J. M. Stone. Debugging concurrent processes: A case study. In ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI), June 1988.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....in the program, which is typically the place to be fixed by the programmer (using DEJAVU to fully replay runs if needed) 4) After the fix, the tests are re run. 2. A SIMPLE EXAMPLE As a simple motivating example, consider the shared queue program IntQueue.java in Figure 2 (adapted from [19]) The IntQueue class realizes a queue of integers. enqueue(elem) enqueues an integer number elem into the queue; dequeue( returns and dequeues the first element from the queue. Internally, the queue is realized as an integer array link where the entry link[elem] is the successor of elem; the ....

....we are unaware of any other technique that would automatically isolate failure inducing differences between schedules. Nonetheless, there is several related work: Manipulating schedules. The core idea of this paper, altering schedules to isolate failure causes, has first been suggested by Stone [19] as speculative replay. Her idea was to reduce the investigation of all possible [schedule] orderings to that of a few selected partial orderings by guiding the replay process according to (human inferred) thread dependencies. In contrast, our method is fully automatic; instead of having ....

J. M. Stone. Debugging concurrent processes: A case study. In Proceedings of ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI), pages 145--154, June 1988.

....all path analyses. Pathspecific datarace analysis can be useful for identifying specific execution paths in a multithreaded program that exhibit dataraces. This information could be useful in determining the input data or thread schedules neces sary to force a program to exhibit certain dataraces [43]. However, since path specific analyses can be expensive, we show in Section 7 how the path specific analyses pre sented in this section and Section 6 can be widened to all path poitts to atalysis and all path datarace aalysis. A dynamic execution of a multithreaded object oriented program ....

J. M. Stone. Debugging concurrent processes: A case study. In P'occcdings of ACM 'IGPLAN Cofe'cncc on P'og'amming Language Design and Implementation (PLDI), pages 14 153, June 1988.

....understood by all agents that belong to the same class. 3 Visual Parallel Programming Our approach to visual parallel programming is based on a visual formalism called the Process Communication Graph (PCG) 32] that originates from the space time diagram (STD) 24] and the concurrency map [37]. The latter two types of diagrams present the execution dynamics of parallel programs as a stream of events in a two dimensional space where one axis represents the time and another the individual processes. They have been used for debugging and tuning of parallel programs, but not for ....

J.M. Stone, "Debugging Concurrent Processes: A Case Study", Proc. ACM SIGPLAN 1988.

....synchronization together with symbols for conventional logical and control flow [5] The control and data flow graphs must be seamlessly combined into a unique graph formalism to provide a complete program description. Our approach originates in the space time diagram [9] and the concurrency map [14]. It combines different levels of abstraction at which parallel programs are expressed into a single Process Communication Graph (PCG) 12] The emphasis in PCG is on modeling the control (e.g. loops, conditionals, alternatives and synchronization) and communication structure (e.g. message passing ....

....0 by two MPI Reduce calls. The syntax for the call is [i]temp( totin. If the required precision was met, process 0 prints the result out, or else the whole procedure repeats. 7. CONCLUSION Visper is a new tool for visual parallel programming that extends the functionality of the concurrency map [14] and space time diagram [9] to handle the composition of parallel programs for MIMD architectures. It is based on MPI [8] and PVM [7] for writing messagepassing applications in a heterogeneous computing environment. Central to Visper is a graphical editor by which the programmer directly inputs a ....

Stone, J. M. Debugging Concurrent Processes: a Case Study. Proceedings of ACM SIGPLAN

....together with the arcs, as a link between the data flow and the control flow graphs. This example also shows that the alternative in PCG has an explicit spatial connotation. 6. Conclusion Visper is a new tool for visual parallel programming that extends the functionality of the concurrency map [11, 12] and the space time diagram [9] to handle the composition of parallel programs for MIMD architectures. Its set of visual symbols and constructs is based on PVM [7] and MPI [8] Central to Visper is a visual editor by which the programmer directly inputs a process communication graph. The graph is ....

Stone J M. Debugging concurrent processes: a case study, Proceedings of ACM SIGPLAN 1988 on Programming Language Design and Implementation, June 1988, pp.145-153.

....or may perform asynchronously. Whatever the possible solution, a visual programming tool must be scalable and easy to use, providing the user with a hardware independent graphical representation of the problem. Process communication graph defined in Visper originates from the concurrency map [10] and the spacetime diagram [9] These diagrams have been used for debugging and tuning of parallel programs, but not for construction. They present the execution dynamics of parallel programs as a stream of events in a twodimensional space where one axis represents the time and another the actual ....

....the execution path may continue from process 2 to process 0, or from process 2 to process 1. In the figure, a similar situation is drawn in D 2 R, for comparison. 9. Conclusion Visper is a new tool for visual parallel programming that extends the functionality of the concurrency map [10] and space time diagram [9] to handle the composition and generation of parallel programs for MIMD architectures. It supports MPI [7] for messagepassing applications in a heterogeneous computing environment. Visper has a graphical editor for the programmer to directly input a process ....

Stone J M. Debugging concurrent processes: a case study, Proceedings of ACM SIGPLAN 1988 on Programming Language Design and Implementation, June 1988, pp.145-153.

.... [11, 14] This approach is not applicable for constructing large parallel programs, since it is a non trivial task to obtain a readable diagram from an arbitrarily drawn graph [13] To avoid this problem, we propose a framework for constructing parallel programs based on Stone s concurrency map [12]. 2.1 Concurrency Map Originally, the concurrency map was designed for displaying information about a parallel program s execution, focusing on exhibiting concurrency in addition to inter process dependencies. It displays the process histories as event streams on a time grid. Each column of the ....

J. M. Stone. Debugging concurrent processes: a case study. In Proceedings of ACM SIGPLAN'88 on Programming Language Design and Implementation, pages 145--153. ACM, June 1988.

....effort, then detecting concurrency would be simple For a first step towards this aim, it might be useful to visualize concurrent regions with a concurrency map. This approach was proposed by Stone, who suggested the use of concurrency maps to support the visual analysis of concurrent processes [68, 69]. To this end, the local event streams of a distributed computation are partitioned into so called dependence blocks. The underlying idea is that all events contained in a dependence block can be regarded as a single, atomic super event , i.e. if one of these events is concurrent with a ....

J.M. Stone. "Debugging Concurrent Processes: A Case Study". Proc. SIGPLAN Conf. on Programming Language Design and Implementation, Atlanta, Georgia, pp. 145-153, June 1988.

....effort, then detecting concurrency would be simple For a first step towards this aim, it might be useful to visualize concurrent regions with a concurrency map. This approach was proposed by Stone, who suggested the use of concurrency maps to support the visual analysis of concurrent processes [Sto88, Sto89]. To this end, the local event streams of a distributed computation are partitioned into so called dependence blocks. The underlying idea is that all events contained in a dependence block can be regarded as a single, atomic super event , i.e. if one of these events is concurrent with a ....

J.M. Stone. "Debugging Concurrent Processes: A Case Study". Proc. SIGPLAN Conf. on Programming Language Design and Implementation, Atlanta, Georgia, pp. 145-153, June 1988.

No context found.

J. M. Stone. Debugging concurrent processes: A case study. In ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI), June 1988.

No context found.

J.M? Stone. Debugging concurrent processes: a case study. Proceedings of SIGPLAN'88 Conference on Programming Language Design and Implementation, ACM SIGPLAN Notices vol 23 no 7, Jul 1988.

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