Abstract:

This report describes a set of experiments to evaluate qualitatively the effectiveness of Data Structure Analysis (DSA) in identifying properties of a program's data structures. We manually inspected several benchmarks to identify linked data structures and their properties, and compared these against the results produced by DSA. The properties we considered are those that were the primary goals of DSA: distinguishing different kinds of data structures, distinct instances of a particular kind, type information for objects within an LDS, and information about the lifetime of such objects (particularly, those local to a function rather than global). We define a set of metrics for the DS graphs computed by DSA that we use to summarize our results concisely for each benchmark. The results of the study are summarized in the last section.

Download:

Paper:

• How Successful is Data Structure Analysis in Isolating and Analyzing Linked Data Structures? (PS)
• How Successful is Data Structure Analysis in Isolating and Analyzing Linked Data Structures? (PDF)

BibTeX Entry:

  @TechReport{DSAEvaluation:TR05,
    Author     = {Patrick Meredith and Balpreet Pankaj and Swarup Sahoo and
		  Chris Lattner and Vikram Adve}, 
    Title      = "How Successful Is Data Structure Analysis in Isolating and  
Analyzing Linked Data Structures?",
    Institution= {Computer Science Dept.,
                  Univ. of Illinois at Urbana-Champaign},
    Number     = {UIUCDCS-R-2005-2658},
    Type       = {Tech. Report},
    Month      = {Nov},
    Year       = {2005}
}