R. Bruce Irvin and Barton P. Miller. Multi-application support in a parallel program performance tool. Technical Report CS-TR-93-1135, Computer Sciences Department, University of Wisconsin - Madison, February 1993. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Experiment Management Support for Parallel Performance Tuning - Karavanic (1999) (Correct)
....efforts have examined methods for using visualization of performance data to compare different versions of programs. Ribler et al. at Virginia [65] have developed alternative methods for visualizing categorical trace data, including recent work on simultaneously visualizing multiple traces. IPS 2 [40] allowed visualization of data from different runs, matching them by normalizing to total execution time. Our previous work combined Paradyn and Devise [42] to enable side by side, linked visualization of data from multiple executions of a parallel application. Open problems remain, for example in ....
R. Bruce Irvin and Barton P. Miller. Multi-application support in a parallel program performance tool. Technical Report CS-TR-93-1135, Computer Sciences Department, University of Wisconsin - Madison, February 1993.
Lydia Workshop - Hellas Load (Correct)
....the complexity is open [6] One important case, however, with a long history and which admits efficient solution is that in which k = 2 and all tasks have the same execution time. The first polynomial time solution O(n 4 ) for that problem was published by Fujii, Kasami and Ninomiya in 1969 [3]. In 1972, Coffman and Graham published an O(n 2 ) algorithm [1] while, in 1982, Gabow [4] found an O(m nff(n) algorithm that, combined with Tarjan s union find result [5] becomes asymptotically optimal ( ff(n) is an inverse of Ackermann s function and is very slow growing) The only ....
....i j k and such that if x precedes y in G, x 2 V i and y 2 V j , then i j. The length of the schedule is t. Here the goal is to find a schedule so that to minimize t. The following theorem gives the fundamental relation between matchings and 2 schedules. Theorem 2.1. Fujii, Kasami and Ninomiya [3]) For G and G defined above, the paired tasks in any optimal schedule for G form a maximum matching in G. Furthermore, any maximum matching in G can be transformed into an optimal schedule for G, leaving the set of unpaired tasks unchanged. Note that not all maximum matchings in G correspond to ....
[Article contains additional citation context not shown here]
R. Bruce Irvin, Barton P. Miller, "Multiapplication Support in a Parallel-Program Performance Tool", IEEE Parallel & Distributed Technology, Spring 1994, pp. 40-50.
Monitoring of Distributed Memory Multicomputer Programs - van Riek, Tourancheau.. (1993) (4 citations) (Correct)
....e2 Proc 1 Proc 2 Observer Figure 3: The effect of totally ordering a partially ordered set of events. The plain arrows represent communications between processors. With this total ordering, the observer is thus wrong from the causal relationship point of view. 2. 10.2 Practical implementations [IM93] shows that the global time precision can influence the understanding of the trace based representation. Despite the fact that enforcing a total order in a set of events theoreticaly results in a loss of information, this method has been generally adopted because it is practically a very good way ....
R. Irvin and P. B. Miller. Multi-application support in a parallel program performance tool. Technical Report 1135, University of Wisconsin-Madison, 1210 W Dayton Street, Madison, Wisconsin 53706, 1993. accessible by anonymous FTP grilled.cs.wisc.edu:technical papers/multiapp.ps.Z.
Experiment Management Support for Performance Tuning - Karavanic, Miller (1997) (12 citations) Self-citation (Miller) (Correct)
....histograms. The result of the query is 9 Paradyn histograms. Here we display the result using xmgr. 4 RELATED WORK There has been extensive work on parallel profiling tools [11,12,3,2,13] and a more limited amount of work on tools that allow displays of performance data from multiple runs [14,21]. In the debugging area, work has been done to find bugs by comparing the execution of a new program version to an old one by comparing program output and user selected variables[16] GUARD compares the program currently being debugged with a reference version of the same program, to detect ....
....views of the PEG: the display on the left has the module machine expanded, while the display on the right results from expanding the process node. Figure 9: Results of the metrics CPUtime, IO Time, and Sync Wait for a Program Event Group. Page 11 point of execution in the other graph. IPS 2 [21] allowed linked visualization of data from different runs, matching them by normalizing to total execution time. Ribler, Abrams et al. Virginia [26] have developed alternative methods for visualizing categorical trace data, including recent work on visualizing multiple traces simultaneously. Our ....
R. Bruce Irvin and Barton P. Miller. Multi-Application Support in a Parallel Program Performance Tool. Technical Report #CS-TR-93-1135, February 1993, Computer Sciences Department, University of Wisconsin - Madison.
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