| P. Wilson and M. Johnstone. Real-Time Non-Copying Garbage Collection. In ACM OOPSLA Workshop on Garbage Collection and Memory Manageeent, September 1993. |
....However, this may be seen as a special case of the region based scheme although objects may move among regions. Under generational collection, objects get grouped within regions according to their age. coordinate the application and the GC we use write barriers instead of read barriers [15]. This decision is motivated because write barriers are more efficient than read barrier, allows cooperation with our region barrier implementation, and the resulting collector can further be easily extended to generational, distributed, and parallel collection. In the context of RTSJ, this ....
P. Wilson and M. Johnstone. Real-Time Non-Copying Garbage Collection. In ACM OOPSLA Workshop on Garbage Collection and Memory Manageeent, September 1993.
....memory management scheme where each running application is assigned a given portion of memory upon execution. A dedicated Gc is then run within each application according to the application s memory usage profile. By default, any real time application will be combined with the Gc proposed in [12]. Finally, a global real time Gc is run for reclaiming the objects created by the environment as well as those shared among applications. In addition, while there exist garbage collection algorithms that offer the necessary features for being compliant with real time constraints, the issue of ....
P.R. Wilson and M.S. Johnstone. Real-Time Non-Copying Garbage Collection. In Workshop on Garbage Collection and Memory Management. OOPSLA, 1993.
....for sweep time reduction: mark sweep using a doubly linked list (which he called the treadmill) where the sweeping is done in constant time. Unfortunately, all objects must be of the same size for it to be usable, so its usefulness in general purpose systems is limited. Wilson and Johnstone [31] extended Baker s algorithm to multiple object sizes using segregated storage. However, unlike Baker s algorithm, where sweeping can be done in constant time, their algorithm required that each dead object be processed before being reused, so sweeping still uses time linear on the heap size ....
Paul R. Wilson and Mark S. Johnstone. Real-time non-copying garbage collection. In Eliot Moss, Paul R. Wilson, and Benjamin Zorn, editors, OOPSLA /ECOOP '93 Workshop on Garbage Collection in Object-Oriented Systems, October 1993.
....applications (though we would restrict our claims to the former, since we currently ignore CPU cycles, cache and paging behaviour) Baker s incremental copying technique [1] might be expected to form the natural basis for a collector with the required behaviour. However, as Wilson and Johnstone [33, 32] indicate, the worst case cost of Baker s read barrier is potentially very high and very unpredictable (they give the example of the cost of traversing a list, which is strongly dependent on whether the list has already been reached and copied by the collector see also [23, 36] Wilson and ....
.... achieve acceptable performance with the same worst case bounds that we require of the allocator, we leave the collection of cycles to further work (the aim of this paper is to present the allocator, not the collector ) Similarly, though there are interesting similarities between our technique and [33], we also leave a direct comparison to further work. In a distributed context, the natural starting point for our garbage collector might be weighted reference counting [4] however, a full implementation (one where cells cannot become unreclaimable due to reference count overflow) suffers from ....
[Article contains additional citation context not shown here]
Paul R. Wilson and Mark S. Johnstone. Real-time non-copying garbage collection. In Eliot Moss, Paul R. Wilson, and Benjamin Zorn, editors, OOPSLA/ECOOP '93 Workshop on Garbage Collection in Object-Oriented Systems, October 1993.
.... such those investigated by [Appel 88, Singhal 92, Moss 92] CX3 could be extended to utilise synchronisation constraints [Decouchant 89, Gunaseelan 91] The GC could also be extended to use generational techniques like generation scavenging [Ungar 84, Ungar 87] and real time GC as suggested in [Wilson 93] In this paper, we presented the design of the CROM system which used a novel approach employing an orthogonal persistence model that was integrated with concurrent threads, nested atomic actions, object swapping, and persistence sensitive garbage collection technology. 9. Acknowledgements ....
Paul R. Wilson and Mark S. Johnstone. Real-time non-copying garbage collection. Position paper for the 1993 OOPSLA Workshop on Memory Management and Garbage Collection, 1993.
....such as control systems demand techniques to avoid memory fragmentation. Using compacting algorithms is one way to solve such problems, while other techniques such as fixed object sizes and pre fragmentation of large objects also can be used to avoid the fragmentation problem [Bak92] Hen93] and [WJ93]. We have chosen to work with a compacting algorithm since it solves a more general and harder problem and also provides a very clean solution to the fragmentation problem. We assume the reader to have a fair understanding of well known GC algorithms and point to the references for details. In ....
P. R. Wilson & M. S. Johnstone. Real-Time Non-Copying Garbage Collection. OOPSLA'93 Workshop on Memory Management and Garbage Collection. Washington DC, Oct 93.
....such as control systems demand techniques to avoid memory fragmentation. Compacting algorithms are one way to solve such problems, while other techniques such as fixed object sizes and pre fragmentation of large objects can be used to avoid the fragmentation problem [Bak92] Hen93] and [WJ93]. We have chosen to work with a compacting algorithm since it provides a very clean solution, and solves a more general problem. For an extensive discussion on incremental GC techniques see [Wil92] 3.1 Development for interactive systems The original batch formulations of the GC algorithms are ....
....processes, since the possible delay of starting a high priority process depends on the time to move a small number of objects. For high priority processes the scheme adds considerable time to the worst case behavior since these operations (allocation, access update pointers) can be very frequent [Wit91, WJ93]. Depending on the timing constraints this might be enough to prohibit use of these operations in a high priority process which can be a rather severe restriction (especially forbidding pointer manipulation) Our approach, making the critical operations very short and of fixed length, has also ....
P. R. Wilson & M. S. Johnstone. Real-Time NonCopying Garbage Collection. OOPSLA'93 Workshop on Memory Management and Garbage Collection. Washington DC, Oct 93.
.... those which run on specialized hardware [14] To allow programmers to use RSK for code which may use cyclic data structures, and to simplify the interpreter, we are implementing a real time garbage collector based on the writebarrier strategy used in Wilson and Johnstone s real time collector [15]. 5 Conclusion We have described dynamically reconfigurable subsystems for sensor based control of robot systems, and presented the Robot Scheme Kernel (RSK) an embedded Scheme interpreter designed for high level management of these subsystems. To allow RSK to evaluate Scheme expressions which ....
P. Wilson and M. Johnstone, "Real-time non-copying garbage collection," in ACM OOPSLA Workshop on Memory Management and Garbage Collection, (Washington D.C.), ACM, September 1993.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC