E. Cooper, S. Nettles, and I. Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In Conference Record of the 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....made interactive response time so poor that users preferred to turn garbage collection off and reboot once the virtual address space was consumed. He also demonstrates that some applications execute fastest with garbage collection turned off. Appel [2] and Cooper, Nettles, and Subramanian [8] describe techniques for determining heap sizes in copying collectors. Unfortunately, neither study specifically examines the impact that their decisions have on application performance. Smith and Morrisett [15] describe a mostlycopying collector that collects garbage whenever the heap 3 Of the ....

Eric Cooper, Scott Nettles, and Indira Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In Conference Record of the

....that garbage collection made interactive response time so poor that users preferred to turn garbage collection o and reboot once the virtual address space was consumed. He also demonstrates that some applications execute fastest with garbage collection turned o . Cooper, Nettles, and Subramanian [8] show how the performance of Standard ML can be improved by applying optimizations to a simple generational collector introduced by Appel [2] In addition to utilizing Mach s support for sparse address spaces and external pagers, they propose and study a modi cation to Appel s algorithm for ....

E. Cooper, S. Nettles, and I. Subramanian. Improving the performance of SML garbage collection using application-specic virtual memory management. In Conference Record of the 1992 ACM Symposium on Lisp and Functional Programming, pages 43-52, San Francisco, CA, June 1992. ACM Press.

....mechanism, and the old UNIX brk system call. In the latter case, the library also provides an implementation of the malloc library. Porting the memory subsystem to new operating systems is straightforward; it should also be possible to take advantage of features, such as MACH s external pagers [CNS92], in some implementations of the subsystem. 6 The new collector The new garbage collector actually consists of two collectors: a minor collector, and a major collector. The minor collector is responsible for collecting the allocation arena. It is designed to have minimal overhead and a small ....

....8 Related work There have been a number of replacements for the SML NJ collector, but most of these require special operating system support, compiler modifications, or both. Cooper, Nettles and Subramanian have described a collector for SML NJ that exploits MACH s external pager mechanism [CNS92]; it should be possible to integrate this approach into our memory subsystem. Nettles, O Toole, Pierce and Haines have developed a replication based collector for SML NJ that has good collector latency. This implementation, however, requires modification to the compiler and has worse overall ....

Cooper, E., S. Nettles, and I. Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In Conference record of the 1992 ACM Conference on Lisp and Functional Programming, June 1992, pp. 43--52.

....server s persistent storage. 6.6. 3 Shrinking and Garbage Collection Cooper et al. describe how they modified Standard ML of New Jersey (SML) 51] and implemented an SML specific external pager for Mach so that VM cache management could take advantage of the knowledge of the SML garbage collector [19]. The garbage collector marked pages as discardable or nondiscardable based on their current role in the garbage collection process. The combination of shrinking and garbage collection in Thor has some of the same character as their system, but Thor applies the idea to individual objects instead ....

Eric Cooper, Scott Nettles, and Indira Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In Proceedings of the ACM Symposium on Lisp and Functional Programming, pages 43--52, 1992.

.... for its bad paging behaviour when running on platforms with insufficient memory (where sufficient memory may be a large amount by conventional language standards) Even when there is enough physical memory to avoid paging entirely, the pauses caused by major collections are intolerably long [9, 20]. Cache performance of SML NJ was also suspected to be bad. It was speculated that 40 of execution time was spent waiting for main memory access, a 66 overhead. Recent work [11] has shown that this is not altogether true, at least for some current architectures and memory subsystem ....

Eric Cooper, Scott Nettles, and Indira Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In 1992 ACM Conference on Lisp and Functional Programming, pages 43--52, San Francisco, California, June 1992.

....the allocationarea size. For example, the allocation area is written from beginning to end between every minor collection. This access pattern has very poor paging and cache locality, and one would certainly want, for example, to limit the size of the allocation area to the available memory (Cooper, Nettles Subramanian [1992]) It has been suggested that making the allocation area small enough to fit in the cache might also improve cache locality (Wilson, Lam Moher [1992] Our collector has a run time option which allows us to limit or fix the size of the new generation allocation area. We examine the operation of ....

