D. Kimelman, P. Mittal, E. Schonberg, P. F. Sweeney, K.-Y. Wang, and D. Zernik. Visualizing the execution of High Performance Fortran (HPF) programs. In IEEE, editor, IPPS '95: 9th International parallel processing symposium --- April 25--28, 1995.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....performance information that can be exploited during program parallelization. Moreover, the SUIF Explorer provides performance analysis for mostly regular applications whereas SCALA also covers irregular applications. An approach for visualizing the performance for HPF programs is described in [31]. Various insights about the interplay between data mapping and communication for HPF programs are offered by this system. In [1] the performance of Fortran D programs is analyzed at the source level which is based on an integration with the Fortran D compiler [21] and the Pablo performance ....

D. Kimelman, P. Mittal, E. Schonberg, P. F. Sweeney, K.-Y. Wang, and D. Zernik. Visualizing the execution of High Performance Fortran (HPF) programs. In IEEE, editor, IPPS '95: 9th International parallel processing symposium --- April 25--28, 1995.

....mainly concentrates on runtime behaviour of data structures. Well known representatives are VIPS [22] for dynamic data structures in Ada and VISTA [12] which shows contents of program variables during execution. Recent work also focusses on HPF and addresses both source based [23] and trace based [24] visualization. Concerning this classification GDDT belongs to the latter class. Unlike most existing systems GDDT focusses on compile time visualization because of productivity reasons. Furthermore all systems are associated with one programming language, while GDDT may be used for the whole ....

....the latter class. Unlike most existing systems GDDT focusses on compile time visualization because of productivity reasons. Furthermore all systems are associated with one programming language, while GDDT may be used for the whole family of HPF like languages due to its modularity. Kimelman et al. [24] report on a set of high level views for visualization of the execution of HPF programs. Mapping relationships between data arrays and processors are displayed similar to GDDT but just up to two dimensions. DDV [23] shows processor utilization in terms of data references. Visualization is based on ....

D. Kimelman, P. Mittal, E. Schonberg, P. Sweeny, K. Wang and D. Zernik, "Visualizing the Execution of High Performance Fortran Programs", IBM Thomas J. Watson Research Center Technical Report, 1994.

....distribution visualization is combined with dynamic data visualization to understand the effects of decomposition on algorithm operation. Similarly, Kimelman et al. show how a variety of runtime information can be correlated to data distribution to better visualize the execution of HPF programs [17]. And in the IVD tool [15] a data distribution specification provided by the user is used to reconstruct a distributed array that has been saved in partitioned form. DAQV II could easily provide distribution and runtime information to these tools from a running application. In this way, DAQV II is ....

D. Kimelman, P. Mittal, E. Schonberg, P. Sweeney, K. Wang, D. Zernik, Visualizing the Execution of High Performance Fortran (HPF) Programs, Proc. 9th Intl. Parallel Processing Symposium (IPPS), IEEE, 1995, pp. 750-757.

....GDDT is oriented to providing more compile time information on the shapes of distributed objects, and estimations of communication costs as data is manipulated or reshaped. Finally, research at the IBM TJ Watson Research Lab produced a prototype tool for visualizing data in running HPF programs [15]. This was strongly tied to the structure of the IBM Research HPF Compiler, and was not represented to be portable to other systems. The current situation is that DAQV can now be a reference for vendors to create proprietary optimized tools for each of their HPF language processing systems. The ....

Doug Kimelman, Pradeep Mittal, Edith Schonberg, Peter F. Sweeney, Ko-Yang Wang, and Dror Zernik. Visualizing the Execution of High Performance Fortran (HPF) Programs. URL http://www.almaden.ibm.com/watson/pv/ipps95abs.html.

....draws a representation of an object hierarchy and then animates the view with execution information such as recordings of object activations and method invocations. Performance views of parallel data structures have been more limited, but recent work in the area of performance visualization [30, 25] has focused on drawing representations of vectors and matrices, annotating these representations with data layouts, and animating the views with traces of computation and communication events. Our work differs from these efforts in that we have not attempted to visualize and animate the actual ....

Doug Kimelman, Pradeep Mittal, Edith Schonberg, Peter F. Sweeney, Ko-Yang Wang, and Dror Zernik. Visualizing the execution of high performance fortran (HPF) programs. Technical report, IBM Thomas J. Watson Research Center, October 1994.

....The run time library is portable across different basic communication systems. Currently it supports both IBM MPL and MPI libraries. The runtime system also provides detailed performance statistics, trace information for debugging by hand, and trace generation for program visualization tools [12]. 3 Optimizations The SPMDizer of pHPF performs several optimizations to reduce both communication costs and the overhead of guards introduced by computation partitioning. Some of these optimizations are well known and have been discussed in the literature [11, 23, 17, 19, 3] while many others ....

D. Kimelman, P. Mittal, E. Schonberg, P. Sweeney, K. Wang, and D. Zernik. Visualizing the execution of high performance fortran (hpf) programs. In Proc. of 1995 Intl' Conf. on Parallel Processing Symposium, April 1995.

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