PhD Preliminary: Ganesha Upadhyaya

Event
Tuesday, December 13, 2016 - 12:00pm
Event Type: 

Title: On Accelerating Big Code Analysis 
Date/Time: December 13th, 2016 @ 12:00 PM
Place: 223 Atanasoff  
Major Professor: Hridesh Rajan
Committee Members: Pavankumar Aduri, Wei Le, Robyn Lutz, Arun Somani(E CPE)

Abstract: 

Big Code analysis is about analyzing massive codebases. Big Code analysis has found numerous applications in defect prediction, discovery of programming patterns, specification inference, etc. Analyzing thousands of source code artifacts contained in Big Code, can be very expensive and needs to be accelerated. The current infrastructure support for Big Code analysis leverages parallelization techniques to accelerate. We believe understanding the interaction between the analysis and the source code artifact, and leveraging it, can further accelerate Big Code analysis. In this work, we present the notion of analysis-graph interactions and leverage it to perform a pre-analysis, that identifies the analysis relevant portions of the source code and performs program compaction to retain only the analysis relevant portions. In this way, running analysis on the reduced source code is much faster than running analysis on the original source code. We applied our technique to analyzing over 7 million control flow graphs (CFGs) and we achieved over 40% reduction in the analysis time. As part of our future work, we propose to apply the concept of analysis-graph interactions for program understanding and program characterization.

Ganesha Upadhyaya Prelim.pdf

Category: