Towards Language-Independent Region Inference

 Region-based memory management is an alternative to garbage
 collection for programming languages with implicit memory
 management. A compile-time analysis, _region inference_
 automatically inserts explicit deallocation instructions at points
 in the program where a _region type system_ can be proved it safe to
 reuse memory. Originally proposed for ML-like languages by Tofte and
 Talpin in 1994, the region-based run-time model and the idea of
 proving deallocations safe at compile time has spread to several
 other languages, but still the region inference that invents them
 automatically has obly been formulated for ML.  This talk gives a
 general high-level introduction to region-based memory management,
 and describes my ongoing work on reformulating a form of region
 inference to be applicable to a wide variety of languages.

 Henning Makholm received a M.Sc. in computer science from the
 University of Copenhagen in 2000, and is currently finishing a Ph.D.
 at the same place. He has worked with type-based program analyses
 for a variety of purposes, including region-based memory management
 and supporting analyses for partial evaluation.

Bis bald,

Andreas Zeller

--
Andreas Zeller    Universitaet des Saarlandes, Saarbruecken, Germany
                 http://www.st.cs.uni-sb.de/~zeller/
Am Freitag, den 27.06. um 10:00 c.t. (Seminarraum 45/328) spricht