This shows you the differences between two versions of the page.

Link to this comparison view

seminars:seminar_10_23_02 [2017/09/20 22:02] (current)
Line 1: Line 1:
 +====== Butterfly Analysis: Adapting Dataflow Analysis to Dynamic Parallel Monitoring ​ ======
 +Tuesday Feb. 23, 2010\\
 +Hamerschlag Hall D-210\\
 +4:00 pm\\
 +**[[www.ece.cmu.edu|Michelle Goodstein]]**\\
 +Online Program monitoring is an effective technique for detecting bugs and 
 +security attacks in running applications. Extending these tools to monitor ​
 +parallel programs is challenging because the tools must account for 
 +inter-thread dependences and relaxed memory consistency models. Existing ​
 +tools assume sequential consistency and often slow down the monitored ​
 +program by orders of magnitude. In this talk, we present a novel approach ​
 +that avoids these pitfalls by not relying on strong consistency models or 
 +detailed inter-thread dependence tracking. Instead, we only assume that 
 +events in the distant past on all threads have become visible; we make no 
 +assumptions of the relative ordering of more recent events on other 
 +threads. To overcome the potential state explosion of considering all the 
 +possible orderings among recent events, we adapt two techniques from 
 +static dataflow analysis, reaching definitions and reaching expressions, ​
 +to the domain of dynamic parallel monitoring and show how our adapted ​
 +analysis can be used in two popular memory and security tools. Significant ​
 +modifications to these techniques are proposed to ensure the correctness ​
 +and efficiency of our approach.
 +Our simulation study on a collection of Splash-2 and Parsec 2.0 benchmarks ​
 +demonstrates the benefits in trading off a low false positive rate for (i) 
 +reduced overhead and (ii) the ability to run on relaxed consistency models.
 +Michelle Goodstein is a Ph.D. student in Computer Science at Carnegie Mellon Uni-versity,​ advised by Todd Mowry. She is interested in dynamic parallel program analysis, and works in the Log-Based Architecture ​ Group. ​ She received dual B.S. degrees in Mathematics and Computer Science from the University of Washington. ​ Michelle is a former recipient of the Clare Booth Luce Fellowship.
 +**[[seminars| Back to the seminar page]]**