Hi! I am a fourth year PhD student at CMU advised by Professor Brandon Lucia (brandonlucia.com).
I am a member of the ABSTRACT research group (http://abstract.ece.cmu.edu/)

Research Interests

My current research interest is in investigating architectural support for optimizing graph processing applications.

I have been exploring techniques to improve the efficiency of graph processing workloads on modern multi-core processors. Graph processing is an important domain that is hard to optimize because of the irregular memory access pattern and the scale of graph inputs. My research focus is to provide better performance by improving cache locality of graph processing workloads. A fundamental tenet of the research is to exploit the structural properties of input graphs to propose input-specific locality optimization techniques.

My previous projects have been on optimizing cache coherence protocols and exploiting approximate computing to improve scalability of parallel applications.

Publications

"When is Graph Reordering an Optimization? Studying the effect of lightweight graph reordering across applications and input graphs",
Vignesh Balaji, and Brandon Lucia,
IEEE International Symposium on Workload Characterization (IISWC), 2018
(BEST PAPER AWARD)
[preprint] [slides] [github]

"Flexible Support for Fast Parallel Commutative Updates",
Vignesh Balaji, Dhruva Tirumala and Brandon Lucia,
Arxiv 2018
[paper]

"An Architecture and Programming Model for Accelerating Parallel Commutative Computations via Privatization",
Vignesh Balaji, Dhruva Tirumala and Brandon Lucia,
Symposium on Principles and Practice of Parallel Programming (PPoPP), 2017
[poster]

"Intermittent Computing: Challenges and Opportunities",
Brandon Lucia, Vignesh Balaji, Alexei Colin, Kiwan Maeng, and Emily Ruppel,
Summit on Advances in Programming Languages (SNAPL), 2017
[paper]

"Overcoming the Data-flow Limit on Parallelism with Structural Approximation",
Vignesh Balaji, Brandon Lucia, and Radu Marculescu,
Workshop on Approximate Computing (WAX) colocated with ASPLOS 2016
[paper]

Contact