J. V. W. Reynders, P. J. Hinker, J. C. Cummings, S. R. Atlas, S. Banerjee, W. F. Humphrey, S. R. Karmesin, K. Keahey, M. Srikant, and M. D. Tholburn. POOMA: A Framework for Scientific Simulations of Paralllel Architectures. In G. V. Wilson and P. Lu, editors, Parallel Programming in C++, pages 547--588. MIT Press, 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....manageable pieces, and creates reusable modules [6] Developers of large scientific software systems have come to depend more and more on layering, since such systems tend to represent a broad spectrum of specialized programming domains. For example, the POOMA framework consists of five layers [16], with the higher layers representing abstractions in the problem domain, such as solvers and complete simulations, and the lower layers representing abstractions in the implementation, such as communication, data distribution, and sequential kernels. Many other systems are similarly layered [4] ....

J. V. W. Reynders, P. J. Hinker, J. C. Cummings, S. R. Atlas, S. Banerjee, W. F. Humphrey, S. R. Karmesin, K. Keahey, M. Srikant, and M. D. Tholburn. POOMA: A framework for scientific simulation on parallel architectures.

....that anticipates and then avoids the overhead of object based programming. Thereby, however, they frequently sacrifice coding styles considerably. Especially in C there now exist high performance libraries, e.g. BLITZ [33] the Parallel Object Oriented Methods and Applications project (POOMA) [24], and the Matrix Template Library (MTL) 28] that are able to compete with the e#ciency of Fortran code, but pay for the performance gains with cryptic code at the user s level. For example, a simple operator expression in these libraries is written as a function call with additional parameters ....

J. V. Reynders, P. J. Hinker, J. C. Cummings, S. R. Atlas, S. Banerjee, W. F. Humphrey, S. R. Karmesin, K. Keahey, M. Srikant, and M. Tholburn. POOMA: A framework for scientific simulations on parallel architectures. In G. V. Wilson and P. Lu, editors, Parallel Programming using C++, pages 553--594. MIT Press, 1996.

....arrays. Since arrays are so frequently used in scientific computing, software support for arrays is very high in programming languages for scientific computing, for instance in Fortran 90 and subsequent versions. Also in C , several array libraries exist, for instance A P , Pooma, Blitz [7, 10, 14]. These libraries allows the programmer, in the same way as Fortran 90 does, to refer to whole (sub)arrays and not necessarily to individual components. The first advantage with this is that the notation becomes more compact, since one no longer has to loop explicitly over individual indices. ....

J. Reynders et al. Pooma: A framework for scientific simulations on parallel architectures. In G. Wilson and P. Lu, editors, Parallel Programming using C++, pages 553--594. MIT Press, 1996.

....that are based on equation (2) in our motivating example, for instance scalar fields, vector fields, and partial derivatives with respect to space coordinates. In 3 software, common names for the key abstraction on this level are Grid Function (see, e.g. 2, 11] or Field (see. e.g. [3, 9]) Coordinate free abstractions On the coordinate free abstraction level we find abstractions that are based on equation (1) Here, we also take into account the dependence of the coordinate system. In current software projects, apart from Sophus [5, 6] this level seems to be less developed. ....

J. Reynders et al. Pooma: A framework for scientific simulations on parallel architectures. In G. Wilson and P. Lu, editors, Parallel Programming using C++, pages 553--594. MIT Press, 1996.

....layer suitable for parallel computing to facilitate extending parallel computing over wide area networks. In general, this is achieved by implementing the component framework layer based on parallel programming languages or mechanisms. Examples include LSA [3] that is based on HPC , and in POOMA [10] the key mechanism to ensure reusability is the field class in which data parallel operations are encapsulated. The main reasons for using parallel languages or mechanisms are: a) Current distributed object systems do not support these parallel programming languages and the underlying parallel ....

.... as improving the efficiency of current distributed object s communication [6, 2] extending the real time and QoS features [11, 14] and supporting the mobile environment in a distributed object framework [13] Some frameworks support the composition of parallel component, such as LSA [3] POOMA [10], PARDIS [8] etc. DoHPC differs Multi threaded ORB in two aspects. Firstly, DoHPC provided partial concurrency control through inter method analysis, which makes the achieving of parallelism in a distributed object more transparent to its clients. Secondly, the implementation of multi threaded ....

J. Reynders, P. J. Hinker et.al, POOMA: A Framework for Scientific Simulation on Parallel Architectures, available in: http://www.acl.lanl.gov/pooma/documentation.

....in encapsulating parallelism so that scientists and engineers developing programs that require high processor bandwidths need not worry about the details of parallelism and optimization in their code. For example, frameworks like the Eiffel Parallel Execution Environment (EPEE) 27] and POOMA [43] have enabled developers to write parallel computing code with the ease of writing code for a single processor machine. All the internal optimization and distribution are handled internally by the framework. This approach removes the need for developing a compiler that targets a particular ....

J. et. al. Reynders. Pooma: A framework for scientific simulations on parallel architectures. In G.V. Wilson and P. Lu, editors, Parallel Programming Using C++, pages 553--594. MIT Press, 1996.

....on implementability and usability of the ET technique. 2 Problem domain, execution models Objects with array like semantics are fundamental to scientific computing, and numerous sophisticated parallel arrayclass libraries have been implemented and are in heavy use, e.g. A P [3] and POOMA [4], often underneath higherlevel C class libraries such as OVERTURE [1] As such the performance of the latter libraries and the applications that use them is directly related to that of the underlying array class libraries. The standard technique for implementing array class libraries in C is ....

