Bernd Mohr, Darryl Brown, and Allen Malony. TAU: A Portable Parallel Program Analysis Environment for pC++. In Proceedings of CONPAR 94 - VAAP VI, Linz, Austria, Springer Verlag, LNCS 854, pages 29-40, September 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....SCALA supports several different trace formats including ALOG and Grace formats for collected performance data which enables the usage of various performance visualization systems. Based on the ALOG trace format we can use well established visualization systems such as Medea [7] TAU [36], and Upshot [23] Medea is a post mortem performance analysis and visualization system. Among others we use Medea to derive and visualize performance metrics together with the input program. This work has been described in detail in [8] Figure 5 shows Medea visualizing various performance ....

....system. Instrumentation is currently restricted to functions and is controlled by a consultant module. Performance bottlenecks are tried to be found automatically through a rule based refinement system. The user can control instrumentation overhead by limiting instrumentation to a threshold. TAU [36] is a sophisticated instrumentation, tracing and profiling system that has been shown to be very useful for various programming paradigms including PC and HPC [30] Forge90 [32] reports on communication costs at the level of a generated message passing code, but not at the level of the input ....

B. Mohr, D. Brown, and A. Malony. TAU: A portable parallel program analysis environment for pC++. In CONPAR, Linz, Austria, 94.

....formalizes trace files by using an event class. This paper does not contain any property which incorporates events, which may change in future work. 5 Note that due to advanced monitoring and profiling technologies (e.g. dynamic profiling [14] hardware profiling [5] and source code profiling [19, 15]) there is basically no barrier to obtain experiment related data for arbitrary parallel and distributed programs. 4.2 Filters and Statistics for Experiment related Data It order to provide flexible mechanisms to describe performance problems we commonly require filters (used to restrict the ....

B. Mohr, D. Brown, and A. Malony. TAU: A portable parallel program analysis environment for pC++. In CONPAR, Linz, Austria, 94.

....a parallel program. Executing these IF conditions is considered to be compiler overhead. Class HPFProcessSummary contains all attributes of class HPFSummary with respect to a specific process (identified by attribute process) Note that dynamic performance related information can both be measured [16, 17, 20, 8] or predicted [10] Some of this information however require a close interaction between performance measurement prediction tools and compilers [10, 8, 2] 10 property is PROPERTY pp name ( arg list ) f [LET def IN] pp condition pp confidence pp severity g; arg is type ident ....

B. Mohr, D. Brown, and A. Malony. TAU: A portable parallel program analysis environment for pC++. In CONPAR, Linz, Austria, 94. 20

....the usage of several performance visualization systems. The ALOG format (part of the MPICH implementation) describes time stamped events and is the de facto trace format for MPI. Based on the ALOG trace format we can use visualization systems of well established systems such as Medea [12] TAU [56], and Upshot [45] Medea is a post mortem performance analysis and visualization system. This system has been extended for performance analysis and visualization of HPF programs that are parallelized by VFC compiler. SCALA is used to instrument programs and to generate ALOG trace files. Among ....

B. Mohr, D. Brown, and A. Malony. TAU: A portable parallel program analysis environment for pC++. In CONPAR, Linz, Austria, 94.

....programs. Scala supports two trace formats ALOG and Scala specific format for collected performance data which enables the usage of several performance visualization systems. Based on the ALOG trace format we can use visualization systems of well established systems such as Medea [6] TAU [31], and Upshot [19] Medea is a post mortem performance analysis and visualization system. Among others we use Medea to derive and visualize performance metrics together with the input program based on ALOG trace files. This work has been described in detail in [7] Scala also generates Grace [17] ....

....examine various data distributions for the diagonalization routines. Thereafter, we plan to parallelize the setup phase of the spherical part HMT . Overall we are very confident that HPF has to potential to parallelize large and substantial portions of the WIEN97 application. 6 Related Work TAU [31] is a sophisticated instrumentation, tracing and profiling system that has been shown to be very useful for various programming paradigms including PC and HPC [26] 14 Forge90 [29] reports on communication costs at the level of a generated message passing code, but not at the level of the ....

