[CS686A] Evaluating branch correlations

Done as a part of the course - Data driven program analysis.
The aim of this assignment is to evaluate correlations between all the branches in a given C program. A strong correlation between two branches (b1, b2) is directly proportional to the probablity of both the branches behaving similarly.

Project Report

Done in collaboration by -

• Jaivardhan Kapoor (150300)
• Shivansh Rai (14658)

Dependencies

The following package is required for generating the CSV file -

• csvtool : The use of this package has been done to take transposition of the generated CSV file.

For running the R file, the following packages are required -

• littler
• r-cran-plyr

Instructions

The shared library path needs to be appropriately updated -

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$rose_dir/lib

NOTE 1: $rose_dir should be updated with path to the local Rose installation directory.
NOTE 2: Intances of ~/rose_install in Makefile needs to be updated with path to the local Rose installation directory.

For generating the CSV file, execute -

make 				# Compilation step
make run 			# Generates the CSV file

This will generate the file output.csv.

To produce the correlation matrix between branches, execute -

make generate_correlations

A sample of the generated output is available at correlations_demo.txt.

The testcase files are placed inside the directory testcases. In case adding new testcase files for instrumentation, their names in the Makefile have to be updated appropriately (in the last line).