We present the first system for estimating and using datadependent expression execution times in a language with first-class procedures and imperative constructs. Thepresence of first-class procedures and imperative constructs makes cost estimation a global problem that can benefit from type information. We estimate expression costs with the aid of an algebraic type reconstruction system that assigns every procedure atype that includes a static dependent cost. A static dependent cost describes the execution time of a procedure in terms of its inputs. In particular, a procedure’s static dependent cost can depend on the size of input data structures and the cost of input first-class procedures. Our cost system produces symbolic cost expressions that contain free variables describing the size and cost of the procedure’s inputs. At run-time, a cost estimate is dynamically computed from the statically determined cost expression and run-time cost and size information. We present experimental results that validate our cost system onthreecompilers and architectures. We experimentally demonstrate the utility of cost estimates in making dynamic parallelization decisions. In our experience, dynamic parallelization meets or exceeds the parallel performance of any fixed number of processors. 1

Dynamic granularity estimation is a new technique for automatically identifying expressions in functional languages for parallel evaluation. Expressions with little computation relative to thread-creation costs should evaluate sequentially for maximum performance. Static identification of such threads is however difficult. Therefore, dynamic granularity estimation has compile-time and run-time components: Abstract interpretation statically identifies functions whose complexity depends on data structure sizes; the run-time system maintains approximations to these sizes. Compiler-inserted checks consult this size information to make thread creation decisions dynamically. We describe dynamic granularity estimation for a list-based functional language. Extension to general recursive data structures and imperative operations is possible. Performance measurements of dynamic granularity estimation in a parallel ML implementation on a shared-memory machine demonstrate the possibility of large...

Parallel programs are more difficult to debug than sequential programs because of their increase in the complexity of interaction. Debugging is complicated both by the properties of multiprocessors in general, and by the specifics of particular multiprocessors, such as their parallelism granularity, interprocessor communication time lags, etc. Additional effort is necessary to assist users both in debugging their code and in helping them to understand it. Previous efforts have focused on either multiprocessor program debugging or program understanding; it is necessary to unify these two tasks. The tasks of program understanding and debugging are similar. By dealing with both tasks through the same program, one can create a flexible program usable both to increase program understanding and debug code. This necessarily would be a event-based tool, operating both during program execution, and during traditional debugging conditions. If successful, this would greatly contribute to increasi...

ogy (MIT), le Professeur David K. Gifford, mon co-locataire de bureau transatlantique Mark A. Sheldon, Jim O'Toole, Franklyn Turbak, Michael "Ziggy" Blair, John Lucassen et Brian Reistad. ffl Au Laboratoire de Recherche en Informatique (LRI) de Paris XI, les Professeurs Jean-Pierre Jouannaud qui m'a invit'e `a pr'esenter ma th`ese d'habilitation `a Paris XI et Jean-Pierre Sansonnet. ffl Au Centre de Recherche Bull, mon co-auteur "industriel" Babak Dehbonei, G'erard Memmi et Nadia Tawbi. ffl A l'Etablissement Technique Central de l'Armement (ETCA), Serge Petiton et Philippe Clermont, maintenant `a HyParTech. ffl Au Programme de Recherches Coordonn'ees C3, Patrice Quinton, pour son financement. ffl A la Direction Recherche Et Technique (DRET), Philippe Sarrazin. De plus, Fran¸cois Irigoin, Corinne Ancourt, Philippe Bazet, Fran¸cois Masdupuy, et Babak Dehbonei ont particip'e `a la relecture de ce manuscr

pour obtenir le titre de DOCTEUR DE L'UNIVERSITE PARIS VI

pour obtenir le dipl^ome d'Habilitation a Diriger des Recherches en Sciences de l'Universite de Paris XI. Pour completement apprecier le contenu de ces recherches, le lecteur pourra se reporter avec pro t au texte complet des articles donnes en annexe. 1 Remerciements L'ensemble de ces travaux n'aurait pu ^etre ce qu'il est sans le concours de nombreux chercheurs et amis qu'il m'est un plaisir de remercier ici: Au Centre de Recherche en Informatique (CRI) de l'Ecole des Mines de Paris, mon colocataire