S. Benkner. HPF+: High Performance Fortran for advanced industrial applications. Lecture Notes in Computer Science, 1401, 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....the efficient selected atomic technique. The approach is effective because the data access pattern is constant for a long time, but a compiler has no way of knowing this. We are currently proposing extensions to OpenMP for this case [8] similar to the SCHEDULE REUSE extensions proposed for HPF [9]. Particle reordering (used above) has a significant effect on the performance of the OpenMP code. For D =3on the Sun the improvement was between 15 and 20 , whereas on the Compaq the figures were between 45 and 65 . In contrast to MPI where the increased performance is a purely serial effect ....

S. Benkner, HPF+ High Performance Fortran for Advanced Industrial Applications, Technical Report TR 98-04, Institute for Software Technology and Parallel Systems, University of Vienna, May 1998.

....transformations and optimizations applied by a compiler. In this paper we introduce P 3 T , the successor tool of P 3 T [22, 15, 16] which models programs, code transformations, and parallel and distributed architectures. The input programs of P 3 T are written in High Performance Fortran [27, 1] which represents the de facto standard of high level data parallel programming. Moreover, P 3 T analyzes Fortran90 message passing programs generated by the underlying compiler (VFC [2] which can be executed on parallel and distributed machines such as network of workstations. P 3 T models ....

S. Benkner. HPF+: High Performance Fortran for advanced industrial applications. Lecture Notes in Computer Science, 1401, 1998.

....phase is specified by indirection array X and distribution of F. If X is not written between consecutive invocations of the inspector phase then the communication schedule can be reused which is a crucial aspect to reduce the inspector overhead. There exist language extensions such as HPF [6] which introduced a REUSE clause for this purpose. For instance, the work distribution and inspector phases of a DO INDEPENDENT loop with a REUSE clause are guarded by means of conditional statements to enforce that they are executed only if the reuse condition is true or if the loop is executed ....

S. Benkner. HPF+: High Performance Fortran for advanced industrial applications. Lecture Notes in Computer Science, 1401, 1998.

....avoidance of redundant inspector executions is a key issue when parallelizing irregular codes. However, existing compile time and runtime analysis techniques [76, 63] are not powerful enough to detect all instances of redundant inspectors. Thus, language support is required for this problem. HPF [3], an improved variant of HPF for advanced industrial applications which has been developed in the Research Project HPF , addresses these issue by providing language features to express the redundancy of schedule computations. Moreover, HPF offers mechanisms to influence the mapping of ....

....2 ,x 2 y,x 2 y 2 Out[1] 0.0000105611 9.18523 10 6 x 2.7142 10 7 x 2 5.36312 10 9 y 2.04169 10 7 x y 7.53052 10 9 x 2 y 1.10728 10 9 y 2 3.07453 10 10 x y 2 1. 94622 10 11 x 2 y 2 In[2] RecursionLimit = 1000 Out[2] 1000 In[3]: WriteList[output 10 to 100.dat,vgl[xyz] In[4] Plot3D[f[x, y] x, 1, 16 , y, 10, 100 , PlotRange 0, 0.00025 , BoxRatios 6, 6, 5 , PlotPoints 40, ViewPoint 2, 5, 2 ] 5 10 15 20 40 60 80 100 0 0.00005 0.0001 0.00015 0.0002 0.00025 5 10 15 20 40 60 ....

S. Benkner. HPF+: High Performance Fortran for advanced industrial applications. Lecture Notes in Computer Science, 1401, 1998.

....transformations and optimizations applied by a compiler. In this paper we introduce P 3 T , the successor tool of P 3 T [20, 13, 14] which models programs, code transformations, and parallel and distributed architectures. The input programs of P 3 T are written in High Performance Fortran [25, 1] which represents the de facto standard of high level data parallel programming. Moreover, P 3 T analyzes Fortran90 message passing programs generated by the underlying compiler (VFC [2] which can be executed on parallel and distributed machines such as network of workstations. P 3 T models ....

S. Benkner. HPF+: High Performance Fortran for advanced industrial applications. Lecture Notes in Computer Science, 1401, 1998.

....is indispensable in modern software engineering. Moreover, a purely data parallel approach is in some cases not appropriate and needs to be accompanied by task parallel techniques. The Approved Extensions of HPF cope with these problems. For our implementation we employed an HPF dialect HPF [1, 3, 4] that, among others, provides most of the features defined in the Approved Extensions. The research compiler Vienna Fortran Compiler (VFC) 2] was used for compiling our HPF codes. In the remainder of this paper we give an overview of the Aurora Financial Management System (Section 2) ....

....of the model building software. 3 Implementing Selected Application Modules in HPF In this section we describe a parallel implementation of the modules described above. Since explicit parallel programming on top of MPI [20] or threads [15] is cumbersome and error prone, we made use of HPF [1, 3, 4], a high level parallel programming language. HPF , an improved version of High Performance Fortran (HPF) 12] has been developed in close cooperation with application designers from industry in the context of an ESPRIT project. HPF provides compiler directives for distributing data on the ....

