Harold W. Cain, Barton P. Miller, and Brain J.N. Wylie. A Callgraph-Based Search Strategy for Automated Performance Diagnosis. Technical report, Computer Science Department, University of Wisconsin-Madison, Dayton Street, Madison, WI 53706-1685, USA, February 2000.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....response to earlier measurements. For diagnosing performance problems, the objective in doing this is to direct the programmer towards opportunities for improving performance. The idea was pioneered in the Paradyn Performance Consultant [9] which uses a callgraph based bottleneck search strategy [4]. This paper presents our progress in exploring this approach in the Java context. We address the following issues: Implementing dynamic instrumentation within a Java Virtual Machine. Several options are available; our implementation used our Veneer Virtual JVM, which o ers the prospect of ....

....the interference caused by the instrumentation it inserts, and throttles the number of ongoing experiments in order to keep the interference within speci ed limits. Re ning the focus by tracing the application s callgraph con nes the search to code which is actually executed (explored in [4]) For nding CPU bottlenecks, this approach is not clearly better than conventional approaches, such as sampling the program counter at random intervals. The real potential for the idea lies in bottlenecks which are harder to characterise where instrumenting for all hypotheses would lead to ....

[Article contains additional citation context not shown here]

H. Cain, B. Wylie, and B. P. Miller. A callgraph based search strategy for automated performance diagnosis. In Arndt Bode et al., editors, Euro-Par 2000 - Munich, volume 1900 of Lecture Notes in Computer Science, pages 108-122, 2000.

....lies in the ability to deploy instrumentation algorithmically. The idea is to formulate a hypothesis about a performance bottleneck, deploy an experiment to test it, then re ne the hypothesis on the basis of the results. The approach has been explored in the Paradyn Performance Consultant [5, 15]. We have a preliminary implementation (as a JUDI ISC) which we are currently developing [3] Aspects. How should a JUDI user specify which program points a given instrument should be attached to Aspect oriented programming (AOP) languages such as AspectJ o er an answer to this [11] AspectJ is ....

H. Cain, B. Wylie, and B. P. Miller. A callgraph based search strategy for automated performance diagnosis. In Arndt Bode, Thomas Ludwig II, Wolfgang Karl, and Roland Wismuller, editors, Euro-Par. Springer Verlag, LNCS 1900.

....Figure 4 shows an excerpt of an OpenMP code together with its associated DRG. Call graph techniques have been widely used in performance analysis. Tools such as Vampir [20] gprof [11, 10] CXperf [14] support a call graph which shows how much time was spent in each function and its children. In [7] a call graph is used to improve the search strategy for automated performance diagnosis. However, nodes of the call graph in these tools represent function calls [10, 14] In contrast our DRG de nes a node as an arbitrary code region (e.g. function, function call, loop, statement, etc. 6 ....

....a variety of performance overheads for Fortran77 OpenMP programs. Paradyn [18] is an automatic performance analysis tool that uses dynamic instrumentation and searches for performance bottlenecks based on a speci cation language. A function call graph is employed to improve performance tuning [7]. Recent work on an OpenMP performance interface [19] based on directive rewriting has similarities to the SIS instrumentation approach in SCALEA and to SCALA [9] the predecessor system of SCALEA. Implementation of the interface (e.g. in a performance measurement library such as TAU) allows ....

Harold W. Cain, Barton P. Miller, and Brian J.N. Wylie. A callgraph-based search strategy for automated performance diagnosis. In Euro-Par

....address to the function name. The result is that we eventually generate a complete call graph for the application. This approach is similar to the one used by the DynInst based Paradyn performance tools, when it automatically instruments an application program in searching for performance problems [2]. Function Argument Parser The function argument parser makes it easy to track the type and name of each parameter to a function in the application, if either the program was compiled with debugging information or the user provides this information. We developed a tool that helps automate this ....

H.W. Cain, B.P. Miller, and B.J.N. Wylie, "A Callgraph-Based Search Strategy for Automated Performance Diagnosis", Euro-Par

....in an optimized group because they are not recognized in the call graph traversal. With additional kernel instrumentation (at an indirect call site) the call graph can be updated when a heretofore unseen callee is encountered, allowing indirect callees to be included in an optimized group [5]. Another candidate for future work is removal of user involvement in the initial step of choosing the group s root function, thus allowing all steps to be performed automatically. Paradyn s Performance Consultant [5, 11] has shown that bottlenecks can be automatically located for non threaded ....