B. Mohr, D. Brown, and A. Malony. TAU: A portable parallel program analysis environment for pC++. In CONPAR, Linz, Austria, 94. 16

....describes, for a particular tool, which of the performance related data required for performance property specification can be obtained from that tool. Moreover, query commands to access this data can be indicated. Examples for such tools are PARADYN [MCCHI 95] SCALA [FaScPa 99] TAU [MoBrMa 94] and VAMPIR [NaArWeHoSo 96] etc. Based on the supplied data specifications, an automated performance analysis environment can use existing tools to access relevant data in the search for performance problems and bottlenecks. An integrated system combining all three components should ....

B. Mohr and D. Brown and A. Malony. TAU: A portable parallel program analysis environment for pC++. CONPAR, Linz, Austria, 1994.

....summing an array and matrix vector multiplication. And let s look at actual program runs. I implemented both algoritms in pC ( 4] run them with all possible mapping specifications (on a Sequent Symmetry with up to 24 processors) and analysed the program runs with the help of the TAU ([5]) environment. Appendix 6 shows the code. 2.3 Summing an array 2.3.1 Discussing the algorithm Let s sum an array of 16 elements (0 to 15) with a tree like parallel algorithm (similar to appendix 1) often referred to as reduction: There are (log 16) 4 steps, and in the first step we update 0 ....

B. Mohr, D. Brown, A. Malony. TAU: a portable parallel program analysis environment for pC++. University of Oregon,

....for the Cray T3D, Prism[27] for the TMC CM 5 and MPPE[18] from Maspar. In the research world, examples include the Pablo system from the Universities of Illinois[24,1] that can trace Fortran D programs and present this information in terms of the source program, and TAU from University of Oregon[20] that can do similar operations for pC programs[2] While presenting performance data at the source code level of a high level parallel language is crucial, it is not always sufficient. The language abstractions can insulate a programmer from the need to specify the low level details, but they ....

Bernd Mohr, Darryl Brown, and Allen Malony. Tau: A portable parallel program analysis environment for pC++. In International Conference on Parallel Systems, pages 29--40. Springer Verlag, September 1994.

.... inadequate for the new style of parallel programming for reasons that we detailed earlier (mainly their lack of handling distributed data structures and relating performance back to the user level source) The University of Oregon is developing performance analysis tools for pC , called Tau [12], and have similar goals of extensibility and customization that we have described in the framework. May and Berman stress the importance of extensibility and creation of new views in the context of a parallel debugger Panorama[11] Both Tau and Panorama have Tcl as their extension language. Our ....

B. Mohr, D. Brown, and A. Malony. Tau: A portable parallel program analysis environment for pc++. Technical report, University of Oregon, 1994. Available as conpar94.ftp.ps from moose.cs.indiana.edu:pub/sage.

.... has also been used to construct a suit of programming environment tools: tuning and analysis utility (TAU) file and class display (Fancy) call graph extended display (Cagey) class hierarchy browser (Classy) routine and data access profile display (Racy) and event and state display (Easy) MBM94, BHMM94] Summary Sage is a convenient tool for building source to source translators. It is not tied to any particular hardware architecture which makes it portable, and it provides basic routines needed by compilers. Though Sage has been extended to accept new source languages several ....

B. Mohr, D. Brown, and A. Malony. TAU: A portable parallel program analysis environment for pC++. In Proceedings of CONPAR 94 - VAPP VI, University of Linz, Austria, September 1994. LNCS 854.

.... from the program to the end user (i.e. to an interactive user interface or to an adaptation algorithm) Instead, they focus on reducing or controlling program perturbation due to performance monitoring[32] A further distinction between Falcon and other projects on performance debugging [37, 30, 3] is Falcon s support of application specific monitoring. Such support is essential when end users wish to use monitoring output to steer their programs or to simply understand their runtime behaviors in terms of familiar quantities (e.g. total energy in the MD application) Related research in ....

B. Mohr, D. Brown, and A. Malony. Tau: A portable parallel program analysis environment for pc++. In Proceedings of CONPAR 94 - VAPP VI, pages 29--40. University of Linz, September 1994. LNCS 854.

