sign up schedule: http://www.cs.utexas.edu/departm

ent/webevent/utcs/events/cgi/show_schedule.cgi?person=GianfrancoBilardi

Colloquium

Speaker Name: Gianfranco Bilardi

Speaker Affi

liation: University of Padova & IBM Research

Date: December 11 20

06

Start Time: 11:00a.m.

Location: ACE 6.304

Host:

Keshav Pingali

Talk Title: How Well Can Physical Machines Approxima

te the Ideal RAM?

Talk Abstract:
It is widely agreed that machi

nes based on sequential programming
are increasingly inadequate to full

y exploit the potential
of present and forthcoming technologies. Howev

er a rigorous
quantitative assessment of the inherent limitations of t

he
sequential approach does not seem to be available. In this
cont

ext we take a look at two interrelated
issues:

(a) How well can

a physical machine where information has
bounded density and speed ap

proximate the performance of
the ideal Random Access Machine?

(b

) How well can the properties of an algorithm that allow
for fast execu

tion be encoded in a sequential program in
ways that a machine can aut

omatically exploit?

We will discuss some results shedding light on t

he above
questions based on novel memory and processor organizations <

br>and related algorithm optimization techniques.