R. Jones and R. Lins. Garbage collection. Algorithms for automatic dynamic memory management. John Wiley and Sons, 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....400MHz processors for our experiments) This paper is about solving this problem and making conservative collectors more widely useful. Of course, realtime or incremental collections [2, 11, 27] have been extensively studied in the past, and the basic ideas to reduce pause time is known. See [19, 26] for good surveys. The basic idea is to mark objects incrementally, maintaining some invariants with appropriate read write barriers so as not to lose any live object. Most previous work, however, assumes the compiler is able to insert fine grain read write barriers as necessary, making them less ....

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. Wiley & Sones, 1996. ISBN 0-471-94148-4.

....the idea of recording history information by using a heap safety automaton. Our free property falls in the compile time garbage collection research domain, where techniques are developed to identify and recycle garbage memory cells at compiletime. Most work has been done for functional languages [5,17,13,15,19]. In this paper, we show a free analysis, which handles a language with destructive updates, that may reclaim an object still reachable in the heap, but not needed further in the run. Escape analysis (e.g. 7] which allows stack allocating heap objects, has been recently applied to Java. In ....

R. Jones. Garbage Collection. Algorithms for Automatic Dynamic Memory Management. John Wiley and Sons, 1999.

....a set of roots. The roots are locations holding addresses of heap objects; generally, roots are processor registers, program stacks and global variables. An object in the heap is locally reachable if its address is held in a root or if its address is held in another locally reachable heap object [14]. The communication library and its host pointers o er the means to refer to objects in remote processes. In this context, the purpose of a distributed garbage collector is to collect objects that are no longer globally reachable. An object o is globally reachable if one of the following ....

....to mutator messages remains high and constant to 66 . On the other hand, the sorted queue strategy generates less messages and only has a fourfold increase: the same ratio diminishes from 9 to 7 . 7. Related Work Literature on distributed garbage collection is abundant; we refer to Jones book [14] for a complete chapter on the subject and the garbage collection home page [13] with more than 1600 references. Section 3 gave an overview and a discussion of the current work on distributed reference counting. Two di erent approaches however deserve a further comparison. Shapiro, Dickman and ....

Richard Jones and Rafael Lins. Garbage Collection. Algorithms for Automatic Dynamic Memory Management. Wiley, 1996.

....be removed by the garbage collector since their counters are zero. However, reference counting methods do not always nd all garbage; they do not work well in cyclic structures since cyclic garbage are not identi ed. Yet, reference counting is cheap to be implemented in distributed environments [17] therefore justifying the option of implementing a scheme in two phases. The phases are implemented in such a way that the reference counting runs more frequently and ideally Location B Location A UTS p2 ts9 1 p3 ts5 ts2 ts3 ts6 ts4 3 1 1 2 2 1 1 2 ts7 1 2 0 p6 ts8 1 UTS p7 ....

....because it only happens once per process; it adds a startup time which does not have in uence on the process total running time. The average overhead observed in Figure 10 was around 5 which is bellow what Jones and Lins have argued as acceptable in terms of overhead due to garbage collection [17]. Yet, this is the worst case of the overhead observed because the system is not bene ting from any improvement done by the garbage collector. It is natural to believe that if a system generates an excessive amount of garbage, the execution of the garbage collector can improve the performance of ....

R. Jones and R. Lins. Garbage Collection | Algorithms for Automatic Dynamic Memory Management. Wiley, 1996.

....passive 2 NO YES YES THE 2nd PHASE . MARKED IN OBJECT. WAS . MARKED IN THE 1st PHASE NO PASSIVE OBJECT ACTIVE ACTIVE OBJECT OBJECT OBJECT GARBAGE Figure 1: Table showing the four possible situations of a cell after the two marking phases (based on a similar table shown in [7]) if there is no agent directly accessing it but it is linked indirectly to one or more processes, by having its handle in UTS for instance. Having said this, in order to do garbage collection in Ligia, one has to show how to build a graph with the information about the usefulness of Ligia ....

....tuples of size 5 n where n is the average size of an object of one of the classes representing some primitive types (Integer, Float and Double) The results are depicted in Figure 7(b) The overhead added by the tuple analyser is 4. 3 of the total time in the case of Figure 7(b) Jones et al. [7] argue that an acceptable garbage collector should impose at most 20 overhead in the system and the experiments shown in Figure 7 shown a total overhead which stays below this number. The 7 experiments above were also done by having the Ligia server placed on a Pentium 166MHz and on a Silicon ....

R. Jones and R. Lins. Garbage Collection | Algorithms for Automatic Dynamic Memory Management. Wiley, 1996.

....in Figure 5(a) Note that the overhead is small and becomes insigni cant if the bene ts brought about by the garbage collector are taken into account. The average overhead was 13.19 . The case of tuple monitoring is depicted in Figure 5(b) The overhead was 4.3 of the total time. Jones et al. [7] argue that an acceptable garbage collector should impose at most 20 overhead in the system and the experiments shown in Figure 5 shown a total overhead which stays below this number. 8 Conclusion This paper has demonstrated how to make use of tuple monitoring and process registration in the ....