.... has also been used to construct a suit of programming environment tools: tuning and analysis utility (TAU) file and class display (Fancy) call graph extended display (Cagey) class hierarchy browser (Classy) routine and data access profile display (Racy) and event and state display (Easy) MBM94, BHMM94] Summary Sage is a convenient tool for building source to source translators. It is not tied to any particular hardware architecture which makes it portable, and it provides basic routines needed by compilers. Though Sage has been extended to accept new source languages several ....

B. Mohr, D. Brown, and A. Malony. TAU: A portable parallel program analysis environment for pC++. In Proceedings of CONPAR 94 - VAPP VI, University of Linz, Austria, September 1994. LNCS 854.

....solve the performance problems, and generates concise hints which are communicated to runtime libraries by a hints file . 2. 3 Previous work Performance analysis research has concentrated on visually displaying performance data [1, 2] relating performance data to high level language constructs [3, 4], or giving the user insights into performance problems using expert analysis [5, 6] Our framework aims to go a step further in the direction of automation : Paradise not only finds performance problems, but also solutions in terms of optimizations for the problem areas, and in co operation with ....

B. Mohr, D. Brown, and A. Malony. TAU: A Portable Parallel Program Analysis Environment for pC++. In Proceedings of the 3rd Joint Conference on Parallel Processing: CONPAR 94 - VAPP VI, September 1994.

....performance tools for functional languages to measure and present information about implicit lowlevel activity. A growing number of performance tools are designed for parallel object oriented languages. Most notably, TAU is the official performance analysis system for the pC programming language [48]. It uses source level event traces to display parallel performance information in terms of class methods and uses a rich program information database [6] to relate events to source code. The Projections tool presents language specific performance information about the distributed object oriented ....

....on the NV model. Data parallel language tools have shown how to provide control (code) views of performance [1,41,60,68] and we adopt their techniques for attaching performance data to code displays. Object oriented language tools offer natural hierarchical structures for organizing noun data [48] and demonstrate techniques for interpreting compiler output for mapping purposes. The standardization of compiler generated mapping data is still an unsolved problem, but these systems demonstrate its potential and we push these techniques further in this dissertation. Event mapping tools ....

Bernd Mohr, Darryl Brown, and Allen Malony. Tau: A portable parallel program analysis environment for pC++. In International Conference on Parallel Systems, pages 29--40. Springer Verlag, September 1994.

....not included. C programs are visualised by using trace visualisation, which is visualisation based on program execution traces. MVD also supports statistical visualisation, in which statistics of program execution is displayed, such as task activity status, and stack high water marks. The TAU [2, 13] visualisation tools are part of the pC programming language system [1] pC itself is an extension of C , with several additional constructs to support data parallelism. The TAU tools are implemented as a graphical hypertool, in which it is a composition of several tools, each of which ....

....many such tools as represented above use language support for providing information for driving program visualisation. For example, C and PARC were specifically created by incorporating features for easily extracting data for visualisation. Therefore, they are visualisation conscious [13]. Other languages, such as LAMINA, do not have such support. Nevertheless, their underlying run time systems permit easy incorporation of such support. However, the majority of concurrent object oriented languages are not visualisation conscious [13] Visualisation is usually an after thought. ....

[Article contains additional citation context not shown here]

B. Mohr, D. Brown and A. Malony. TAU: A Portable Parallel Program Analysis Environment for pC++. In B. Buchberger and J. Volkert (editors), Lecture Notes in Computer Science volume 854, Proceedings of the International Conference on Vector and Parallel Processing, CONPAR'94, pages 29--40. Springer-Verlag; Berlin, Germany, 1994.

....latency and low monitoring perturbation are conflicting attributes, Falcon offers several mechanisms to balance the tradeoff between latency and perturbation and this paper includes a study of the such tradeoffs. A further distinction between Falcon and other projects on performance debugging [40, 32, 3] is derived from Falcon s support of application specific monitoring. Such support is essential when end users wish to use monitoring output to steer their programs or to simply understand their runtime behaviors in terms of familiar quantities (e.g. total energy in the MD application) This work ....

B. Mohr, D. Brown, and A. Malony. Tau: A portable parallel program analysis environment for pc++. In Proceedings of CONPAR 94 - VAPP VI, pages 29--40. University of Linz, September 1994. LNCS 854.

