S. Atlas, S. Banerjee, et al, "POOMA: A high performance distributed simulation environment for scientific applications", Proc. SuperComputing'95, Nov. 1995  Home/Search   Document Not in Database   Summary   Related Articles   Check

....is sometimes because it doesn t support the operations (operator overloading, for example) and sometimes because it does the work for you (automatic garbage collection, for example) In recent years numerous variations on the theme of C for parallel computing have appeared. See, for example [8, 17, 23, 28, 31, 55]. 21 2.4.3 C# C# [20] which is an important part of the new Microsoft .NET platform, is a modern, object oriented and type safe programming language based on C and C . Header files, Interface Definition Language, and complicated interfaces are not needed in C#. C# is supposed to o#er an ....

Susan Atlas, Subhankar Banerjee, Julian C. Cummings, Paul J. Hinker, M. Srikant, John V. W. Reynders, and Mary Dell Tholburn. POOMA: A high performance distributed simulation environment for scientific applications. In Supercomputing `95, 1995.

....earlier is critical for e#ciency. In addition, due to array indexing, some data packing and unpacking are necessary. In these procedures, moving a block of data, rather than moving one array element at a time, will increase speed. A number of existing software supports variable ghost cell layers [1, 13, 14]. 7 3 Test problem Although the original motivation for this work is on atmosphere and ocean models, we test the ghost cell expansion method on a simpler 2D static heat distribution problem, to clearly illustrate some performance issues. The 2D problem is governed by the Laplacian ....

S. Atlas, S. Banerjee, J.C. Cummings, P.J. Hinker, M. Srikant, J.V.W. Reynders, and M. Tholburn. "POOMA: A high performance distributed simulation environment for scientific applications." Proc. Supercomputing '95, Nov 1995. see also http://www.acl.lanl.gov/pooma/.

....towards both system software and programming environments for scientific computing. The industry is already moving in this direction and in the past years several projects took on migrating legacy software to modern techniques (e.g. object oriented wrappers, component based development, etc. [13, 14, 19, 27, 59]) It is still unclear to what degree performance losses due to the software layer (system and programming environment) are unacceptable. Our thesis is that the main gain in performance comes from the architectural and algorithmic scalability. Thus, the software layer may negatively affect ....

Susan Atlas, Subhankar Banerjee, Julian Cummings, Paul J. Hinker, M. Srikant, John V. W. Reynders, and Marydell Tholburn. Pooma: A high performance distributed simulation environment for scientific applications. In Proceedings of SuperComputing 95, 1995.

....in broadcasts and reductions, and migrate as needed. Each element of the array is identified by a unique array index, which may be variable length. Because elements can be individually scheduled and migrated, an object array is quite distinct from the array objects found in HPF, POOMA[13], P [14] Global Arrays[12] and elsewhere. In our construct, each element of the array is a relatively coarse grained 1 C object, with full support for remote method invocation. Our work is quite similar to pC [8] but adds migration and reductions. Unlike Concurrent Aggregates[9] Linda[7] ....

S. Atlas, S. Banerjee, J. C. Cummings, and others (presented by J. Reynders). "POOMA: A high performance distributed simulation environment for scientific applications," Supercomputing

....the actual Uintah Material Point Method (MPM) task graph concerned with advancing Newtonian material point motion on a single patch for a single timestep. The idea of the dataflow graph as an organizing structure for execution is well known. The SMARTS dataflow engine that underlies the POOMA [6] toolkit shares goals and philosophy with UPI. SISAL compilers used dataflow concepts at a much finer granularity to structure code generation and execution. Dataflow is a simple, natural and efficient way of exposing parallelism and managing computation, and is an intuitive way of reasoning ....

S. Atlas, S. Banerjee, J.C. Cummings, P.J. Hinker, M. Srikant, J.V.W. Reynders, and M. Tholburn. POOMA: A high-performance distributed simulation environment for scientific applications. In Supercomputing '95 Proceedings, December 1995.

.... Java is a descendant of C . C and C are used increasingly in scienti c programming; they are already used almost universally by implementers of parallel libraries and compilers. In recent years numerous variations on the theme of C for parallel computing have appeared. See, for example [7, 28, 11, 2, 12, 21]. 3 Java omits various features of C and C that are considered dicult notably, pointers. Poor compiler analysis has often been blamed on these features. The inference is that Java, like Fortran, may be a suitable source language for highly optimizing compilers (although direct evidence ....

....o set is in units of the buf array element or the base type of datatype not of any compound type. The Object o set presentation is reminiscent of the interface of the arrayCopy utility in the standard Java API. 12 class Life void main(String args) MPI.Init(args) int dims [ new int [2] ; Set dims , etc Cart p = new Cart(MPI.WORLD, dims, periods, false) int coords = new int [2] p.Get(dims, periods, coords) Compute local blockSizeX , blockBaseX , blockSizeY , blockBaseY . Create block , allowing for ghost cells. int sX = blockSizeX 2 ; int sY = ....

