Non-compacting Memory Allocation and Real-time Garbage Collection Dissertation Proposal

Garbage collection is the automatic reclamation of computer storage Knu73, Coh81, Wil92, Wil95]. While in many systems, programmers must explicitly reclaim heap memory at some point in their program by using a \free" or \dispose" statement, garbage collected systems free the programmer from this burden. In spite of its obvious attractiveness for many applications, garbage collection for real-time programs is not popular. This is largely due to the perceived cost and disruptiveness of garbage collection in general, and of incremental garbage collection in particular. Most existing \real-time" garbage collectors are not in fact usefully real-time, largely due to the use of a read barrier to trigger incremental copying of data structures being traversed by the running application. This may slow down running applications unpredictably, even though individual increments of garbage collection work are small and bounded. We have developed a hard real-time garbage collector which uses a write barrier to only coordinate collection work with modiications of pointers in data structures, therefore making coordination costs cheaper and more predictable. We combine this write barrier approach with implicit non-copying reclamation (\fake copying"), which has most of the advantages of copying collection (notably avoiding both the sweep phase required by mark-sweep collectors, and the touching of garbage objects when reclaiming their space), without the disadvantage of having to actually copy the objects. It has long been believed that fragmentation will be prohibitively high in a non-copying garbage collection algorithm. We address these fragmentation issues with our garbage collector, and compare the fragmentation produced by a number of traditional memory allocation algorithms. In doing this comparison we produce some surprising results demonstrating that methodologies used to study fragmentation , dating back as far as 1961, are fundamentally unsound and biased. We have conducted sound studies that suggest that fragmentation can be kept very low for most real programs using well-known, non-copying policies. Finally, we have extended our garbage collector with generational techniques to attempt to provide excellent average-case eeciency while preserving good soft real-time responsiveness.