Hiranandani, S., Kennedy, K., Tseng, C.-W., and Warren, S. The D editor: A new interactive parallel programming tool. In Proceedings of Supercomputing (1994).  Home/Search   Document Not in Database   Summary   Related Articles   Check

....for a programmer to decide whether an unexpected invariant is caused by a weakness in the proof system or a flaw in the program. In addition, the existing system processes only a first order language, though Bourdoncle explains how to extend the analysis [3:Section 5] Several environments [30, 6, 26, 44, 41] have been built for parallel programming languages to expose dependencies, thus allowing the programmer to tune programs to minimize these dependencies. In particular, the ParaScope [6, 30] and D editors [26] have many similarities to MrSpidey. Both MrSpidey and the editors provide information at ....

....explains how to extend the analysis [3:Section 5] Several environments [30, 6, 26, 44, 41] have been built for parallel programming languages to expose dependencies, thus allowing the programmer to tune programs to minimize these dependencies. In particular, the ParaScope [6, 30] and D editors [26] have many similarities to MrSpidey. Both MrSpidey and the editors provide information at varying levels of granularity; both retain source correlation through transformations; and both depict dependencies graphically. However, unlike MrSpidey, the editors process a language with extremely simple ....

Hiranandani, S., Kennedy, K., Tseng, C.-W., and Warren, S. The D editor: A new interactive parallel programming tool. In Proceedings of Supercomputing (1994).

....for a programmer to decide whether an unexpected invariant is caused by a weakness in the proof system or a flaw in the program. In addition, the existing system processes only a first order language, though Bourdoncle explains how to extend the analysis [3:Section 5] Several environments [30, 6, 26, 44, 41] have been built for parallel programming languages to expose dependencies, thus allowing the programmer to tune programs to minimize these dependencies. In particular, the ParaScope [6, 30] and D editors [26] have many similarities to MrSpidey. Both MrSpidey and the editors provide information at ....

....explains how to extend the analysis [3:Section 5] Several environments [30, 6, 26, 44, 41] have been built for parallel programming languages to expose dependencies, thus allowing the programmer to tune programs to minimize these dependencies. In particular, the ParaScope [6, 30] and D editors [26] have many similarities to MrSpidey. Both MrSpidey and the editors provide information at varying levels of granularity; both retain source correlation through transformations; and both depict dependencies graphically. However, unlike MrSpidey, the editors process a language with extremely simple ....

Hiranandani, S., Kennedy, K., Tseng, C.-W., and Warren, S. The D editor: A new interactive parallel programming tool. In Proceedings of Supercomputing (1994).

....for a programmer to decide whether an unexpected invariant is caused by a weakness in the proof system or a flaw in 11 the program. In addition, the existing system processes only a first order language, though Bourdoncle explains how to extend the analysis [2, Section 5] Several environments [17, 4, 14, 26, 24] have been built for parallel programming languages to expose dependencies, thus allowing the programmer to tune programs to minimize these dependencies. In particular, MrSpidey has many similarities to the ParaScope [17, 4] and D editors [14] Both MrSpidey and the editors provide information at ....

....[2, Section 5] Several environments [17, 4, 14, 26, 24] have been built for parallel programming languages to expose dependencies, thus allowing the programmer to tune programs to minimize these dependencies. In particular, MrSpidey has many similarities to the ParaScope [17, 4] and D editors [14]. Both MrSpidey and the editors provide information at varying levels of granularity; both retain source correlation through transformations; and both depict dependencies graphically. However, unlike MrSpidey, the editors process a language with extremely simple control and data flow facilities, ....

Hiranandani, S., Kennedy, K., Tseng, C.-W., and Warren, S. The D editor: A new interactive parallel programming tool. In Proceedings of Supercomputing (1994). 13

....a speedup of 8 on 11 processors for a 127 by 127 grid point North Atlantic simulation. This contrasts with the results using the SGI Power Fortran Analyzer (a loop based parallelization tool) which only achieved a speedup of 5 on 16 processors for the same problem. 6 Related Work The D Editor [14] is an interactive tool to help create efficient programs in the Fortran D and HPF languages. It interfaces to the Fortran D compiler to get feedback about the program being compiled, then presents this information to the user graphically. The D Editor relates information about communication and ....

Seema Hiranandani, Ken Kennedy, Chau-Wen Tseng, and Scott Warren. The D Editor: A New Interactive Parallel Programming Tool. In Proceedings of Supercomputing '94, 1994.

