Arts & Sciences Events
This calendar presented by
Arts & Sciences
[PAST EVENT] Mingzhou Zhou, Computer Science - Ph.D. defense
June 29, 2015
10:30am - 12:30pm
Abstract:
Modern computing has been reshaped by the emergence of new platforms and growing diverse and large volume of data. It has several prominent features: the agile software development, which encourages early delivery, continuous improvement, rapid and flexible response to change, has been widely adopted by many companies; Responsive analytic applications become crucial for business, however, due to the increasing variety and complexity of inputs to those applications, effective data driven optimization becomes more and more important; With the popularity of mobile devices, which typically have less powerful computation resource, limited memory and storage capacity, many applications for mobile platforms are demanding compact size and high responsiveness, such as game and real-time applications. Feedback-Driven Optimization (FDO) is a technique that has been widely adopted by modern compilers. By allowing the compiler to gather profiles of a program?s dynamic behaviors and perform optimization based on the profiles, it often substantially enhances the quality of generated executable. However, an important open question is how to advance FDO to adapt to the arising trends in modern computing. We aim to answer the question through a systematic exploration from three key aspects: How to reduce profiling overhead in the context of frequent software enhancement and upgrading? How to improve the usefulness of FDO when the profile is gathered through sampling, which is a common technique to alleviate profiling overhead? How to space-efficiently address the input sensitivity problem, which refers to the case when applying optimization based on one profile to a program leads to inferior performance on a different input? We developed several novel techniques, namely profile migration, profile rectification and space-efficient versioning to address those problems. We evaluated our approaches on SPEC benchmarks and the experiment results show our techniques are very promising.
Bio:
Mingzhou Zhou is a PhD candidate in the Department of Computer Science at The College of William and Mary, under the supervision of Dr. Xipeng Shen. His research interest lies in compiler and runtime system techniques, program behavior analysis and optimization, with an emphasis on feedback driven optimization. Prior to The College of William and Mary, Mingzhou received his B.S. from Tsinghua University, Beijing, China in 2009.
Modern computing has been reshaped by the emergence of new platforms and growing diverse and large volume of data. It has several prominent features: the agile software development, which encourages early delivery, continuous improvement, rapid and flexible response to change, has been widely adopted by many companies; Responsive analytic applications become crucial for business, however, due to the increasing variety and complexity of inputs to those applications, effective data driven optimization becomes more and more important; With the popularity of mobile devices, which typically have less powerful computation resource, limited memory and storage capacity, many applications for mobile platforms are demanding compact size and high responsiveness, such as game and real-time applications. Feedback-Driven Optimization (FDO) is a technique that has been widely adopted by modern compilers. By allowing the compiler to gather profiles of a program?s dynamic behaviors and perform optimization based on the profiles, it often substantially enhances the quality of generated executable. However, an important open question is how to advance FDO to adapt to the arising trends in modern computing. We aim to answer the question through a systematic exploration from three key aspects: How to reduce profiling overhead in the context of frequent software enhancement and upgrading? How to improve the usefulness of FDO when the profile is gathered through sampling, which is a common technique to alleviate profiling overhead? How to space-efficiently address the input sensitivity problem, which refers to the case when applying optimization based on one profile to a program leads to inferior performance on a different input? We developed several novel techniques, namely profile migration, profile rectification and space-efficient versioning to address those problems. We evaluated our approaches on SPEC benchmarks and the experiment results show our techniques are very promising.
Bio:
Mingzhou Zhou is a PhD candidate in the Department of Computer Science at The College of William and Mary, under the supervision of Dr. Xipeng Shen. His research interest lies in compiler and runtime system techniques, program behavior analysis and optimization, with an emphasis on feedback driven optimization. Prior to The College of William and Mary, Mingzhou received his B.S. from Tsinghua University, Beijing, China in 2009.
Contact
vlthompsondopp@wm.edu