[Article contains additional citation context not shown here]

Susan Atlas, Subhankar Banerjee, Julian C. Cummings, Paul J. Hinker, M. Srikant, John V. W. Reynders, and Mary Dell Tholburn. POOMA: A high performance distributed simulation environment for scientic applications. In Supercomputing `95, 1995.

....some reasons to think that Java may be a good language for scientific and parallel programming. Java is a descendant of C . C and C are used increasingly in scientific programming. In recent years numerous variations on the theme of C for parallel computing have appeared. See, for example [26, 6, 9, 2, 10, 18]. Java omits various features of C and C that are considered difficult notably, pointers. Poor compiler analysis has often been blamed on these features. The inference is that Java, like Fortran, may be a suitable source language for highly optimizing compilers (although direct evidence ....

....buffer area. But the data address for the Java array could be moved during the lifetime of the request object. Our current Java binding omits non blocking communication, so we have not addressed this problem. 6 class Life void main(String args) MPI.Init(args) int dims [ new int [2] ; Set dims , etc Cart p = new Cart(MPI.WORLD, dims, periods, false) int coords = new int [2] p.Get(dims, periods, coords) Compute local blockSizeX , blockBaseX , blockSizeY , blockBaseY . Create block , allowing for ghost cells. int sX = blockSizeX 2 ; int sY = ....

[Article contains additional citation context not shown here]

Susan Atlas, Subhankar Banerjee, Julian C. Cummings, Paul J. Hinker, M. Srikant, John V. W. Reynders, and Mary Dell Tholburn. POOMA: A high performance distributed simulation environment for scientific applications. In Supercomputing `95, 1995.

....or element wise, and each element has access to other elements via a pointer to the collection itself. In addition to data parallel operations, IC also provides the ability to spawn new threads to execute statements. The Parallel Object Oriented Methods and Applications (POOMA) Framework [2] is a C class library which supports data parallel programming. Like my implementation, POOMA is a layered collection of C classes. In particular, the Global and Local Layers provide Global Data Types, which are data parallel classes for several data types including Fields (multidimensional ....

Susan Atlas, Subhankar Banerjee, Julian C. Cummings, Paul J. Hinker, M. Srikant, John V. W. Reynders, and Marydell Tholburn. POOMA: A high performance distributed simulation environment for scientific applications. In Supercomputing '95, December 1995. 15

....programming. Distributed arrays are supported as templated classes with support for various data parallel operations. Compiler supported approaches such as the PC [13] use syntactic extensions for supporting HPF style data distributions and data parallel operations. Frameworks such as POOMA[6] and Overture[10] use mechanisms provided in C itself (such as operator overloading and templates) and provide a rich library of operators useful in scientific engineering applications. 3.4.4 Other approaches Several additional languages are being used in parallel CSE applications. We will ....

Susan Atlas, Subhankar Banerjee, Julian C. Cummings, Paul J. Hinker, M. Srikant, John V. W. Reynders, and Marydell Tholburn. Pooma: A high performance distributed simulation environment for scientific applications. In Supercomputing '95, 1995.

....analysis and optimization to improve performance [8, 4] Alternatively, some specialization can be built into the library instantiation process, if the library is written in a language that uses some form of dynamic macro expansion. This is the approach used to optimize the POOMA library [2]. Unfortunately, these approaches suffer from the disadvantage that program compilation times, either for scripts or expansion of data structures, can be long. To avoid this, we have developed a new approach called telescoping languages, in which the libraries that provide component operations ....

S. Atlas, S. Banerjee, J. C. Cummings, P. J. Hinker, M. Srikant, J. V. W. Reynders, and M. Tholburn. POOMA: A high performance distributed simulation environment for scientific applications. In Proceedings of Supercomputing 95, San Diego, CA, Dec. 1995.

....parallel programs in terms of dataflow; our work differs in that it addresses patterns of both dataflow and computation. 162 Distributed objects. The mesh spectral archetype is based to some extent on the idea of distributed objects, as discussed for example in work on pC [12] and POOMA [4]. Our work on archetypes differs from this work in that we focus more on the pattern of computation and on identifying and exploiting patterns of computation and communication. Communication libraries. Many researchers have investigated and developed reusable general libraries of communication ....

S. Atlas, S. Banerjee, J. C. Cummings, P. J. Hinker, M. Srikant, J. V. W. Reynders, and M. Tholburn. POOMA: A high performance distributed simulation environment for scientific applications. h http://www.acl.lanl.gov/PoomaFramework/papers/SCPaper-95.html i, 1995.

