APPEL, A. W., ELLIS, J. R., AND LI, K. Real-time concurrent collection on stock multiprocessors. In Proceedings of the SIGPLAN'88 Conference on Programming Language Design and Implementation (Atlanta, Georgia, June 1988). SIGPLAN Notices, 23, 7 (July), 11--20.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....immortal, or immutable objects) 8, 14] although we do not explore this type of segregation in this paper. Any references into an increment must be tracked if that increment is to be collected independently. Pointer tracking may use remembered sets [34] card marking [39] hardware support [4, 9, 16, 26], or a combination of techniques. Card tables [39] are a common alternative to the remsets we use in Beltway. Card tables trade a fast write barrier (typically two or three machine instructions) for increased work scanning at collection time. A marked entry in the card table means that one or more ....

Andrew W. Appel, John R. Ellis, and Kai Li. Real-time concurrent collection on stock multiprocessors. ACM SIGPLAN Notices, 23(7):11--20, 1988.

....into to space before returning the pointer to the mutator. There is therefore a tight coupling between the operations of the mutator and the scheduling of operations by the collector. Examples are Baker s copying algorithm [5] which uses an explicit read barrier, and the Appel Ellis Li collector [2], which uses virtual memory protection. Both of these collectors have the property that mutator utilization is very poor right after the collector starts, when the fault rate is high. An alternative is to use a replicating collector which maintains a from space invariant, and to perform mutator ....

....his seminal paper on real time garbage collection by stating that a real time list processing system is one in which the time required by the elementary list operations . is bounded by a small constant. This approach has been the basis for most of the later work on real time collection [2, 6, 10, 11, 17, 19, 24]. However, this is implicitly a work based approach, and as we have seen in Sections 4 and 7, at the small time intervals that are typically of interest in real time systems, work based collectors may be subject to very poor utilization. Baker attempts to finesse this problem by interleaving the ....

[Article contains additional citation context not shown here]

APPEL, A. W., ELLIS, J. R., AND LI, K. Real-time concurrent collection on stock multiprocessors. In Proceedings of the SIGPLAN'88 Conference on Programming Language Design and Implementation (Atlanta, Georgia, June 1988). SIGPLAN Notices, 23, 7 (July), 11--20.

.... Steele [69] 75 barrier, object locks MSC Kung, Song [47] barrier Mark Sweep Ungar [72] barrier Gen Copy Liebermann, Hewitt [50] 83 ffl barrier, write barrier Gen Copy Halstead [36] ffl ffl barrier, object locks Copy Crammond [19] Appel, Ellis, Li [2] barrier, VM support Copy Yuasa [78] ffl ffl barrier, object handles Mark Sweep (Remap) DeTreville [20] Ref. Count Lins [52] Mark Sweep, Ref. Count Sharma [67] barrier Copy Herlihy, Moss [38] ffl ffl barrier Copy Engelstad al. 28] ms read barrier ....

....the functions that the mutator calls directly and Figure 4.7 shows the functions that form the main body of the collector. Together with the code in Figure 4.5 and some helper functions from Appendix A, these form the entire algorithm for the collector. 1 shared int Which = 1; 2 shared int Prev[2] = 0,0; 3 shared int Cur[2] 0,0; 4 shared int Wait[2] 0,0; 5 void EnterRoom(int i) f 6 oldWait = FetchAndAdd( Wait[i] 1) 1; 7 while (oldWait Cur[i] 8 if (CompareAndSwap( Which, 1,i) 1) f 9 Cur[i] Wait[i] 13 int ExitRoom(finalizer t f)f 14 oldPrev = ....

[Article contains additional citation context not shown here]

Andrew W. Appel, John R. Ellis, and Kai Li. Real-time concurrent collection on stock multiprocessors. ACM SIGPLAN Notices, 23(7):11-- 20, 1988.

.... While numerous concurrent, multiprocessor garbage collectors for general purpose programming languages have been described in the literature [12, 14, 16, 18, 19, 21, 22, 24, 29, 30] the number that have been implemented is quite small and of these, only a few actually run on a multiprocessor [2, 12, 18, 16, 17, 27]. DeTreville s work on garbage collectors for Modula 2 on the DEC Firefly workstation [12] is the only comparative evaluation of multiprocessor garbage collection techniques. His algorithm is based on Rovner s reference counting collector [29] backed by a concurrent tracing collector for cyclic ....

