Personal tools
You are here: Home Events ICSA/NAIS/EPCC COLLOQUIUM TALK

ICSA/NAIS/EPCC COLLOQUIUM TALK

— filed under:

Talk and presentation by Grigori Fursin - INRIA, France Title: Collaborative application characterization, optimization and run-time adaptation

What
  • Colloquium Series
When Mar 09, 2012
from 11:10 AM to 12:15 PM
Where Appleton Tower. Room 2.12
Add event to calendar vCal
iCal

****THIS IS A JOINT ICSA/ NAIS/ EPCC TALK****

Abstract:

Designing 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 tuning choices combined with complex interactions between all system components.

 

I will present a possible solution to this fundamental problem based on collective participation of users combined with empirical tuning, statistical analysis, machine learning and run-time adaptation. I will describe public Collective Tuning Portal, Infrastructure and Repository (cTuning) for collaborative application and architecture characterization and optimization. With continuously increasing and systematized design and optimization knowledge, users should be able to quickly obtain advices about possible solutions to effectively balance performance, power consumption, code size, compilation time and other important characteristics of their computer systems depending on usage scenarios.

 

Bio:

 

Grigori Fursin obtained BS in physics and electronics from MIPT (Russia), 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 while serving as the project leader and the head of application characterization and optimization group where he prepared research directions on adaptive computer systems for the exascale era.

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 one of the leaders of the MILEPOST project (2006-2009) developing the first publicly available machine learning compiler. Grigori also a founder of the collaborative tuning portal (cTuning.org) with the repository and open source tools to discover, share and systematize knowledge about program and architecture characteristics and optimizations.

Grigori continuously transfers his technology to industry and has been collaborating with multiple companies and universities including IBM, CAPS Entreprise, Intel, Google and others.

 

Document Actions