RESEARCH
Research Interests
Parallel Computing
Software Transactional Memory (STM)
General Purpose computing on Graphics Processing Units (GPGPU) Computing
Distributed Computing
High Performance Computing (HPC)
Convergence of HPC and Artificial Intelligence (AI)
Blockchain
Quantum Computing
PhD Thesis Title: "Concurrent & Parallel Systems: Robust Enhancement of Performance"
Advisors: Prof. R. K. Shyamasundar, Indian Institute of Technology Bombay, and
Dr. A. K. Bhattacharjee, Bhabha Atomic Research Centre, Mumbai.
Summary
In this work, we provided solutions to problems pertaining to the following challenges: Performance, Productivity and Robustness, for concurrent and parallel systems. In the context of concurrent systems, an efficient Software Transactional Memory (STM) was proposed, which provides a good way to address both application performance and productivity. In the context of parallel systems, powerlist data structure was used to provide a high-level framework for performance and productivity. Powerlist is a robust data-structure that allows us to utilize recursion and parallelism in an unified manner. Further, to address robustness concerns, the variety of deadlocks possible in GPU applications was discussed and two lock-based deadlock-free synchronization mechanisms were presented: Prev-Deadlockfree and Par-Deadlockfree, for GPU architectures that overcome these issues.