I am working with Prof. David I. August in Liberty Research Group on compiler optimization techniques for multi-core architecture. My research interest includes parallel compiler optimization, programming language and artificial intelligence. I am currently working on multithreaded program analysis for optimization

Upcoming Paper Deadlines

• PPoPP Abstract July 10, 2009
• CGO Sep, 2009

Recent Readings

• Foundations of the C++ Concurrency Memory Model, H.J. Boehm, S. Adve
• Pointer Analysis for Multithreaded Program, R. Rugina, M. Rinard
• Context-Sensitive Interprocedural Points-to Analysis in the Presence of Function Pointers, M. Emami, R. Ghyiya, L. Hendren, 1994
• Graph Algorithm (Bedside story book)

Publications

• Revisiting Sequential Programming Model for the Multi-core Era(PDF), M. bridges, N. Vachharajani, Y. Zhang, T. Jablin, D. I. August, IEEE Micro, January 2008 (IEEE Micro's "Top Picks" special issue for papers "most relevant to industry and significant in contribution to the field of computer architecture" in 2007.)
• Revisiting Sequential Programming Model for the Multi-core(PDF), M. bridges, N. Vachharajani, Y. Zhang, T. Jablin, D. I. August, Proceedings of the 40th IEEE/ACM International Symposium on Microarchitecture (MICRO), December 2007
• Runtime Empirical Selection of Loop Schedulers on Hyperthreaded SMPs(PDF), Y. Zhang, M. Voss, Proceedings of the 19th IEEE International Parallel and Distributed Processing Symposium, April 2005
• An Adaptive OpenMP Loop Scheduler for Hyperthreaded SMPs(PDF), Y. Zhang, M. Burcea, V. Cheng, R. Ho, M. Voss, Proceedings of the 17th International Conference for Parallel and Distributed Computing Systems, September 2004

Talks and Presentations

• My talk slides on alias analysis using separation logic in my generals exam (qualification exam in Princeton) can be found here

Research Links

• Alias Analysis Papers
• Multithreade Program Analysis

Crazy Research Ideas

• On demand pointer analysis? We are spending lots of efforts on pointer analysis these days. Context/Flow sensitive analysis is still not practical. Shall we simply abandon the current pointer memory management system or shall we look for on demand pointer analysis? First we figure out which pointers are important then we spend lots of time on that instead of analyzing every pointer extensively.
• How about passing high level type information down to low level pointer analysis to achieve both accurancy and optimal order of optimization?
• Why is there a huge gap between fp and int programs? what are fundamentally preventing general purpose programs to execute as fast as scientific applications?
• Is there a universal way to measure dynamic optimization efficiency? When we cannot find anything statically, we always want to go online. But how do we know whether online optimization is gonna work as we think? How do we know whether the price we pay pays us back? Is there a framework that we can employ to prevent us from doing the wrong thing at the very beginning?
• What are the domain specific language mean? How do we categorize domains and know how to get parallelism?
• Since we have machines with 100,000 processors, why do we want to do speculation? We can just throw two or more versions there and pick the right one later. We dont know how to use those idling 99, 999 processors anyway. This leads to a fundamental question: do we want to do things 100% correctly at the beginning, or do we allow us to make some tolerable mistakes and being able to fix it at a very low cost?