UTCS Colloquium: Hugh Leather/University of Edinburgh: "Machine Learning in Compilers: The Future of Optimisation" TAY 3.128 Wednesday, May 27, 2009 2:00 p.m.

Contact Name: 
Jenna Whitney
Date: 
May 27, 2009 2:00pm - 3:00pm

Type of Talk: UTCS Colloquium

Speaker/Affiliation:&nbs

p; Hugh Leather/University of Edinburgh

Date/Time:  Wednesday

, May 27, 2009  2:00 p.m.

Location:  TAY 3.128

Hos

t:  Martin Burtscher (for Keshav Pingali)

Talk Title:  &q

uot;Machine Learning in Compilers: The Future of Optimisation"

Talk Abstract:

Developing an optimising compiler is a highly skilled
and arduous process and there is inevitably a software delay whenever a ne

wprocessor is designed. It takes several generations of a compiler to start
to effectively exploit the processors'' potential, by which time a new pr

ocessor appears and the process starts again. This never ending game of cat

ch-up means that we rarely fully exploit a shipped processor and it inevita

bly delays time to market. As we move to multi and many core platforms, th

is problem increases.

This talk will look at research undertaken
in the EU MilePost http://www.milepost.eu/ project which uses machine lear

ning to automatically learn how to optimise for a reconfigurable platform.

It will also outline a variety of ongoing work to use such an approach for

compiler/architecture co-design space exploration, trans-architecture lear

ning compilation and multi-core optimisation.

Speaker Bio:

Hug

h Leather is completing a PhD at the University of Edinburgh. He worked pre

viously at Trilogy in Texas as a senior engineer and before that at Microso

ft in Redmond as a developer. He obtained his undergraduate degree also fro

m the University of Edinburgh. His research is in using machine learning to
automatically create optimising compilers