....of Domani et al. [17] This is a generational collector based on the work of Doligez et al. [16, 15] for which generations were shown to sometimes provide significant improvements in throughput. No response time measurements were provided. The other implemented concurrent multiprocessor collectors [2, 18, 16, 27] are all tracing based algorithms for concurrent variants of ML, and generally have significantly longer maximum pause times than our collector. In addition, ML produces large amounts of immutable data, thereby simplifying the collection process. The collector of Huelsbergen and Larus [18] for ML ....

APPEL, A. W., ELLIS, J. R., AND LI, K. Real-time concurrent collection on stock multiprocessors. In Proceedings of the SIGPLAN Conference on Programming Language Design and Implementation (Atlanta, Georgia, June 1988), ACM Press, New York, New York. SIGPLAN Notices, 23, 7 (July), 11--20.

.... Collection While numerous concurrent, multiprocessor collectors for general purpose programming languages have been described in the literature [8, 10, 11, 14, 15, 18, 19, 21, 26, 27] the number that have been implemented is quite small and of these, only a few actually run on a multiprocessor [2, 8, 14, 11, 13, 24]. DeTreville s work on garbage collectors for Modula 2 on the DEC Firefly workstation [8] is the only comparative evaluation of multiprocessor garbage collection techniques. His algorithm is based on Rovner s reference counting collector [26] backed by a concurrent tracing collector for cyclic ....

....for Java that we know of is the work of Domani et al. [13, 12] This is a generational collector based on the work of Doligez et al. [11] for which generations were shown to sometimes provide significant improvements in throughput. The other implemented concurrent multiprocessor collectors [2, 14, 11, 24] are all tracing based algorithms for concurrent variants of ML, and generally have significantly longer maximum pause times than our collector. In addition, ML produces large amounts of immutable data, thereby simplifying the collection process. The garbage collector of Huelsbergen and Larus ....

APPEL, A. W., ELLIS, J. R., AND LI, K. Real-time concurrent collection on stock multiprocessors. In Proceedings of the SIGPLAN Conference on Programming Language Design and Implementation (Atlanta, Georgia, June 1988). SIGPLAN Notices, 23, 7 (July), 11--20.

....allocated in the region being collected. There are many previous concurrent and or incremental copying collectors, most of which use a read barrier to insure that mutators see only the N copies of copied objects. Examples include the algorithms of Baker [3] Brooks [5] and Appel, Ellis, and Li [1], in addition to Nettles and O Toole. Because of their N space only invariant, these algorithms cannot flip incrementally without adding a read barrier. There are some refinements of replicating collection, such as special treatment of immutable data [10] important for ML, but not as helpful for ....

A. W. Appel, J. R. Ellis, and K. Li. Realtime concurrent collection on stock multiprocessors. In Proceedings of the ACM SIGPLAN '88 Conference on Programming Language Design and Implementation, pages 11--20, Atlanta, Georgia, June

....run on a single thread after all program threads have all been stopped (the so called stop the world concept) This causes bad processor utilization, and hinders scalability. In order to make better use of a multiprocessor, concurrent collectors have been presented and studied (see for example, [5, 35, 36, 16, 2, 12, 13, 7, 18, 30, 17]) A concurrent collector is a collector that does most of its collection work concurrently with the program without stopping the program threads. Most of the concurrent collectors need to stop all program threads at some point during the collection, in order to initiate and or nish the ....

Andrew W. Appel, John R. Ellis, and Kai Li. Real-time concurrent collection on stock multiprocessors. ACM SIGPLAN Notices, 23(7):11-20, 1988.

....are given in figure 3. Note that this language bears some similarities 7 Although some recent concurrent collectors (e.g. WGH92] do not lock when copying, they require additional memory and add overhead to node access and allocation. Moreover, note that collectors such as the Appel EllisLi [AEL88] collector are sequential; the concurrency arises from a single process performing collection while other (mutator) process(es) are executing the program. 7 program : fundef; Delta Delta Delta ; fundef fundef : f(pat; pat) expr expr : if expr then expr else expr j d(expr; ....

A. Appel, J. Ellis and K. Li, Real-time concurrent collection on stock multiprocessors, ACM Symp. on Programming Language Design and Implementation, 1988.

....of delivering events one at a time) This thesis presents and evaluates a new mechanism for propagating event information through the Linux kernel into applications. There are several types of applications that may bene t from a better event propagation mechanism such as garbage collectors [8, 19], persistent storage systems [32] and distributed shared memory [23] We focus our discussion on the Linux operating system and simple event driven web servers and the kinds of events of interest to them (events providing information of activity on TCP IP sockets) Such web servers present a ....