....implement an algorithm or a library in a natural way (see, e.g. SparseLib and IML [8] Such an approach will (by design) hide computational costs from the user and degrade performance. One approach to providing performance and abastraction is through the use of lazy evaluation (see, e.g. [2]) but this approach can have other performance penalties as well as implementation difficulties. One of the most important concerns in obtaining high performance on modern workstations is proper exploitation of the memory hierarchy. That is, a high performance algorithm must be cognizant of the ....

S. Atlas et al., POOMA: A high performance distributed simulation environment for scientific applications, in Proc. Supercomputing '95, 1995.

....architectures, appears in [36] Examples of specific algorithms include one deep mergesort [20] and one deep quicksort [13, 35] Distributed objects. The mesh spectral archetype is based to some extent on the idea of distributed objects, as discussed for example in work on pC [5] and POOMA [1]. We differ from this work in that we focus more on the pattern of computation and on identifying and exploiting patterns of computation and communication. Communication libraries. Many researchers have investigated and developed reusable general libraries of communication routines; MPI [29] is a ....

S. Atlas, S. Banerjee, J. C. Cummings, P. J. Hinker, M. Srikant, J. V. W. Reynders, and M. Tholburn. POOMA: A high performance distributed simulation environment for scientific applications. h http:/ /www.acl.lanl.gov/PoomaFramework/papers/SCPaper-95.html i, 1995.

....relative to a particular software system. More specifically, binding between a component CDS and its implementation is the code that for a particular software system will bind (or translate) a CDS definition into format used by the component implementation. For example, a binding to a POOMA [1] system might translate a CDS definition of a multi dimensional grid into a multi dimensional POOMA field. Binding allows us to compromise between generality of the CDS descriptions and the efficiency needed by a particular implementation. For example, a grid fragment defined in CDS and exchanged ....

S. Atlas, S. Banerjee, J. Cummings, P. J. Hinker, M. Srikant, J. V. W. Reynders, and M. Tholburn. POOMA: A High Performance Distributed Simulation Environment for Scientific Applications. In Supercomputing '95 Proceedings, December 1995.

....scientific computing, primarily because they do not support efficient parallel communication channels between components. Abstractions suitable for high performance computing are needed. The existence of many successful high performance languages and libraries such as HPC [24] POOMA [4], ISIS [12] SAMRAI [29] and PETSc [5] testifies that such abstractions can enable the user to develop more efficient programs faster. Similarly, we need abstractions that capture high performance concepts in component architectures. For example, PARDIS [37] and PAWS [6] successfully show ....

S. Atlas, S. Banerjee, J. Cummings, P. J. Hinker, M. Srikant, J. V. W. Reynders, and M. Tholburn. POOMA: A highperformance distributed simulation environment for scientific applications. In Supercomputing '95 Proceedings, December 1995.

....number of platforms. Existing sparse matrix collections like the Harwell Boeing suite [8] are important tools for designers of numerical algorithms and partitioning algorithms. And existing software libraries that support sparse matrix computations are useful tools for application programmers [2, 7, 11]. However, to date the designers of parallel systems have not had access to a set of public domain sparse matrix kernels that are simple, realistic, and portable. Towards that end, we introduce the Spark98 kernels, a set of 10 SMVP kernels for shared memory and message passing systems. 1 We ....

.... ( v) 3] double ( w) 3] int firstrow, int numrows) f 3 int i; 4 int Anext, Alast, col; 5 double sum0, sum1, sum2; 6 7 for (i = firstrow; i (firstrow numrows) i ) f 8 Anext = Aindex[i] 9 Alast = Aindex[i 1] 10 11 sum0 = A[Anext] 0] 0] v[i] 0] A[Anext] 0] 1] v[i] 1] A[Anext] 0][2] v[i] 2] 12 sum1 = A[Anext] 1] 0] v[i] 0] A[Anext] 1] 1] v[i] 1] A[Anext] 1] 2] v[i] 2] 13 sum2 = A[Anext] 2] 0] v[i] 0] A[Anext] 2] 1] v[i] 1] A[Anext] 2] 2] v[i] 2] 14 15 Anext ; 16 while (Anext Alast) f 17 col = Acol[Anext] 18 sum0 = A[Anext] 0] 0] v[col] 0] ....

[Article contains additional citation context not shown here]

ATLAS, S., BANERJEE, S., CUMMINGS, J., HINKER, P., SRIKANT, M., REYNDERS, J., AND THOLBURN, M. POOMA: A high performance distributed simulation environment for scientific applications. In Proceedings of Supercomputing '95 (Washington, D.C., Nov. 1995).