....a performance model, programmers are left to guess how the compiler will optimize their code. They must often write to idioms to enable optimizations in a particular compiler, severely increasing programming effort and degrading performance. The proliferation of parallelizing restructuring tools [6, 7, 11] attest to the former and Ngo has demonstrated the latter [15] This research was supported in part by DARPA Grant E30602 97 1 0152, NSF Grant CCR9710284 and the Intel Corporation. DO 100 i = 2 , n 1 DO 100 j = 2 , n 2 r=aa(j,i) d(j 1,i) d(j,i) 1.0 (dd(j,i) aa(j 1,i) r) ....

Seema Hiranandani, Ken Kennedy, Chau-Wen Tseng, and Scott Warren. The D Editor: A new interactive parallel programming tool. In Supercomputing '94, pages 733--42, November 1994.

....front end which allowed interactive transformations to be applied to a program. It had a library to examine and manipulate Intermediate Representation of the program. It was intended more as a restructuring tool than for performance programming. The Parascope programming environment, the D editor [8] and the subsequent work at Rice on compilation and performance evaluation environment is similar in spirit to ours. They focus more on data parallel programs in a message passing environment. We plan to incorporate results from data distributions for parallel programs running on Scalable Shared ....

Seema Hiranandani, Ken Kennedy, Chau-Wen Tseng, and ScottWarren. The D Editor: A New Interactive Parallel Programming Tool. In Proceedings of Supercomputing 1994. IEEE Computer Society, November 1994. Available from the CRPC technical archive http://www.crpc.rice.edu/CRPC/softlib/TRs online.html.

....Their system differs from ours in that their focus is on issues such as alias and dataflow analysis in the context of C and C programs. We also plan to incorporate results of the alias analysis from our compiler into our programming tool. The Parascope programming environment, the D editor [3] and the subsequent work at Rice on compilation and performance evaluation environment is similar in spirit to ours. They focus more on data parallel programs in a message passing environment. We plan to incorporate results from data distributions for parallel programs running on Scalable Shared ....

Seema Hiranandani, Ken Kennedy, Chau-Wen Tseng, and ScottWarren. The D Editor: A New Interactive Parallel Programming Tool. In Proceedings of Supercomputing 1994. IEEE Computer Society, November 1994. Available from the CRPC technical archive http://www.crpc.rice.edu/CRPC/softlib/TRs online.html.

....Decisions to improve program performance require knowledge about the performance characteristics of the program. Interactive programming tools may use dynamic profiling and or static performance estimation to identify code segments critical for the overall program performance [AMCA 95, MCAK94, HKTW94, Ree94] This information can be used to guide the user s program tuning efforts or to direct more aggressive compile time analysis and code improvement techniques to the identified performance hot spots . Compilers use static performance models to predict the performance benefits of program ....

S. Hiranandani, K. Kennedy, C.-W. Tseng, and S. Warren. The D Editor: A new interactive parallel programming tool. In Proceedings of Supercomputing '94, Washington, DC, November 1994.

....parallel systems have recently become a matter of considerable user and developer interest. Thinking Machines Corporation s commercial Prism environment [SAB 92] successfully demonstrated the principle, and motivated research groups to investigate similar support environments for Fortran D [HKTW94] and Vienna Fortran [PZ95] as well as other programming languages. Numerous more restricted efforts have also partially integrated various editors, compilers, debuggers and analysis tools in research and commercial contexts, some now coordinated under the auspices of the Parallel Tools ....

Seema Hiranandani, Ken Kennedy, Chau-Wen Tseng, and Scott Warren. The D Editor: A new interactive parallel programming tool. In Proc. Supercomputing'94 (Washington DC). IEEE Comp. Soc. Press, November 1994.

....for a programmer to decide whether an unexpected invariant is caused by a weakness in the proof system or a flaw in the program. In addition, the existing system processes only a first order language, though Bourdoncle explains how to extend the analysis [2:Section 5] Several environments [16, 4, 13, 27, 24] have been built for parallel programming languages to expose dependencies, thus allowing the programmer to tune programs to minimize these dependencies. In particular, MrSpidey has many similarities to the ParaScope [16, 4] and D editors [13] Both MrSpidey and the editors provide information at ....

....[2:Section 5] Several environments [16, 4, 13, 27, 24] have been built for parallel programming languages to expose dependencies, thus allowing the programmer to tune programs to minimize these dependencies. In particular, MrSpidey has many similarities to the ParaScope [16, 4] and D editors [13]. Both MrSpidey and the editors provide information at varying levels of granularity; both retain source correlation through transformations; and both depict dependencies graphically. However, unlike MrSpidey, the editors process a language with extremely simple control and data flow facilities, ....