....is encountered, allowing indirect callees to be included in an optimized group [5] Another candidate for future work is removal of user involvement in the initial step of choosing the group s root function, thus allowing all steps to be performed automatically. Paradyn s Performance Consultant [5, 11] has shown that bottlenecks can be automatically located for non threaded user programs, via a call graph traversal. Other than emitting all hot chunks before any cold chunks, the relative placement of functions within a group is arbitrary. With future work, basic block positioning can be ....

H. Cain, B.P. Miller, and B.J.N. Wylie. A Callgraph-Based Search Strategy for Automated Performance Diagnosis. European Conference on Parallel Computing (Euro-Par), Munich, Germany, August 2000.

....productive using an automated performance problem search tool; the analysis expertise is embodied in the tool s search strategy. The capability to automate searches for performance problems is enhanced by incorporating structural information about the application under study such as its call graph [4], or by pruning and prioritizing the space in which the search takes place based on the application s behavior during previous runs [13] We have developed a new technique that uses sampling [1,2,8] to augment an automated search for performance problems to attack the problem of code size ....

....found the application as a whole to be spending too much time doing synchronization, it refines its search to consider whether the application is spending too much time performing synchronization operations in the processes of the application. The current Performance Consultant search strategy [4] uses the application s call graph structure to guide its refinement through the application code. For example, if the Performance Consultant has found that an MPI application as a whole is spending too much time sending messages, it evaluates whether the application is spending too much time ....

[Article contains additional citation context not shown here]

H. W. Cain, B. P. Miller, and B. J. N. Wylie, "A Callgraph-Based Search Strategy for Automated Performance Diagnosis", Euro-Par 2000, Munich, Germany, August 2000.

No context found.

H.W. Cain, B.P. Miller, and B.J.N. Wylie, "A Callgraph-Based Search Strategy for Automated Performance Diagnosis", Euro-Par 2000.

....With an automated search tool, the user need not be a performance analysis expert to find application performance problems because the expertise is embodied in the tool. Automated search tools benefit from the use of structural information about the application under study such as its call graph [4] and by pruning and prioritizing the search space based on the application s behavior during previous runs [12] To attack the problem of scalability with respect to application code size, we have developed Deep Start, a new algorithm that uses sampling [1,2,8] to augment automated search. Our ....

....PC is collecting data for the experiment) and whether the experiment s data has proven the experiment s hypothesis to be true, false, or not yet known. The Deep Start search algorithm augments the PC s current call graph based search strategy with stack sampling. The PC s current search strategy [4] uses the application s call graph to guide refinement. For example, if it has found that an MPI application is spending too much time sending messages, the PC starts at the main function and tries to refine its search to form experiments that test the functions that main calls. If a function s ....

Cain, H.W., Miller, B.P., Wylie, B.J.N.: A Callgraph-Based Search Strategy for Automated Performance Diagnosis. In Bode, A., Ludwig, T., Karl, W., Wismller, R. (eds): LNCS

....after each procedure call, to stop and restart the event accumulation, respectively. Second, inclusive metrics help find performance bottlenecks more quickly; traversing the call graph with an inclusive time metric has been shown effective in automatically searching for bottlenecks in user code [14]. 1.4 Evolving Kernels The final contribution of this dissertation is a framework for evolving kernels. An evolving kernel is one that, automatically and at run time, adapts and modifies its code in response to the run time environment. Evolving kernels are motivated the the need for kernel ....

....for a performance bottleneck, for example) to instrument only a few top level routines at a time. If the results show a bottleneck in a function, then the instrumentation for it can be removed and applied to the function s callees. This strategy can be used to automate the search for bottlenecks [14, 42], with little run time overhead, because at any given time only a few functions are instrumented. This algorithm is prohibitive with any static instrumentation system, because the program would have to be re run every time the instrumentation changes. Static instrumentation systems typically ....

[Article contains additional citation context not shown here]

H. Cain, B.P. Miller, and B.J.N. Wylie. A Callgraph-Based Search Strategy for Automated Performance Diagnosis. European Conference on Parallel Computing (Euro-Par), Munich, Germany, August 2000.

No context found.

Harold W. Cain, Barton P. Miller, and Brain J.N. Wylie. A Callgraph-Based Search Strategy for Automated Performance Diagnosis. Technical report, Computer Science Department, University of Wisconsin-Madison, Dayton Street, Madison, WI 53706-1685, USA, February 2000.

No context found.

H. W. Cain, B. P. Miller, and B. J. N. Wylie. A Callgraph-Based Search Strategy for Automated Performance Diagnosis. In Proc. of the 6th International Euro-Par Conference, volume 1999.

No context found.

Harold W. Cain, Barton P. Miller, and Brian J.N. Wylie. A callgraph-based search strategy for automated performance diagnosis. In Euro-Par 2000.

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