.... The benefit of data abstraction in objectoriented languages on parallel software development has been demonstrated by various efforts [8, 16] Research efforts in providing suitable object oriented parallel languages libraries for certain classes of applications have also been abundant [3, 5, 2, 14, 20]. Our VGDS effort distinguishes itself from others in two aspects. First, instead of tackling one particular data structure, we propose an integrated framework incorporating a diverse set of data structure classes that are essential in a broad base of scientific, engineering, and commercial ....

....layered object oriented design framework for a diverse set of distributed data structures, where data distribution, data sharing, data coherence, and synchronization between data references are mediated by the runtime system. Our VGDS effort has similar goals and approaches to the POOMA package [2] and the Chaos library [17] POOMA supports a set of distributed data structures (fields, matrices, particles) for scientific simulations. To our knowledge, POOMA has not supported adaptive data structures as VGDS does. Chaos is a general purpose runtime library that supports pointer based ....

Susan Atlas, Subhankar Benerjee, Julian C. Cummings, Paul J. Hinker, M. Srikant, John V.W. Reynders, and Marydell Tholburn. Pooma: A high performance distributed simulation environment for scientific applications. In Supercomputing95, 1995.

....reasons to think that Java may be a good language for scientific and parallel programming. ffl Java is a descendant of C . C and C are used increasingly in scientific programming. In recent years numerous variations on the theme of C for parallel computing have appeared. See, for example [25, 6, 9, 2, 10, 17]. ffl Java omits various features of C and C that are considered difficult notably, pointers. Poor compiler analysis has often been blamed on these features. The inference is that Java, like Fortran, may be a suitable source language for highly optimizing compilers (although direct ....

....is in units of the buf array element or the base type of datatype not of any compound type. The Object offset presentation is reminiscent of the interface of the arrayCopy utility in the standard Java API. class Life void main(String args) MPI.Init(args) int dims [ new int [2] ; Set dims , etc Cart p = new Cart(MPI.WORLD, dims, periods, false) int coords = new int [2] p.Get(dims, periods, coords) Compute blockSizeX , blockBaseX , blockSizeY , blockBaseY . Create block , allowing for ghost cells. int sX = blockSizeX 2 ; int sY = blockSizeY 2 ; ....

[Article contains additional citation context not shown here]

Susan Atlas, Subhankar Banerjee, Julian C. Cummings, Paul J. Hinker, M. Srikant, John V. W. Reynders, and Mary Dell Tholburn. POOMA: A high performance distributed simulation environment for scientific applications. In Supercomputing `95, 1995.

....not yet a fully satisfactory solution. For a truly seamless integration, the sequence will map directly to constructs present in the programmer s package (such as for example distributed vector in HPC PSTL [3] We are currently working on formulating direct mappings for the HPC PSTL and POOMA [1] libraries. 2.3 General Design Components of PARDIS PARDIS is a distributed software system consisting of an IDL compiler, communication libraries, object repository databases and facilities responsible for locating and activating objects. The relationship between these components is depicted in ....

S. Atlas, S. Banerjee, J. Cummings, P. J. Hinker, M. Srikant, J. V. W. Reynders, and M. Tholburn. POOMA: A High Performance Distributed Simulation Environment for Scientific Applications. In Supercomputing '95 Proceedings, December 1995.

No context found.

S. Atlas, S. Banerjee, et al, "POOMA: A high performance distributed simulation environment for scientific applications", Proc. SuperComputing'95, Nov. 1995

No context found.

S. Atlas, S. Banerjee, J. Cummings, P.J. Hinker, M. Srikant, J.V.W. Reynders, and M. Tholburn, POOMA: A high performance distributed simulation environment for scientific applications, SuperComputing '95

No context found.

S. Atlas, S. Banerjee, J. C. Cummings, P. J. Hinker, M. Srikant, J. V. W. Reynders, and M. Tholburn. POOMA: A high-performance distributed simulation environment for scientific applications. In Supercomputing '95 Proceedings, December 1995.

No context found.

S. Atlas, S. Banerjee, J.C. Cummings, P. J. Hinker, M. Srikant, J. V. W. Reynders, and M. Tholburn, POOMA: A High Performance Distributed Simulation Environment for Scientific Applications, Supercomputing '95 Proceedings, December 1995.

No context found.

S. Atlas, S. Banerjee, J.C. Cummings, P. J. Hinker, M. Srikant, J. V. W. Reynders, and M. Tholburn, POOMA: A High Performance Distributed Simulation Environment for Scientific Applications, Supercomputing '95 Proceedings, December 1995.

No context found.

S. Atlas, S. Banerjee, J.C. Cummings, P. J. Hinker, M. Srikant, J. V. W. Reynders, and M. Tholburn, POOMA: A High Performance Distributed Simulation Environment for Scientific Applications, Supercomputing '95 Proceedings, December 1995.

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