R. Jones and R. Lins. Garbage Collection | Algorithms for Automatic Dynamic Memory Management. Wiley, 1996.

....to the bridge a bridge becomes wider or narrower but is never duplicated. 5 Collecting the Garbage An object in Linda is considered garbage if there is no experiment which can be made to detect this object by the outside world (similarly to what is de ned for Actors [1] by Jones and Lins [7]) The UTS and I O nodes of the distributed graphs represent the boundary to the outside world. This is the reason, then, for having them as roots of a tracing based algorithm. The algorithm proposed in this paper is divided into three main phases. There is no particular order of execution except ....

....collection in this phase is done using a conservative approach. The algorithm aims to avoid false reclaims of live objects. In a mark and sweep algorithm a live object would only be falsely garbage collected if this object becomes invisible to the collector but is still needed by the mutator [7]. In a mark and sweep algorithm visited objects are marked as black. Blacked objects are not revisited. If in the course of the marking phase, the algorithm reaches a black object it does not further explore that path. Therefore if a black object has a reference to a white object (due to changes ....

R. Jones and R. Lins. Garbage Collection | Algorithms for Automatic Dynamic Memory Management. Wiley, 1996.

....management should not be left to be dealt with by users. In distributed environments (including Linda) users are unable to know whether a memory cell is being used by others in the system. As a matter of fact, if the task of reclaiming cell was left to users, the creation of dangling references [JL96] would be common. Garbage collection is the process of searching and reclaiming unused memory objects automatically. These two operations (search and reclaim) are often executed in separate phases. The search phase can be considered the core of garbage collection since there is nothing special ....

....not go into the details of these searching algorithms for it should be clear that our problem is to maintain the necessary information to be used by a searching algorithm and not to choose a suitable searching algorithm for Linda. For good surveys on garbage collection see [Coh81, PS94, Wil94, JL96] 4 Supporting Garbage Collection Within Linda The goal of this section is to describe how we intend to maintain in Linda the information required by any garbage collector. Existing Linda implementations do not provide any information that can be directly used by a garbage collector. Thus, we ....

Richard Jones and Rafael Lins. Garbage Collection | Algorithms for Automatic Dynamic Memory Management. Wiley, 1996.

....for collecting distributed cycles of garbage in a message passing system; the other is an algorithm for DGC in a system with caching and or replication. Both have been implemented in real systems and are in actual use. This article assumes some familiarity with centralised GC, now a mature area [11, 27]. We use the following vocabulary. Objects connected by references form a graph; a remote reference may point into another process. An application or mutator enters the graph via roots, allocates objects, and performs assignment of reference variables. The system or collector reclaims garbage, ....

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. Wiley, Chichester (GB), 1996. ISBN 0-471-941484.

....This is because Larchant applies the so called Union Rule, whereby an object is garbage only if it is unreachable in the union of all replicas [9] 3.1 Reference listing and tracing Larchant runs a tracing GC algorithm on every site. The trace can use either mark and sweep or copy collection [12, 25]. It collects each cluster independently from others; furthermore, each replica of a cluster is traced independently of other replicas. The root of a trace is the set of scions along with PerDiS root objects. Since we are collecting the persistent store (not an application s private replica) we ....

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. Wiley, Chichester (GB), 1996. ISBN 0-471-94148-4.

....and therefore waitfree redistribution requires the strength of consensus variables. It seemed to us that a mark and sweep garbage collection algorithm as in [Jon92] is not likely to yield sufficient performance. We therefore decided to aim at garbage collection by means of reference counting, cf. [JoL96]. By our simplifying assumptions, the usual objection that reference counting does not discern cyclic structures, does not apply. 2 THE INTERFACE 3 Algorithms in which concurrent processes manipulate a shared pointer structure are error prone. We therefore provide a proof of the algorithm by ....

Jones, R., Lins, R.: Garbage Collection, algorithms for automatic dynamic memory management. Wiley, 1996.

....when handling so large heaps, and this is probably also why the performance of reference counting was gradually getting worse as heap became larger in Bin tree and Puzzle. 5 Related Work A number of algorithms for reference counting and global mark and sweep have been proposed in the literature[9, 13]. Performance studies are rare, presumably because most of them are not implemented. Existing performance studies [10, 14, 16, 19] only show performance of either a global mark and sweep or a reference counting and do not compare them. Most of previous works about distributed garbage collection ....

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. Joen Wiley & Sons, 1996.

