Swarnendu BiswasComputer Science and Engineering
Ohio State University
Room#395, Department of CSE,
Ohio State University,
Dreese Labs, 2015 Neil Avenue,
Columbus, Ohio 43210-1277.
Email: biswass [AT] cse.ohio-state.edu
My areas of interest are Programming Languages, Compilers and Runtime Systems, Computer Architecture, Software Engineering, Embedded and Real-Time systems.
I will be joining the ISS group with Prof. Keshav Pingali at UT Austin for my postdoc. I worked in the PLaSS group under the supervision of Dr. Michael Bond for my PhD. My PhD research focused on analyzing concurrency correctness issues in multi-threaded programs and memory models. For example, I have worked on devising dynamic program analyses to check for atomicity violations and data races, which are the two most common source of concurrency errors. The challenges in developing such analyses lie in balancing precision, correctness, coverage, scalability, and efficiency for practical usage on different systems. My work on concurrency is mostly from the perspective of Java programs and is developed in a Java research virtual machine. My work on memory models aims to strengthen existing programming language memory models by providing strong semantic guarantees even for programs with data races. I have worked on developing software-only solutions which detect violations of region serializability and can run on commodity hardware. I have also explored more ambitious hardware-only techniques to ensure region serializability with modest extensions to the hardware. My simulated hardware-based work proposes to replace the existing cache coherence protocols with novel techniques to ensure region serializability and snapshot isolation.
I interned at Google Inc., Mountain View, during the summer of 2013 with the Java Platform Team. The goal of my project was to have an efficient data race detector targeted towards production systems. My work involved implementing a prototype of a low-overhead data race detector for Java programs in the Hotspot VM in OpenJDK. We used hardware breakpoints and debug registers available on commodity hardware for tracking conflicting accesses to memory locations.
My MS research focused on developing improved
techniques for automating regression testing of embedded programs. The work had two parts: developing an automated
technique for selecting regression test cases, and then optimizing the set of selected test cases further.
Winter Quarter 2012:
CSE 202: Introduction to Programming and Algorithms for Engineers and Scientists U 4
About Me | more
During my MS, I worked as a research assistant at IIT Kharagpur (September 2008 to August 2011). The
project was sponsored by General Motors India. My work was on developing automated and efficient regression test
selection approaches for automotive software.
I worked as a software developer at Wipro Technologies for three years (August 2005 to August 2008). I was involved in software development in the Unified Communications and Automotive Software domains.