UTCS Colloquium: Mario Mendez/University of New Mexico: Set Sharing and Resource Analysis of Java Bytecode ACES 2.402 1:30pm Thursday July 3 2008

Contact Name: 
Jenna Whitney
Date: 
Jul 3, 2008 1:30pm - 2:30pm

There is a signup schedule for this event (UT EID required).

Ty

pe of Talk: UTCS Colloquium

Speaker/Affiliation: Mario Mendez/Univ

ersity of New Mexico

Date/Time: Thursday July 3 2008 1:30 p.m.
Location: ACES 2.402

Host: Keshav Pingali

Talk Title:
Set Sharing and Resource Analysis of Java Bytecode

Talk Abstract:Abstract Interpretation-based frameworks usually depend on
at least t

hree building blocks: the compiler that translates the
program to be an

alyzed into a more machine-friendly intermediate representation an abstrac

t domain that captures the property of
interest and a fixpoint algorit

hm that ties them together. In most
frameworks the three components ar

e bundled together thus is
difficult to incorporate new analyses. In m

y talk I will explain our
proposed (Prolog-like) IR and describe an e

fficient fixpoint algorithm
based on that IR. The solution is generic e

nough to be used for the
analysis of Java bytecode and Prolog.

I

n the more specific context of abstractions that formally prove
propert

ies about a Java program I will describe my work in resource
and heap

analysis. Resource analysis refers to properties that are
more high-lev

el that those computed by traditional cost analysis
(i.e. time and mem

ory): bytes sent or received by an application
files left open etc. M

y recent work focuses in heap analysis; more
precisely I designed an

abstract domain (set sharing) that captures
which program variables def

initely do not reach a same object in
memory. While the domain was show

n to more precise than a
existing alternative the exponential complexi

ty of some of its
operations raised questions about its scalability. Ho

wever we later
discovered that set sharing can be efficiently implemen

ted by using
Zero-supressed BDDs (ZBDDs) which drastically reduce both

memory
consumption and running time.

Speaker Bio:
Mario Mendez

-Lojo is a Ph.D. candidate in the Computer Science
Department at the Un

iversity of New Mexico with a focus in the area
of Static Analysis. He
received an M.S. in Computer Science from the
University of New Mexico

in 2006 and a B.S. in Computer Science
from the University of La Coruna
(Spain) in 2000. His latests research
interests are in how to design s

calable analyses that can be incorporated
into a compiler more precise

alternatives to the standard Abstract
Interpretation model and the crea

tion of common analysis tools that
can be applied to a variety of langua

ges.

A more detailed CV can be found at http://cs.unm.edu/%7Emario.