....to provide a parallel programming environment, including an editor, debugger and an automatic data partitioner. Polaris allows programmers to provide instructions to the compiler through source code assertions. Sage provides a rich set of tools for pC named Tuning and Analysis Utilities, TAU [8, 29]. TAU includes tools for file and class display, call graph display, class hierarchy browsing, routine and data access profile display, and event and state display. Almost all of these systems are research tools that encourage user experimentation. Experimentation is further facilitated by having ....

B. Mohr, D. Brown, and A. Malony. TAU: A portable parallel program analysis environment for pC++. In Proceedings of CONPAR 94 - VAPP VI, University of Linz, Austria, September 1994. LNCS 854.

....describes, for a particular tool, which of the performance related data required for performance property specification can be obtained from that tool. Moreover, query commands to access this data can be indicated. Examples for such tools are PARADYN [MCCHI 95] SCALA [FaScPa 99] TAU [MoBrMa 94] and VAMPIR [NaArWeHoSo 96] etc. Based on the supplied data specifications, an automated performance analysis environment can use existing tools to access relevant data in the search for performance problems and bottlenecks. An integrated system combining all three components should ....

B. Mohr and D. Brown and A. Malony. TAU: A portable parallel program analysis environment for pC++. CONPAR, Linz, Austria, 1994.

....performance analysis tools developed partly by other groups [6, 7, 8, 9] The toolset provides support for accessing static information about the program and for querying and analyzing dynamic data obtained from program execution. A more detailed discussion of the tools can be found in [10, 11, 12]. 4 A Tour of the Sort Program: Static Analysis The pC implementation of the Sort program is easily derived from the class object defined above. A pC collection of Pvector objects is created. The member functions of the Pvector class are divided into two categories: those functions that ....

B. Mohr, D. Brown, A. Malony, TAU: A Portable Parallel Program Analysis Environment for pC ++ , Proc. of CONPAR 94 - VAPP VI, Linz, Austria, Springer Verlag, LNCS 854, pp. 29--40, Sept. 1994.

....from OACIS, and by IBM R D contract MHVU3704. y Department of Computer and Information Science, University of Oregon, Eugene, Oregon 97403, fdarrylb, hacks, malony, mohrg cs.uoregon.edu. 2 Malony et al. of the static and dynamic analysis tools have been covered in detail elsewhere (see [9, 10]) Instead of reiterating that discussion (a brief description is given in x2.1 and x2.2) we focus on the latest tool for breakpoint based program analysis in x3. Future directions are discussed in x4. 2 The pC Program Analysis Environment We advocate the development of language specific ....

....The toolset provides support for accessing static information about the program and for querying and analyzing dynamic data obtained from program execution. The static and dynamic tools of the environment are briefly described below; a more detailed discussion of these tools can be found in [9, 10]. 2.1 Static Analysis Tools One of the basic motivations behind using C as the base for a new parallel language is its proven support for developing and maintaining complex and large applications. However, to apply the C language capabilities effectively, users require support tools to manage ....

B. Mohr, D. Brown, A. Malony, TAU: A Portable Parallel Program Analysis Environment for pC++ , Proc. of CONPAR 94 - VAPP VI, Linz, Austria, Springer Verlag, LNCS, Sept. 1994.

No context found.

B. Mohr, D. Brown, and A. Malony. TAU: A Portable Parallel Program Analysis Environment for pC++. In B. Buchberger and J. Volkert, editors, Proceedings of CONPAR 94--VAPP VI, volume 854 of Lecture Notes in Computer Science, pages 29--40. Springer-Verlag, September 1994.

....its current state is given in Section 1.6. Section 1.7 discusses future development of , particularly its extension to other parallel language environments and the incorporation of more sophisticated tools. 1. 2 Design Requirements and Goals An earlier version of this material was published in [9]. was designed to improve parallel programming productivity by combining advances in parallel debugging, performance evaluation, and program visualization tools. We feel that the requirements the design and the tools had to address are common to next generation parallel programming ....