....is the broad problem of recovering unused space from the heap without explicit instructions from the programmer. It includes both reference counting, garbage collection, and hybrids of the two, like that reported here. Although the second term is often used to include all of storage management [7, 25], we choose the original taxonomy [26, p. 412] because we find it more descriptive: reference counting is a better analog of curbside recycling than of trash collection. Because of the central role played by heap management, it has now become a problem worthy of hardware support. Indeed, others ....

R. Jones & R. Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management, New York, John Wiley & Sons (1996).

....They typically consist of local collections and a global collection and local collections are typically scheduled independently. On distributedmemory machines, global collection schemes are roughly classified into two categories, which are reference counting schemes and distributed marking schemes [1, 39, 56]. Reference counting schemes keep track of how many references exist for each object and delete objects to which there are no references. Distributed marking schemes are natural extensions to local marking collectors. They traverse the entire object graph, which spans multiple processors, ....

....The most straightforward approach would be to make an independent local collector interruptible. An independent collection would periodically poll the network and schedule incoming messages, even in the middle of a collection. This is just an adaptation of incremental collection techniques [39]. Expenses include additional memory overhead, polling overhead, and implementation complexity. The viability of this approach will depend on memory requirements of the application. If each processor has plenty of available memory, additional memory overhead caused by incremental collection will ....

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. John Wiley & Sons, 1996.

....They typically consist of local collections and a global collection and local collections are typically scheduled independently. On distributed memory machines, global collection schemes are roughly classified into two categories, reference counting schemes and distributedmarking schemes [1, 17, 27]. Reference counts keep track of how many references exist for each object and delete objects to which there are no references. Distributed marking schemes are natural extensions to local marking collectors. They traverse the entire object graph, which spans multiple processors, exchanging mark ....

....exchanging mark messages to trace remote references. 2. 1 Local Collection Reference Counts There have been many proposals that extend reference counts to distributed memory machines [6, 11, 26, 27] Implementations of distributed programming languages typically favor reference counts [7, 17] and some parallel languages also adopt them [16] Compared to distributed marking schemes, reference counts seem to have many advantages; for example, processors do not have to be synchronized and only processors that have ever used a reference need to be involved in reclaiming a particular ....

[Article contains additional citation context not shown here]

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. John Wiley & Sons, 1996.

....10 98, Vancouver, Canada. eases programming. This idea has been taken on board in many languages, including the recent and widespread language Java. Distributed garbage collection, which extends the idea of automatic memory management to distributed memory, has been the focus of much attention [17, 30]. Distributed reference counting has been a popular implementation technique of distributed garbage collection because it is simple to implement and can be nicely integrated with sequential garbage collectors [2, 24, 28, 39] Its variant, reference listing , 4, 30, 34, 36] associates objects ....

Richard Jones and Rafael Lins. Garbage Collection. Algorithms for Automatic Dynamic Memory Management. Wiley, 1996.

....This is because Larchant applies the so called Union Rule, whereby an object is garbage only if it is unreachable in the union of all replicas [9] 3.1 Reference listing and tracing Larchant runs a tracing GC algorithm on every site. The trace can use either mark and sweep or copy collection [12, 25]. It collects each cluster independently from others; furthermore, each replica of a cluster is traced independently of other replicas. The root of a trace is the set of scions along with PerDiS root objects. Since we are collecting the persistent store (not an application s private replica) we ....

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. Wiley, Chichester (GB), 1996. ISBN 0-471-94148-4.

....an overview of our approach. Sections 4 and 5 detail our implementation and present some preliminary performance figures. The final section presents conclusions and further work. 2 Related Work There is an overwhelming number of papers concerning garbage collection: a useful summary is given in [5]. It is important to distinguish between distributed versus concurrent versus parallel garbage collection. As previously mentioned distributed garbage collection addresses issues such as fault tolerance and avoidance of tight coordination. Concurrent garbage collection entails running the 2 Java ....

Richard Jones and Rafael Lins. Algorithms for Automatic Dynamic Memory Management. John Wiley & Sons, 1996.

No context found.

R. Jones and R. Lins. Garbage collection. Algorithms for automatic dynamic memory management. John Wiley and Sons, 1996.

No context found.

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. JOHN WILEY & SONS, NY 10158-0012, USA, 1996.

No context found.

R. Jones, R. Lins, Garbage Collection. Algorithms for Automatic Dynamic Memory Management, John Wiley & Sons Ltd. Bans Lane, Chichester, 1996.

No context found.

R. Jones and R. Lins. Garbage Collection. Algorithms for Automatic Dynamic Memory Management. John Wiley & Sons, Ltd., San Francisco, 1999.

No context found.

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. Wiley, 1996.

No context found.

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. Wiley, 1996.

No context found.

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. Wiley & Sones, 1996.

No context found.

Richard Jones and Rafael Lins. Garbage Collection, Algorithms for Automatic Dynamic Memory Management. Wiley & Sones, 1996.

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