A.W. Appel, J.R. Ellis, and K. Li. Real-Time Concurrent Collection on Stock Multiprocessors. In Proceedings of the ACM SIGPLAN '88 Conference on Programming Language Design and Implementation, pages 11-20, June 1988.

.... Modula 3, a garbagecollected language [10] Our local collector is a slightly modified version of the SRC Modula 3 incremental, mostly copying collector [4] Synchronisation between the mutator and the local collector is provided by a page wise read only barrier supported by the operating system [1]. Network Objects uses reference lists rather than counts: any client process holds at most one surrogate for any given network, or concrete, object. Our intable is represented by that part of (a modified version of) the Network Objects object table, that contains references, or wireReps, to ....

Andrew W. Appel, John R. Ellis, and Kai Li. Real-time concurrent collection on stock multiprocessors. ACM SIGPLAN Notices, 23(7):11--20, 1988.

.... WORK While numerous concurrent, multiprocessor collectors for general purpose programming languages have been described in the literature [3, 6, 8, 10, 12, 13, 15, 16, 18, 21, 22] the number that have been implemented is quite small and of these, only a few actually run on a multiprocessor [2, 6, 12, 10, 11, 20]. DeTreville s work on garbage collectors for Modula 2 on the DEC Fire y workstation [6] is the only comparative evaluation of multiprocessor garbage collection techniques. His algorithm is based on Rovner s reference counting collector [21] backed by a concurrent tracing collector for cyclic ....

....of Domani et al. [11] This is a generational collector based on the work of Doligez et al. [10, 9] for which generations were shown to sometimes provide signi cant improvements in throughput. No response time measurements were provided. The other implemented concurrent multiprocessor collectors [2, 12, 10, 20] are all tracing based algorithms for concurrent variants of ML, and generally have signi cantly longer maximum pause times than our collector. In addition, ML produces large amounts of immutable data, thereby simplifying the collection process. The collector of Huelsbergen and Larus [12] for ML ....

Appel, A. W., Ellis, J. R., and Li, K. Real-time concurrent collection on stock multiprocessors. In Proceedings of the SIGPLAN Conference on Programming Language Design and Implementation (Atlanta, Georgia, June 1988), ACM Press, New York, New York, pp. 11-20.

.... Collection While numerous concurrent, multiprocessor collectors for general purpose programming languages have been described in the literature [6, 8, 10, 12, 13, 15, 16, 18, 22, 23] the number that have been implemented is quite small and of these, only a few actually run on a multiprocessor [2, 6, 12, 10, 11, 21]. DeTreville s work on garbage collectors for Modula 2 on the DEC Fire y workstation [6] is the only comparative evaluation of multiprocessor garbage collection techniques. His algorithm is based on Rovner s reference counting collector [22] backed by a concurrent tracing collector for cyclic ....

....of Domani et al. [11] This is a generational collector based on the work of Doligez et al. [10, 9] for which generations were shown to sometimes provide signi cant improvements in throughput. No response time measurements were provided. The other implemented concurrent multiprocessor collectors [2, 12, 10, 21] are all tracing based algorithms for concurrent variants of ML, and generally have signi cantly longer maximum pause times than our collector. In addition, ML produces large amounts of immutable data, thereby simplifying the collection process. The garbage collector of Huelsbergen and Larus [12] ....

Appel, A. W., Ellis, J. R., and Li, K. Realtime concurrent collection on stock multiprocessors. In Proceedings of the SIGPLAN Conference on Programming Language Design and Implementation (PLDI) (Atlanta, Georgia, June 1988), ACM Press, New York, New York, pp. 11-20.

....programming language implementors have been striving for a pause free garbage collector. So far, no one has found a perfect solution. Each scheme has its own weakness; for example reference counting fails to reclaim circular structures; Baker s algorithm [Bak77] and its descendents, including [LH83, AEL88]) require a potentially costly read barrier; Generation Scavenging [Ung84, Ung86, UJ88] may fill up memory with tenured garbage, opportunistic strategies [Wil89] may fail to receive adequate opportunities, multiple generation stop and copy algorithms [CW86] may experience long waterfall pauses ....

A.W. Appel, J. R. Ellis, K. Li, "Real-Time Concurrent Collection on Stock Multiprocessors," ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI'88), June 1988, 11-20.

....store instructions. RPA can perform a snapshot at beginning store barrier by inserting an extra load just before the store instruction, and then profiling that load instruction. Several concurrent GC algorithms use virtual memory page protection to replace in lined barriers. Appel, Ellis and Li [1] use this technique for an algorithm based on Baker s algorithm. Baker s algorithm divides to space into three areas. The scanned area contains objects that have been scanned for references to from space. All references in this area have been redirected; the scanned area contains only references ....