J.V.W. Reynders et al. POOMA: A Framework for Scientific Simulations on Parallel Architectures, volume Parallel Programming using C++ by Gregory V. Wilson and Paul Lu, chapter 16, pages 553--594. MIT Press, 1996.

....FORTRAN 77 performance. INTRODUCTION The use of object oriented C frameworks has significantly simplified the development of numerous complex parallel scientific applications at Los Alamos National Laboratory (LANL) and elsewhere; examples at LANL include OVERTURE (Brown et al. 1997) and POOMA (Reynders, J.V.W. et al. 1996) In spite of considerable use of, and commitment to, these frameworks, concerns about performance are nonetheless a significant issue; simply put, performance very close to that of carefully hand crafted C or FORTRAN 77 with message passing must be realized before the acceptance and use of such ....

Reynders, J.V.W. et al. 1996. POOMA: A Framework for Scientific Simulations on Parallel Architectures.

....on implementability and usability of the ET technique. 2 Problem Domain and Execution Models Objects with array like semantics are fundamental to scientific computing, and numerous sophisticated array class libraries have been implemented and are in heavy use, e.g. A P [5] and POOMA [6], often underneath higher level C class libraries such as OVERTURE [1] which add support for complex geometry, adaptive mesh refinement, moving grids, and other features to meet more sophisticated applications requirements. As such the performance of the latter libraries and the applications ....

J.V.W. Reynders et al. POOMA: A Framework for Scientific Simulations on Parallel Architectures, volume Parallel Programming using C++ by Gregory V. Wilson and Paul Lu, chapter 16, pages 553--594. MIT Press, 1996.

....on our profiling and tracing tools. The ACTS toolkit is a group of parallel class libraries and scientific software frameworks. It provides a rich set of data parallel abstractions, exploiting advanced features of C . Most significantly, one of its central components the POOMA framework [17] makes extensive use of templates [18] to hide the mechanics of message passing behind data parallel arrays. Our first two requirements are thus . Abstraction. Our tools must handle data parallel abstractions, providing access to individual instantiation as well as aggregate behavior. ....

J.V.W. Reynders et. al., "Pooma: A Framework for Scientific Simulation on Parallel Architectures," in G.V. Wilson and P. Lu (Eds.), Parallel Programming using C++, pp. 553-594, MIT Press, 1996.

....to the master. Fig. 4. Callstack on node 0 of a POOMA 2D Diffusion equation simulation Figure 4 shows the callstack view of a two dimensional diffusion equation which tracks the progression of the diffusion of a heat source on a mesh with respect to time. It was implemented using the POOMA [10] object oriented scientific computing framework. 7 Conclusions The merging of the TAU and DAQV II systems described above has been implemented for performance callstack sampling. Callstack analysis and visualization tools have also been constructed. The TAU portable profiling library captures ....

Reynders, J. et. al.: Pooma: A Framework for Scientific Simulation on Parallel Architectures. In: Wilson, G., Lu, P. (Eds.): Parallel Programming using C++. M.I.T. Press (1996) 553--594

....support general concepts of scientific simulation on parallel computers and facilitate rapid application development. The Advanced Computing Laboratory (ACL) at Los Alamos is providing such support for the ASCI mission through its Parallel Object Oriented Methods and Applications (POOMA) project[1]. The POOMA team has developed a C class library that contains physics based data abstractions and parallel numerical algorithms designed to accelerate development of scientific simulations on parallel architectures. The primary goal of the POOMA project is to use modern, object oriented ....

J. V. W. Reynders et al., POOMA: A Framework for Scientific Simulations on Parallel Architectures, in Parallel Programming using C++, MIT Press, Cambridge, MA, 1996.

No context found.

J. V. W. Reynders, P. J. Hinker, J. C. Cummings, S. R. Atlas, S. Banerjee, W. F. Humphrey, S. R. Karmesin, K. Keahey, M. Srikant, and M. D. Tholburn. POOMA: A Framework for Scientific Simulations of Paralllel Architectures. In G. V. Wilson and P. Lu, editors, Parallel Programming in C++, pages 547--588. MIT Press, 1996.

No context found.

J. Reynders, et al. POOMA: A framework for scientific simulations of parallel architectures, Parallel Programming in C++, G. Wilson and P. Lu (editors), pp. 547-588, MIT Press, 1996.

No context found.

John V. Reynders, Paul J. Hinker, Julian C. Cummings, Susan R. Atlas, Subhankar Banerjee, William F. Humphrey, Steve R. Karmesin, Kataryzna Keahey, M. Srikant, and MaryDell Tholburn. POOMA: A framework for scientific simulations on parallel architectures. In Gregory Wilson and Paul Lu, editors, Parallel Programming using C++, pages 553--594. MIT Press, 1996.

No context found.

John V. W. Reynders, Paul J. Hinker, Julian C. Cummings, Susan R. Atlas, Subhankar Banerjee, William F. Humphrey, Steve R. Karmesin, Katarzyna Keahey, M. Srikant, and Mary Dell Tholburn. POOMA: A framework for scientific simulation on parallel architectures.

No context found.

J. Reynders, et al. POOMA: A framework for scientific simulations of parallel architectures, Parallel Programming in C++, G. Wilson and P. Lu (editors), pp. 547-588, MIT Press, 1996.

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