Speaker: Jeremy Singer, University of Glasgow

Title: Controlling Heap Size in Poly/ML

Abstract:

Typical theorem-proving workloads on the Poly/ML [1] runtime system
may execute for several hours, occupying multi-gigabyte heaps.
The heap size may be fixed at execution startup time, or it may be
allowed to vary dynamically. To date, runtime heap size growth
has been implemented using simple hard-coded heuristics.
In this talk, I will argue that a mathematically rigorous approach
to heap sizing, based on control theory, is more appropriate.

1 - http://www.polyml.org/FAQ.html