Andrew W. Appel, John R. Ellis, Kai Li, "Real-Time Concurrent Collection on Stock Multiprocessors," 1988 Conf. on Programming Language Design and Implementation, pp. 11-20, June 1988.

....of the activation record. Most CPS based compilers [20, 17] allocate certain closures on a stack if their pattern of usage can be sufficiently analyzed at compile time, so that they can be efficiently and promptly deallocated. However, using a stack complicates the garbage collector interface[6], and greatly complicates the efficient implementation of call with current continuation [13] The tech niques we present here neither require nor preclude the use of a stack. We will avoid any technique that calls for side effects to existing closures. In the first place, this corrupts the ....

Appel, Andrew W., Ellis, John R., and Li, Kai. Real-time concurrent collection on stock multiprocessors. SIGPLAN Notices (Proc. SIGPLAN '88 Conf. on Prog. Lang. Design and Implementation), 23, 7 (1988) 11 20.

....general conclusions about what user programs require from the operating system and hardware. Concurrent garbage collection A concurrent, real time, copying garbage collection algorithm can use the page fault mechanism to achieve medium grain synchronization between collector and mutator threads [4]. The paging mechanism provides synchronization that is coarse enough to be efficient and yet fine enough to make the latency low. The algorithm is based on the Baker s sequential, real time copying collector algorithm [6] Baker s algorithm divides the memory heap into two regions, from space ....

....must be performed on every fetch. Furthermore, the mutator and the collector must alternate; they cannot operate truly concurrently because they might simultaneously try to copy the same object to different places. Instead of checking every pointer fetched from memory, the concurrent collector [4] uses virtual memory page protections to detect from space memory references and to synchronize the collector and mutator threads. To synchronize mutators and collectors, the algorithm sets the virtual memory protection of the unscanned area s pages to be no access. Whenever the mutator tries to ....

[Article contains additional citation context not shown here]

Andrew W. Appel, John R. Ellis, and Kai Li. Real-time concurrent collection on stock multiprocessors. SIGPLAN Notices (Proc. SIGPLAN '88 Conf. on Prog. Lang. Design and Implementation), 23(7):11--20, 1988.

No context found.

APPEL, A. W., ELLIS, J. R., AND LI, K. Real-time concurrent collection on stock multiprocessors. In Proceedings of the SIGPLAN'88 Conference on Programming Language Design and Implementation (Atlanta, Georgia, June 1988). SIGPLAN Notices, 23, 7 (July), 11--20.

No context found.

Andrew W. Appel, John R. Ellis, and Kai Li. Real-time concurrent collection on stock multiprocessors. ACM SIGPLAN Notices, 23(7):11-20, 1988.

No context found.

A. W. Appel, J. R. Ellis, and K. Li. Real-time concurrent collection on stock multiprocessors. ACM SIGPLAN Notices, 23(7):11--20, 1988.

No context found.

A. W. Appel, J. R. Ellis, and K. Li. Real-time concurrent collection on stock multiprocessors. ACM SIGPLAN Notices, 23(7):11--20, 1988.

No context found.

APPEL, A. W., ELLIS, J. R., AND LI, K. Real-time concurrent collection on stock multiprocessors. In Proceedings of the SIGPLAN'88 Conference on Programming Language Design and Implementation (Atlanta, Georgia, June 1988). SIGPLAN Notices, 23, 7 (July), 11--20.

No context found.

APPEL, A. W., ELLIS, J. R., AND LI, K. Real-time concurrent collection on stock multiprocessors. In Proceedings of the SIGPLAN Conference on Programming Language Design and Implementation (Atlanta, Georgia, June 1988). SIGPLAN Notices, 23, 7 (July), 11--20.

No context found.

APPEL,A.W.,ELLIS,J.R.,AND LI, K. Real-time concurrent collection on stock multiprocessors. In Proceedings of the SIGPLAN'88 Conference on Programming Language Design and Implementation (Atlanta, Georgia, June 1988). SIGPLAN Notices, 23, 7 (July), 11--20.

No context found.

Andrew W. Appel, John R. Ellis, and Kai Li. Real-time concurrent collection on stock multiprocessors. ACM SIGPLAN Notices, 23(7):11-20, 1988.

No context found.

Andrew W. Appel, John R. Ellis, and Kai Li. Real-time concurrent collection on stock multiprocessors. ACM SIGPLAN Notices, 23(7):11--20, 1988.

First 50 documents  Next 50

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