....tool application is well understood. However, this cannot be fully realized unless tools can leverage other programming system technologies (e.g. use the compiler to implement instrumentation) 1.3 TAU Overview Parts of this material (Section 1.3 and Section 1.3. 1) were previously published in [9]. The architecture defines how its components interoperate and fit in the pC language system. Below, we describe the design and show how it addresses the programming productivity requirements of pC . is not a general solution to the problem of parallel program analysis. Instead, our goal ....

[Article contains additional citation context not shown here]

Bernd Mohr, Darryl Brown, and Allen Malony. TAU: A Portable Parallel Program Analysis Environment for pC++. In Bruno Buchberger and Jens Volkert, editors, Proceedings of CONPAR 94--VAPP VI, volume 854 of Lecture Notes in Computer Science, pages 29--40. Springer-Verlag, September 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 ....

.... browser called racy (Routine and data ACcess profile displaY) an event trace browser called easy (Event And State displaY) and a breakpoint debugger called breezy (BReakpoint Executive Environment for visualiZation and data displaY) A more detailed discussion of the t tools can be found in [4,12,14]. 3 ExtraP A Performance Extrapolation Tool for pC ExtraP is a performance extrapolation system for pC that has been integrated into t in the guise of speedy. This section explains the modeling approach of extrapolation and the techniques used by ExtraP. 3.1 Performance Extrapolation ....

B. Mohr, D. Brown, A. Malony, TAU: A Portable Parallel Program Analysis Environment for pC++, Proc. of CONPAR 94 - VAPP VI, Linz, Austria, Springer Verlag, LNCS 854, pp. 29-40, September 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 ....

....called breezy (BReakpoint Executive Environment for visualiZation and data displaY) To generate the dynamic execution data for these tools, profiling and tracing instrumentation and measurement support has been implemented in pC . A more detailed discussion of the t tools can be found in [5,16,18]. 3 ExtraP A Performance Extrapolation Tool for pC ExtraP is a performance extrapolation system for pC that has been integrated into t in the form of speedy. This section explains the modeling approach of extrapolation and the techniques used by ExtraP. 3.1 Performance Extrapolation ....

B. Mohr, D. Brown, A. Malony, TAU: A Portable Parallel Program Analysis Environment for pC++, Proc. of CONPAR 94 - VAPP VI, Linz, Austria, Springer Verlag, LNCS 854, pages 29-40, September 1994.

No context found.

Bernd Mohr, Darryl Brown, and Allen Malony. TAU: A Portable Parallel Program Analysis Environment for pC++. In Proceedings of CONPAR 94 - VAAP VI, Linz, Austria, Springer Verlag, LNCS 854, pages 29-40, September 1994.

No context found.

B. Mohr, D. Brown, and A. Malony. TAU: A portable parallel program analysis environment for pC++. In CONPAR, Linz, Austria, 94.

No context found.

Mohr B., Brown D., Malony A.: TAU: A Portable Parallel Program Analysis Environment for pC++. In \Proceedings of CONPAR 94{VAPP VI", University of Linz, Austria 1994.

No context found.

B. Mohr, D. Brown, and A. Malony, "TAU: A portable parallel program analysis environment for pC++", CONPAR 94 - VAPP VI, Linz, Austria, September 1994.

No context found.

B. Mohr, D. Brown, and A. Malony. TAU: A portable parallel program analysis environment for pC++. In CONPAR, Linz, Austria, 94.

No context found.

B. Mohr, D. Brown, and A. Malony, Tau: A portable parallel program analysis environment for pc++, tech. report, University of Oregon Department of Computer and Information Science, Eugene, Oregon 97403.

No context found.

B. Mohr, D. Brown and A. Malony. TAU: A Portable Parallel Program Analysis Environment for pC++. In B. Buchberger and J. Volkert (editors), Lecture Notes in Computer Science volume 854, Proceedings of the International Conference on Vector and Parallel Processing, CONPAR'94, pages 29--40. Springer-Verlag; Berlin, Germany, 1994.

No context found.

B. Mohr, D. Brown, and A. Malony. TAU: A Portable Parallel Program Analysis Environment for pC++. In Proceedings of the 3rd Joint Conference on Parallel Processing: CONPAR 94 - VAPP VI, September 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