Flynn-Hummel, and Sava Mintchev. Highperformance parallel programming in Java: Exploiting native libraries. In ACM 1998 Workshop on Java for High-Performance Network Computing. Palo Alto, February 1998, Concurrency: Practice and Experience, 1998. To appear.

....as the typical parallel machine today. There are a large number of e orts trying to provide support for a distributed execution of Java programs on clusters, mainly by using Java s remote method invocation facility (e.g. 7, 8, 13, 14] or by grafting an existing message passing library (e.g. [10, 11]) onto Java, in order to connect multiple Java Virtual Machines. In contrast, we view a cluster as executing a single Java Virtual Machine (JVM) where the nodes are resources for the distributed execution of Java threads with true concurrency. Distribution must be transparent to the programmer ....

V. Getov, S. Flynn-Hummell, and S. Mintchev. High-performance parallel programming in Java: Exploiting native libraries. In Proceedings of the ACM 1998 Workshop on Java for High-Performance Network Computing, pages 4554, Palo Alto, California, February 1998.

....Java Virtual Machines and libraries still have some performance limitations, but since this seems certain to continually improve it is worthwhile to consider the use of Java for high performance message passing systems now. This is an active area with several groups attempting implementations [2, 11, 21]. Developers seem to split into two major camps those attempting to provide Java bindings for the MPI [13] interface and those who have implemented non conforming systems. We find good arguments for both these approaches. Considerable work for message passing in a procedural or imperative style ....

Vladimir Getov, Susan Flynn-Hummel, and Sava Mintchev. High-performance parallel programming in Java: Exploiting native libraries. In 1998 ACM Workshop on Java for High Performance Network Computing, 1998.

....Java Virtual Machines and libraries still have some performance limitations, but since this seems certain to continually improve it is worthwhile to consider the use of Java for high performance message passing systems now. This is an active area with several groups attempting implementations [2, 11, 21]. Developers seem to split into two 1 major camps those attempting to provide Java bindings for the MPI [13] interface and those who have implemented non conforming systems. We nd good arguments for both these approaches. Considerable work for message passing in a procedural or imperative ....

Vladimir Getov, Susan Flynn-Hummel, and Sava Mintchev. High-performance parallel programming in Java: Exploiting native libraries. In 1998 ACM Workshop on Java for High Performance Network Computing, 1998.

....on two dual processor (200MHz Ultrasparc) Solaris workstations with 256Mbytes of DRAM. The mpiJava overhead under WMPI was about 100 ms whereas the MPICH overhead was between 250 and 300 s, that is signi cantly lower. Java to C Interface Generator (JCI) The Java to C Interface Generator (JCI) [21] has been developed at IBM T. J. Watson Research Center. JCI is a Java interface generator to C similar to Java Native Interface. Although JCI is not intended to be an environment for high performance computing, Java programmers can use JCI to bene t from native libraries such as MPI, to improve ....

Getov V, Flynn-Hummel S, Mintchev S. 1998. High-Performance Parallel Programming in Java: Exploiting Native Libraries. In Proceedings of the ACM 1998 Workshop on Java for High Performance Network Computing; Palo Alto, California, USA, February 1998.

....0 10 20 30 40 50 60 0 10 20 30 40 50 60 cpus LEQ RMI CCJ Figure 10: Speedup for the LEQ application 5. RELATED WORK The driving force in high performance Java is the Java Grande Forum (www.javagrande.org) There are also many other research projects for parallel programming in Java [1, 6, 7, 14, 16, 25]. Most of these systems, however, do not support collective communication. Taco [24] is a C template library that implements collective operations, however without exploiting MPI s concept of collective invocation by the participating processes. JavaNOW [26] implements some of MPI s collective ....

V. Getov, S. Flynn-Hummel, and S. Mintchev. High-performance Parallel Programming in Java: Exploiting Native Libraries. In ACM 1998 workshop on Java for High-performance network computing, Feb. 1998.

....develop plug ins; in fact a plug in is not required to be a monolithic entity and may depend on any number of other classes. Besides, if efficiency requirements are too strict to be fulfilled by pure Java code, it is possible to exploit the JNI to either manually wrap native code into Java classes [7] or to exploit one of the tools that perform automatic wrapping of legacy code [8] Harness provides a flat mapping of Internet host names onto computational resource names. However, the framework allows each user to adopt his or her own computational resource naming and grouping scheme as long ....

V. Getov, S. Flynn-Hummel and S. Mintchev, High Performance Parallel Programming in Java: Exploiting Native Libraries, to appear on Concurrency: Practice and Experience, 1998.

....with the project specifications, modification records, test procedures, test results, marketing information etc. It is in this environment that we see the major benefits of integrating Java, with its support for Webbased computing, with traditional high performance programming using Fortran and C (Getov et al., 1998). This vision is of a new type of multi language multi paradigm programming environment for high performance applications which makes use of innovative Web based and GUI technologies provided by Java and provides a high performance and simple API and a high level of portability and flexibility for ....

Getov, V., Hummel, S.F. and Mintchev, S., 1998, "High-Performance Parallel Programming in Java: Exploiting Native Libraries", Concurrency: Practice & Experience, to appear (1998).

.... by the Java Native Interface (JNI) 7] The JNI allows to develop the parts of the framework that are most critical to efficient application execution in ANSI C language and to introduce into them the desired level of architecture dependent optimization at the cost of increased development effort [8][9] The use of native code requires a different implementation of a service for each type of heterogeneous computational resource that need to deliver that service. This fact implies a development effort multiplied for each plug in including native code. However, if a version of the plug in for ....

V. Getov, S. Flynn-Hummel and S. Mintchev, High Performance Parallel Programming in Java: Exploiting Native Libraries, to appear on Concurrency: Practice and Experience, 1998.

....plug ins. In fact, a plug in is not required to be a monolithic entity, on the contrary it may depend on any number of other classes. Besides, if the efficiency requirements are too strict to be fulfilled by pure Java code, the JNI allows developers to manually wrap native code into Java classes [7] while tools to achieve automatic wrapping of legacy code are currently studied [8] 3.3 Harness DVMs Topology and Protocols The Harness metacomputing framework allows the definition and establishment of DVMs. The kernel level services of a Harness DVM are delivered by a distributed system ....

V. Getov, S. Flynn-Hummel and S. Mintchev, High Performance Parallel Programming in Java: Exploiting Native Libraries, to appear on Concurrency: Practice and Experience, 1998.

No context found.

V. Getov, S. Flynn-Hummel, and S. Mintchev. High-performance Parallel Programming in Java: Exploiting Native Libraries. In ACM 1998 Workshop on Java for High-performance Network Computing, Feb. 1998.

No context found.

Getov, V., Flynn-Hummel, S., and Mintchev, S. "High-Performance Parallel Programming in Java: Exploiting Native Libraries," Harrow, England and Yorktown Heights, NY: October, 1997.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC