B. Chapman, H. Zima, M. Haines, P. Mehrotra, and J. V. Rosendale. Opus: A coordination language for multidisciplinary applications. Scientific Programming, 6(4):345--362, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....as adaptive mesh refinements, answers become better (e.g. more precise) over time. It can be very important to extract results as progress is made toward an acceptable solution for steering, early termination, or other reasons. A typical application of this type is the design of an aircraft wing [2, 3, 5, 12]. Certain scientific and mathematical processes, like Newtons method of successive approximation for roots of an equation, can also utilize this type of extraction [4] The traditional RPC result extraction model does not lend itself well to progressive extractions. 1.2.4 Summary Decomposition ....

B. Chapman, et al., Opus: A Coordination Language for Multidisciplinary Applications, ICASE Tech Report 97-30, Jun 1997.

....the topology it is ready to process a message, and then forwards the results to the father node. Note the declarative Prolog like style for implementing Points of View. Note also the use of predefined communication view primitives (such as csMsgSendAssDep) throughout the code. 3.2. 16 Opus Opus ([19]) is e#ectively a coordination superlanguage on top of High Performance Fortran (HPF) for which it was designed, for the purpose of coordinating concurrent execution of several dataparallel components. Interaction between concurrently executing tasks is achieved via the ShareD Abstraction (SDA) a ....

B. Chapman, M. Haines, P. Mehrotra, J. V. Rosendale and H. Zima, "Opus: A Coordination Language for Multidisciplinary Applications", Scientific Programming 6 (2), 1997.

....at a high level include LINDA [3] Strand [29] CC [11] and Fortran M [28] Recently, a major research effort has been put on the issue of integrating task and data parallelism. Examples include Orca [7] Fx [76, 77] Approved Extensions for Data and Task Parallelism in HPF 2. 0 [25] and Opus [12]. In this work we develop compiler optimizations for reducing remappings in data parallel languages. Although we use the de facto standard HPF as base language, the problem is relevant for other data parallel languages, as well, if data remappings are supported. In the following section we ....

....which is constructed by putting modules from a variety of related scientific disciplines together. Each module owns a best data mapping. Hence, remappings are required between the individual phases. Similar issues arise in multidisciplinary optimization. For example in modern aircraft design [12], one has a variety of interacting disciplines: aerodynamics, propulsion, structural analysis and design, and so forth. The various disciplines form a single application with an overall optimization loop of the design process whereby remappings might be necessary between the disciplines. The ....

B. Chapman, H. Zima, M. Haines, P. Mehrotra, and J. Van Rosendale. OPUS: A Coordination Language for Multidisciplinary Applications. Technical Report TR 95-6, Institute for Software Technology and Parallel Systems, University of Vienna, Vienna, Austria, Oktober 1995.

....messagepassing library such as MPI [10] Low level message passing primitives are provided which allow the exchange of distributed data among concurrent HPF tasks. The integration of task and data parallelism has also been proposed within an object oriented coordination language such as Opus [8]. By using a syntax similar to that of HPF, Opus programmers can de ne classes of objects which encapsulate distributed data and data parallel methods. In this paper we present COLT HPF (COordination Layer for Tasks expressed in HPF) a portable coordination layer for HPF tasks. COLT HPF is ....

B.Chapman et al. Opus: a Coordination Language for Multidisciplinary Applications. Scientic Programming, 6(2), April 1997.

....patterns [3,7,12] The capability of integrating task and data parallelism into a single framework is the subject of many proposals since it allows the This work has been partially supported by HPPC SEA contract number EU1063. number of addressable applications to be considerably enlarged [11,22,10,8,6,15]. Also the latest version of the HPF standard, HPF 2.0, provides extensions which allows restricted forms of task parallelism to be exploited [16] Unfortunately, the extensions are not suitable for expressing complex interactions among asynchronous tasks as required by multidisciplinary ....

....message passing library such as MPI [10] Low level message passing primitives are provided which allow the exchange of distributed data among concurrent HPF tasks. The integration of task and data parallelism has also been proposed within an object oriented coordination language such as Opus [8]. By using a syntax similar to that of HPF, Opus programmers can de ne classes of objects which encapsulate distributed data and data parallel methods. In this paper we present COLT HPF (COordination Layer for Tasks expressed in HPF) a portable coordination communication layer for HPF tasks. ....

[Article contains additional citation context not shown here]

B.Chapman et al. Opus: a Coordination Language for Multidisciplinary Applications. Scientic Programming, 6(2), April 1997.

....adaptive mesh refinements, answers become better (e.g. more precise) over time. It can be very important to extract results as progress is made toward an acceptable solution for steering, early termination, or other reasons. A typical application of this type is the design of an aircraft wing [2, 3, 5, 12]. Certain scientific and mathematical processes, like Newton s method of successive approximation for roots of an equation, can also utilize this type of extraction [4] The traditional RPC result extraction model does not lend itself well to progressive extractions. Decomposition of the ....

B. Chapman, M. Haines, P. Mehrotra, H. Zima and J. Van Rosendale, "Opus: A Coordination Language for Multidisciplinary Applications," ICASE Technical Report 97-30, June 1997.

....are provided which allow the exchange of distributed data among concurrent HPF tasks. We believe that the exploitation of a message passing interface is too much low level for e ective usability. Other approaches extend an objectoriented coordination languages with data parallelism (e.g. Opus [4]) These models show a greater dynamism with respect to taskHPF and for this reason delegate much of the optimization e ort to the run time. In addition, they do not support the use of common organization patterns. Finally, patterns have been proposed to coordinate data parallel modules in ....

B. Chapman et al. Opus: a Coordination Language for Multidisciplinary Applications. Scientic Programming, 6(2), April 1997.

....data parallel potential (such as the simulation module of Section 3.2) may be employed in a task parallel manner. HPF provides some basic support for task parallelism which is specified in the Approved Extensions. More advanced task parallel features are offered by the coordination language Opus [8, 9] which provides an object based extension to HPF supporting the coordination of data parallel HPF tasks. A detailed description of Opus is beyond the scope of this paper and the interested reader is referred to [8, 9, 17 19] In the remainder of this section we summarize the tasking support of ....

....advanced task parallel features are offered by the coordination language Opus [8, 9] which provides an object based extension to HPF supporting the coordination of data parallel HPF tasks. A detailed description of Opus is beyond the scope of this paper and the interested reader is referred to [8, 9, 17 19]. In the remainder of this section we summarize the tasking support of HPF and its realization within VFC. Language Features. Task parallelism can be achieved by mapping different computations onto different subsets of a processor array. As a consequence of such a mapping, specific statements are ....

B. Chapman, M. Haines, P. Mehrotra, J. Van Rosendale, and H. Zima. OPUS: A Coordination Language for Multidisciplinary Applications. Scientific Programming, 6/9:345--362, Winter 1997.

....support high performance parallel computing. In particular, it is a non trivial task to integrate parallel applications, written e.g. in MPI or HPF, efficiently into distributed Java or Corba applications. In this paper we present OpusJava, a Java [12] interface to the coordination language Opus [6, 7] that provides a high level, comprehensive model for DHPC. OpusJava combines the strengths of Opus (in particular the integration of HPF [16] style data parallelism with flexible task parallel features) with the strengths of Java (in particular the dynamic networking features) OpusJava provides a ....

....component interaction, hybrid forms of parallelism, and synchronization at a high level. An OpusJava implementation of this scenario is presented in Section 5, after an introduction to the salient features of Opus and OpusJava in the next two sections. 4 3 Overview of the Opus Language Opus [6, 7] introduces a small set of extensions to HPF [16] designed to coordinate an efficient parallel execution of multiple, independent HPF program modules. At the heart of these extensions is an abstract data type called a ShareD Abstraction, or SDA, whose purpose is to provide a means for ....

B. Chapman, M. Haines, P. Mehrotra, J. Van Rosendale, and H. Zima. OPUS: A Coordination Language for Multidisciplinary Applications. Scientific Programming, 6/9:345--362, Winter 1997.

....with a non rectilinear iteration space. Furthermore, the support for distributed arrays of derived types, as well as the distribution of components of derived types, will be extended. Task parallel features based on on directives and task regions and coordination mechanism as proposed in OPUS [12] are currently being implemented. For an efficient handling of such features and for other optimizations we plan to utilize one sided communication primitives of MPI 2 [20] ....

B. Chapman, H. Zima, M. Haines, P. Mehrotra, J. Van Rosendale. OPUS: A Coordination Language for Multidisciplinary Applications. Journal of Scientific Programming, 1995.

.... , the complexity of simulations being tackled by scientists and engineers is increasing exponentially. In order to utilize the available architectures efficiently, several levels of parallelism need to be exploited by such simulations. We recently introduced the coordination Language Opus [4, 5] which allows a high level management of data parallel tasks. Its central concept is the shared abstraction (SDA) which generalizes Fortran 90 HPF modules using an object based approach and imposing monitor semantics. SDAs can be internally data parallel while task parallelism is exploited ....

B. Chapman, M. Haines, P. Mehrotra, J. Van Rosendale, and H. Zima. OPUS: A Coordination Language for Multidisciplinary Applications. Scientific Programming, 6/9:345-- 362, Winter 1997.

.... few processors, while most processors are reserved for more scalable parts of the application [2, 7] The capability of integrating task and data parallelism into a single framework is the subject of many proposals since it allows the number of addressable applications to be considerably enlarged [6, 19, 5, 3, 1, 10]. Also the latest version of the HPF standard, HPF 2.0, proposes an unintuitive and in exible approach to express pipelining among HPF tasks [11] A di erent approach regards the design of an HPF binding for a standard message passing library such as MPI [5] The integration of task and data ....

.... approach to express pipelining among HPF tasks [11] A di erent approach regards the design of an HPF binding for a standard message passing library such as MPI [5] The integration of task and data parallelism has also been proposed within an object oriented coordination language such as Opus [3]. By using a syntax similar to that of HPF, Opus programmers can de ne classes of objects which encapsulate distributed data and data parallel methods. In this paper we present COLT HPF (COordination Layer for Tasks expressed in HPF) a portable coordination communication layer for HPF tasks. COLT ....

B.Chapman et al. Opus: a Coordination Language for Multidisciplinary Applications. Scientic Programming, 6(2), April 1997.

....of future research. The Coordination Language Opus The Coordination Language Opus (or simply Opus in the following) has been designed in a joint effort of ICASE, NASA Langley Research Center, and the Institute for Software Science to support the development of multidisciplinary applications (see [54, 55, 133, 138]) In this chapter we present the basic design of Opus, which we refine and augment in some minor aspects, in the context of multidisciplinary applications. Issues in the context of more general DHPC applications are sketched in Section 2.7 and elaborated in Chapter 3. Multidisciplinary ....

....for DHPC focusing on their support for three major issues in DHPC: heterogeneity, interoperability, and hybrid parallelism. 8.1 Contributions The primary contributions of this thesis are summarized below. ffl Extensions of the Opus language We have refined and extended the Opus language [55] to meet the key requirements of DHPC. These extensions include: The SDA concept has been refined to provide more flexibility with respect to SDA creation, termination, and synchronization. Specifically, we introduced new intrinsic methods for SDA creation status inquiries (CREATED) immediate ....

B. Chapman, M. Haines, P. Mehrotra, J. Van Rosendale, and H. Zima. OPUS: A Coordination Language for Multidisciplinary Applications. Scientific Programming, 6(9):345--362, Winter 1997.

....the past decade in high level language design for distributed memory architectures and clusters of SMPs, as well as the related compiler and runtime technology. We mention only a few among the many relevant references. In particular, languages such as Kali, Vienna Fortran, Fortran D, HPF, and Opus [26, 9, 14, 21, 10], and compilation systems such as SUPERB, the Fortran D compiler, Paradigm, and the Vienna Fortran Compiler [35, 14, 4, 5] as well as the huge amount of work on irregular problems have strongly influenced the macroserver design. Macroservers address some weaknesses of existing systems (in ....

....project [17, 18] they were conceptualized in the HTMT Petaflops architecture project [31] Another important contribution of the DIVA project is its direct management of pointers for on chip manipulation of irregular data structures. An important precursor of macroservers is the Opus language [10] developed jointly at ICASE, NASA Langley Research Center, and the University of Vienna. Opus provides a hybrid programming model for expressing multidisciplinary applications focusing on shared abstractions (SDAs) which establish an objectbased layer on top of HPF. Other languages contributing ....