S. Benkner. HPF+: High Performance Fortran for Advanced Industrial Applications. In Proceedings HPCN Europe 1998, Lecture Notes in Computer Science, No. 1401. Springer, April 1998.

....by a compiler. P 3 T , the successor tool of P 3 T [15, 11, 12] is a performance estimator for distributed and parallel programs which models programs, code transformations, and parallel and distributed architectures. The input programs of P 3 T are written in High Performance Fortran [17, 1] which represents the de facto standard of high level data parallel programming. Moreover, P 3 T analyzes Fortran90 message passing programs generated by the underlying compiler (VFC [2] and which can execute on parallel and distributed machines such as network of workstations. P 3 T ....

S. Benkner. HPF+: High Performance Fortran for advanced industrial applications. Lecture Notes in Computer Science, 1401, 1998.

....the SPiDER needs, and its functionality will be extended to be able to visualize data distribution on I O devices. In the next step, we will work on improvements to debugger s performance by adding extensions to the Base Debugging Subsystem and we will support a selected subset of optimizations [1, 3]. It will be developed in coordination with the compiler developers that deal with optimizations of irregular codes and address parallel I O. ....

S. Benkner. HPF+. High Performance Fortran for Advanced Industrial Applications. In Proc. Int. Conf. High Performance Computing and Networking, volume 1401 of Lecture Notes in Computer Science, pages 797--808, Amsterdam, The Netherlands, April 1998. Springer.

....take advantage of the communication library which is best suited to the particular computer hardware. On the other hand, it is recognized that the current HPF standard does not provide sufficient support for dealing with unstructured mesh applications. Optimized versions of HPF have been pursued [1], which helped to identify which extensions to the language are desirable. Here, we report on the program development strategy, the optimization steps, and the obtained parallel performance for two MHD application codes: the Versatile Advection Code (VAC) 16, 18] and the HEating by Resonant ....

....suited for vector and data parallel simulations of magnetized plasmas. The parallelism expressed in High Performance Fortran works well for present day distributed memory supercomputers. This optimistic viewpoint may not be applicable for algorithms exploiting highly irregular data structures [1, 13]. We provided ample implementation details, which should be useful for HPF code developers. During our efforts to parallelize our F90 codes, compiler bugs were identified and resolved in successive releases of the still maturing HPF compilers. Our choice of using only a subset of the HPF language ....

S. Benkner, HPF+ High Performance Fortran for Advanced Industrial Applications, in: P. Sloot, M. Bubak, and B. Hertzberger, eds., Lecture Notes in Computer Science, Vol. 1401, Proceedings HPCN Europe 1998 (SpringerVerlag, Berlin, 1998) 797-808.

....will be adapted to the DeHiFo needs, and its functionality will be extended to be able to visualize data distribution on I O devices. In the next step, we will work on improvements to debugger s performance by adding extensions to the BDS and we will support a selected subset of optimizations [1, 3]. It will be developed in coordination with the compiler developers that deal with optimizations of irregular codes and address parallel I O. ....

S. Benkner. HPF+. High Performance Fortran for Advanced Industrial Applications. In Proc. Int. Conf. High Performance Computing and Networking, volume 1401 of Lecture Notes in Computer Science, pages 797--808, Amsterdam, The Netherlands, April 1998. Springer.

....GDDT will be adapted to the DeHiFo s needs, and its functionality will be extended to be able to visualize data distribution on I O devices. In the next step, we will work on improvement of the debugger s performance by adding extensions to the BDS and support a selected subset of optimizations [2]. Acknowledgements The development of DeHiFo is a joint effort of research teams from several universities headed by M. Bubak (AGH Krak ow) R. Koppler (University of Linz) and R. Wismuller (TU Munich) We are grateful for many usefull suggestions and comments of J. Fritscher (Vienna) S. ....

S. Benkner. HPF+. High Performance Fortran for Advanced Industrial Applications. In Proc. Int. Conf. High Performance Computing and Networking, volume 1401 of Lecture Notes in Computer Science, pages 797--808, Amsterdam, The Netherlands, April 1998. Springer.

....functionality, and introduces new features to provide the compiler with essential information on the applicability of crucial optimizations. In the following we give a brief overview of the HPF language features supported by VFC. For a more detailed description of HPF the reader is referred to [3]. 2.2.1 Processor Arrays Arbitrary multi dimensional processor arrays may be defined. The program may be parallelized for an arbitrary number of processors by using the intrinsic function NUMBER OF PROCESSORS( in processor declarations. An explicit relationship between different processor arrays ....

S. Benkner. HPF+: High Performance Fortran for Advanced Industrial Applications. Proceedings International Conference on High Performance Computing and Networking (HPCN'98), Amsterdam, April 1998, Springer Verlag.

No context found.

S. Benkner. HPF+: High Performance Fortran for Advanced Industrial Applications. Proceedings HPCN'98, Amsterdam, April 1998, Springer Verlag.

No context found.

S. Benkner. HPF+: High Performance Fortran for advanced industrial applications. Lecture Notes in Computer Science, 1401, 1998.

No context found.

S. Benkner. HPF+: High Performance Fortran for advanced industrial applications. Lecture Notes in Computer Science, 1401, 1998.

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