********************************************************************* * Ecole Normale Supe'rieure * * * * Se'minaire * * SEMANTIQUE ET INTERPRETATION ABSTRAITE * * P. Cousot * * * * Vendredi, 14h00--15h30 * * Salle S16, etage -1 * * DI ENS 45 rue d'Ulm 75005 Paris * ********************************************************************* *** Vendredi 25 novembre 2005 *************************************** Olin Shivers (Georgia Tech) Environment analysis with $\Delta$CFA Abstract: The lambda expression provides, as a single linguistic construct, all three fundamental program structures: data structure, environment structure, and control structure. This universality is one of the sources of its power and centrality in programming languages, but it is also the reason why lambda is so difficult to analyse. Because these three kinds of structure come together in lambda, a program analysis that handles lambda must simultaneously handle all three facets of the construct. Thus a control-flow analysis for lambda-based programs must be, at the same time, an environment- and data-flow analysis, and vice versa. While techniques to analyse lambda-based programming languages have been around for quite some time, the weakest facet of these analyses has been the handling of environment structure. For example, the nature of the environment abstractions used in the $k$-CFA hierarchy acts as a fundamental barrier to certain classes of analysis. I will present a new analysis, $\Delta$CFA, which exploits an environment/control abstraction based on a variant of \emph{procedure strings}, a record of procedure entries and exits. The new analysis permits more precise reasoning about environment structure, enabling new classes of optimisation. This work is a joint effort with my student, Matthew Might. ---------------------------------------------------------------------------- Olin Shivers received his undergraduate degree in Computer Science and Mathematics from Yale University; his undergraduate advisor was Alan Perlis. His doctoral work in program analysis was done at Carnegie Mellon, under the supervision of Peter Lee and Allen Newell. Since then, he has held positions at Bell Labs, and the University of Hong Kong and the MIT AI Lab, before joining the faculty at Georgia Tech, where he currently teaches and does research in programming languages. Prof. Shivers has also started several companies; his most recent is Smartleaf, which provides scalable portfolio-analysis technology to the financial industry. ********************************************************************* Pour recevoir l'annonce par courrier electronique: WWW: http://www.di.ens.fr/~cousot/annonceseminaire.shtml *********************************************************************