B. Chapman, M. Haines, P. Mehrotra, J. Van Rosendale, and H. Zima. Opus: A coordination language for multidisciplinary applications. Scientific Programming (to appear), 1997.

....a programming language de nition its emphasis is on the semantics of the model. However, since it was obviously necessary to adopt some programming language notation, in particular for examples, we decided to use Fortran 95 syntax, with ad hoc extensions mainly motivated by HPF [31] and Opus [15]. 2 Processor in Memory For more than a decade, research experiments have been conducted with semiconductor devices that merged both logic and static RAM cell blocks on the same chips. Even earlier, simple processors and small blocks of SRAM could be found on simple control processors for ....

....macroserver 13 4.5 Macroserver Management The model currently provides two commands for the managementofmacroservers. We discuss migration and destruction below; additional commands maybeintroduced to deal with persistency, allowing the saving and retrieval of macroservers in long term storage [15]. 4.5.1 Migration An existing macroserver may be migrated in the virtual memory space: ####### (mv, cstr, sv) where mv is a macroserver variable referring to a macroserver S, cstr is a region constraint with at least one solution (Section 4.4.1) and sv is a status variable. The execution of ....

[Article contains additional citation context not shown here]

B. Chapman, M. Haines, P. Mehrotra, J. Van Rosendale, and H. Zima. Opus: A Coordination Language for Multidisciplinary Applications. Scientic Programming, 1998.

....Such applications are often easier to write in a task parallel language. Since both task and data parallelism thus have their strengths and weaknesses, it is attractive to integrate both forms in one model. Recently, several languages have been proposed that try to achieve such integration [4, 6, 8, 12, 16, 17, 19]. Systems like Fx, Opus, and Fortran M add taskparallel constructs to HPF (or other data parallel Fortran dialects) with the goal of making HPF more general purpose. Other research projects, such as Orca and Braid, are working in the opposite direction, and add data parallel constructs to a ....

....Fx is not sufficient for dynamic programming problems in which traditional concurrent programming is used, or for coordinating the execution of multiple data parallel applications. In both cases, Fx lacks general communication and synchronization capabilities. Opus Programming model Opus [4] adds task parallel programming constructs to HPF for the purpose of coordinating concurrent execution of several data parallel components. Therefore, a task in Opus is defined as the execution instance of a data parallel program. In this regard, Opus tasks are very coarse grained as compared with ....

[Article contains additional citation context not shown here]

Barbara Chapman, Matthew Haines, Piyush Mehrotra, John Van Rosendale, and Hans Zima. Opus: A coordination language for multidisciplinary applications. Scientific Programming, 6(2), April 1997.

....SMP clusters [13] In effect this model exposes the second level of parallelism to the programmer, with the advantage that it is more flexible than HPF, but the disadvantage of increased code complexity. Other approaches to coordinating multiple data parallel tasks are based on new languages, e.g. [14], or compiler analysis guided by directives to extract implicit task parallelism, e.g. 15] In contrast to our approach, these techniques require sophisticated compiler technology and or the adoption of new languages. 6 Conclusions We have presented an implementation of KeLP HPF, in which the ....

Chapman, B., Haines, M., Mehrotra, P., Van Rosendale, J., Zima, H.: Opus: A Coordination Language for Multidisciplinary Applications, Sci. Prog., 1997.

....a programming language definition its emphasis is on the semantics of the model. However, since it was obviously necessary to adopt some programming language notation, in particular for examples, we decided to use Fortran 95 syntax, with ad hoc extensions mainly motivated by HPF [31] and Opus [15]. 2 Processor in Memory For more than a decade, research experiments have been conducted with semiconductor devices that merged both logic and static RAM cell blocks on the same chips. Even earlier, simple processors and small blocks of SRAM could be found on simple control processors for ....

....13 4.5 Macroserver Management The model currently provides two commands for the management of macroservers. We discuss migration and destruction below; additional commands may be introduced to deal with persistency, allowing the saving and retrieval of macroservers in long term storage [15]. 4.5.1 Migration An existing macroserver may be migrated in the virtual memory space: MIGRATE (mv, cstr, sv) where mv is a macroserver variable referring to a macroserver S, cstr is a region constraint with at least one solution (Section 4.4.1) and sv is a status variable. The execution of ....