Hiranandani, S., Kennedy, K., Tseng, C.- W., and Warren, S. The D editor: A new interactive parallel programming tool. In Proceedings of Supercomputing (1994).

....with parallel programming systems. However, recent efforts on developing portable high level parallel language systems has motivated work in integrated program analysis environments where performance debugging concerns are more closely coupled with the language s use for program development [10,14]. Of particular interest is the incorporation of performance prediction support in the programming environment for giving feedback to the user on algorithm implementation or to the compiler on optimization strategies [5,11] In most instances, however, there is a dependence on actual machine ....

S. Hiranandani, K. Kennedy, C.-W. Tseng, S. Warren, The D Editor: A New Interactive Parallel Programming Tool, Proc. Supercomputing'94, IEEE Computer Society Press, pp. 733-742, November 1994.

....with parallel programming systems. However, recent efforts on developing portable high level parallel language systems has motivated work in integrated program analysis environments where performance debugging concerns are more closely coupled with the language s use for program development [14,18]. Of particular interest is the incorporation of performance prediction support in the programming environment for giving feedback to the user on algorithm implementation or to the compiler on optimization strategies [6,7,15] In most instances, however, there is a dependence on actual machine ....

....a workstation environment without ever having to run the programs on target machines. FIGURE 7. SPEEDY Main Control Panel 14 Our experiments were done on a Sun workstation with a non preemptive threads package called Awesime. The simulation parameters were tuned to match a CM 5 machine [14,19,24]. Our first experiment is designed to show how various design choices can be made using speedy. We chose MatMul, a simple matrix multiplication program written in pC , where the two dimensional data distributions can be changed to improve the performance. We ran the experiments with two ....

S. Hiranandani, K. Kennedy, C.-W. Tseng, S. Warren,The D Editor: A New Interactive Parallel Programming Tool, Proc. Supercomputing '94, IEEE Computer Society Press, pages 733-742, November 1994.

....of messages can overlap communication with computation, hide message latency, and exploit pipeline parallelism. I also developed algorithms for interprocedural compilation of Fortran D [11] data parallel optimizations for Fortran 90 [5] and providing interactive user feedback in the D Editor [21]. Experiments on representative applications show the prototype Fortran D compiler produced code which approached the performance of hand optimized programs, and was four or more times faster on the CM 5 than code output by the best commercial compiler (CMF version 2.1 beta) at the time [20] The ....

S. Hiranandani, K. Kennedy, C.-W. Tseng, and S. Warren. The D Editor: A new interactive parallel programming tool. In Proceedings of Supercomputing '94, Washington, DC, November 1994.

....requires fewer language extensions, and relies on less run time support. The Fortran D compiler computes information to support the D Editor, a interactive parallel programming tool for presenting both static and dynamic information on interprocessor communication and parallel performance [27, 28, 29]. The compiler has also been used for evaluating an integer programming system for performing automatic data decomposition [30] Few researchers have published experimental results for large programs. Pingali Rogers apply message vectorization, message pipelining, and reduction recognition in ....

S. Hiranandani, K. Kennedy, C.-W. Tseng, and S. Warren. The D Editor: A new interactive parallel programming tool. In Proceedings of Supercomputing '94, Washington, DC, November 1994.

....Using the ParaScope Editor as a base, I built the D Editor to provide feedback for data parallel languages such as HPF and Fortran D. The D Editor provides novel information on partitioning, parallelism, and communication based on compile time analysis at the level of the original Fortran program [39]. It uses color coding and a collection of graphical displays to help the user to zoom in on portions of the program containing sequentialized code or expensive communication. The D Editor forms the centerpiece of the D System for data parallel languages [1, 35] ParaScope Dependence Analyzer ....

S. Hiranandani, K. Kennedy, C.-W. Tseng, and S. Warren. The D Editor: A new interactive parallel programming tool. In Proceedings of Supercomputing '94, Washington, DC, November 1994.

No context found.

S. Hiranandani, K. Kennedy, C.-W. Tseng, and S. Warren. The D Editor: A new interactive parallel programming tool. In Proceedings of Supercomputing '94, Washington, DC, November 1994.

No context found.

Seema Hiranandani, Ken Kennedy, Chau-Wen Tseng, and Scott Warren. The D Editor: A New InteractiveParallel Programming Tool. In Proceedings of Supercomputing '94, 1994.

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