Designing, porting and optimizing applications for rapidly evolving computing systems
is often complex, ad-hoc, repetitive, costly and error prone process due to an enormous
number of available design and optimization choices combined with the complex interactions
between all components.
I will present a possible solution to this fundamental problem based on collective
participation of users combined with empirical tuning and predictive modeling.
I will describe public Collective Tuning Repository and new projects for collaborative
application characterization and optimization. With continuously increasing and systematized
knowledge about behavior of computer systems, users should be able to obtain scientifically
motivated advices about anomalies in the behavior of their applications
and possible solutions to effectively balance performance and power consumption.
I believe that such approach will be vital for developing efficient Exascale computing systems.
Grigori Fursin obtained BS in electronics and MS in computer engineering from MIPT (Russia), PhD in computer science from the University of Edinburgh (UK), and is currently a tenured research scientist at INRIA (France). In 2010-2011, Grigori was on sabbatical helping to establish Intel Exascale Lab in France and serving as the head of application characterization and optimization group.
Grigori's main interests are in code and architecture characterization and auto-tuning for performance, power and other characteristics using empirical, statistical, collective and machine learning techniques. He was the technical leader of the MILEPOST project (2006-2009) developing the first public machine learning based compiler. Grigori also established collaborative portal (cTuning.org) with the repository and open source tools to share and systematize knowledge about program and architecture design and optimization. Grigori is collaborating with multiple companies and universities including Intel, CAPS Entreprise, IBM, Google, University of Edinburgh, UIUC, Paris South University, ICT and others.
Attachment | Size |
---|---|
Presentation_Fursin_UCAR_SEA2012.pdf | 4.09 MB |