[Article contains additional citation context not shown here]

B. Chapman, M. Haines, P. Mehrotra, J. Van Rosendale, and H. Zima. Opus: A Coordination Language for Multidisciplinary Applications. Scientific Programming, 1998.

....that supports execution of multidisciplinary applicatins. In doing so, it must address multiple issues of coming task and data parallelism. There are a number of other projects that are looking to provide useful models for integrating task and data parallelsim, including Fx [7] Orca [8] Opus [2,3,9]. Fx provides Fortran language extensions that allow data parallel sections of code to communicate through channels (similar to MARS) Each of these data parallel executions is called a task, and communication between tasks (via channels) occurs only at task entry and exit points, resembling the ....

B. Chapman, M. Haines. P. Mehrotra, J.V. Rosendale, H. Zima, Opus: A coordination language for multidisciplinary applicatins. To appear in Scientic Programming, 1996. 8

....that supports execution of multidisciplinary applicatins. In doing so, it must address multiple issues of coming task and data parallelism. There are a number of other projects that are looking to provide useful models for integrating task and data parallelsim, including Fx [7] Orca [8] Opus [2,3,9]. Fx provides Fortran language extensions that allow data parallel sections of code to communicate through channels (similar to MARS) Each of these data parallel executions is called a task, and communication between tasks (via channels) occurs only at task entry and exit points, resembling the ....

B. Chapman, M. Haines, P. Mehrotra, H. Zima and J. Van Rosendale, Opus: A Coordination Language for Multidisciplinary Applications, ICASE Technical Report 97-30, June 1997.

....of a complex body of code and a concise formulation of the associated coordination and control mechanisms. Below we describe the features of Opus and then show how to use these features for a simple application code. A full description of the language features and their usage can be found in [8]. 4.1 Features of Opus Opus introduces a small set of features for defining and using SDA objects and accessing SDA data. It provides language constructs to define SDA types, declare SDA variables, create, initialize, terminate, and save SDA objects, as well as activate SDA methods both ....

B. Chapman, M. Haines, P. Mehrotra, J. Van Rosendale, and H. Zima. Opus: A Coordination Language for Multidisciplinary Applications. To be published in Scientific Programming, 1997 (Also available as ICASE Report No. 97-30 ICASE, NASA Langley Research Center, Hampton VA.)

....in nature, requiring several independent programs working together to generate a solution. Such applications do not fit well into the data parallel paradigm, since there are no facilities for controlling the interaction between multiple, independent program units. The Opus coordination language [6, 7] was recently designed to fill this gap. Opus defines language extensions to HPF that support the coordination of data parallel tasks. In doing so, Opus introduces coarse grained task parallelism on top of an existing data parallel model, where each task is represented by an independent program ....

....explicit messagepassing constructs throughout the code; ffl task level parallelism between modules and data parallelism within each module; and ffl flexible and general synchronization mechanisms that allow the programmer maximal freedom in exploiting task level parallelism. In a previous paper [7], we presented the syntax and semantics of the Opus language. In this paper, we describe the design of a system that efficiently implements the Opus constructs. We provide an overview of the whole system, ranging from the low level virtual machine up to the high level compiler component. In ....

B. Chapman, M. Haines, P. Mehrotra, J. Van Rosendale, and H. Zima. OPUS: A Coordination Language for Multidisciplinary Applications. Scientific Programming, 6/9:345--362, Winter 1997.

No context found.

B. Chapman, H. Zima, M. Haines, P. Mehrotra, and J. V. Rosendale. Opus: A coordination language for multidisciplinary applications. Scientific Programming, 6(4):345--362, 1997.

No context found.

B. Chapman, M. Haines, P. Mehrotra, J. Van Rosendale, and H. Zima. OPUS: A Coordination Language for Multidisciplinary Applications. Scientific Programming, 6/9:345-362, Winter 1997.

No context found.

B. Chapman, M. Haines, P. Mehrotra, J. Van Rosendale, and H. Zima. OPUS: A Coordination Language for Multidisciplinary Applications. Scientific Programming, 6/9:345--362, Winter 1997.

First 50 documents

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