E Cooper, S Nettles & I Subramanian [June 1992], "Improving the performance of SML garbage collection using application specific virtual memory management, " in SIGPLAN Symposium on Lisp and Functional Programming, San Francisco, California.

....the allocation area size. For example, the allocation area is written from beginning to end between every minor collection. This access pattern has very poor paging and cache locality, and one would certainly want, for example, to limit the size of the allocation area to the available memory (Cooper, Nettles Subramanian [1992]) It has been suggested that making the allocation area small enough to fit in the cache might also improve cache locality (Wilson, Lam Moher [1992] Our collector has a run time option which allows us to limit or fix the size of the new generation allocation area. We examine the operation of ....

E Cooper, S Nettles & I Subramanian [June 1992], "Improving the performance of SML garbage collection using application specific virtual memory management, " in SIGPLAN Symposium on Lisp and Functional Programming, San Francisco, California.

....to control page replacement policy. These schemes do not provide resource allocation policies that satisfy our design principles to guarantee replacement performance. Furthermore, they are not concerned with file caching. Previous research on user level virtual memory page replacement policies [1, 9, 12, 15, 27] shows that application tailored replacement policies can improve performance significantly. With certain modifications, these user level policies might be used as user level file caching policies in our two level replacement. Recent work on prefetching [22, 10, 7] can be directly applied in ....

Eric Cooper, Scott Nettles, and Indira Subramanian. Improving the Performance of SML Garbage Collection using Application-Specific Virtual Memory Management. In Proceedings of the 1992 ACM Conference on LISP and Functional Programming, June 1992.

....before issuing these writes in order to ensure that RVM is not holding any old log records that apply to stable to space. It would be much more efficient to discard these log records instead. Garbage collectors sometimes benefit from similar features in data caches and virtual memory systems [7]. When the garbage collector reclaims a semi space it is better that cache lines and virtual memory pages that contain reclaimed data be reset to a zero fill on demand status because there is no need for the underlying memory system to preserve their old contents. Recently the designers of RVM ....

Eric Cooper, Scott Nettles, and Indira Subramanian. Improving the Performance of SML Garbage Collection using Application-Specific Virtual Memory Management. In Proceedings of the ACM Conference on Lisp and Functional Programming, pages 43--52, June 1992.

....collection work from the elapsed execution time. 4.1 Benchmarks Three benchmarks were used to test our implementation. Each was chosen because it stressed the memory management system in a different way. All benchmarks require many major and minor garbage collections during execution. See [5] for more details about these benchmarks. ffl Primes is a prime number sieve implemented in a simple lazy language which is in turn interpreted by an SML program. It allocates memory at a very high rate (approximately 10Mb sec) but few objects survive garbage collection. It is typical of ....

Eric Cooper, Scott Nettles, and Indira Subramanian. Improving the Performance of SML Garbage Collection using Application-Specific Virtual Memory Management. In Proceedings of the ACM Conference on Lisp and Functional Programming, pages 43-- 52, June 1992.

....to the overhead. 4.1 Benchmarks Three benchmarks were used to test our implementation. Each was chosen because it stressed the memory management system in a different way. All benchmarks require several minutes to execute and require many major and minor garbage collections during execution. See [6] for more details about these benchmarks. ffl Primes is a prime number sieve implemented in a simple lazy language which is in turn interpreted by an SML program. It allocates memory at a very high rate (approximately 10Mb sec) but few objects survive garbage collection. It is typical of ....

Eric Cooper, Scott Nettles, and Indira Subramanian. Improving the Performance of SML Garbage Collection using Application-Specific Virtual Memory Management. In Proceedings of the ACM Conference on Lisp and Functional Programming, pages 43--52, June 1992.

No context found.

E. Cooper, S. Nettles, and I. Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In Conference Record of the 1992.

No context found.

E. Cooper, S. Nettles, and I. Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In Conference Record of the 1992.

No context found.

E. Cooper, S. Nettles, and I. Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In Conference Record of the 1992.

No context found.

E. Cooper, S. Nettles, and I. Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In Conference Record of the 1992.

No context found.

E. Cooper, S. Nettles, and I. Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In Conference Record of the 1992.

No context found.

E. Cooper, S. Nettles, and I. Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In Proceedings of the 1992.

No context found.

E. Cooper, S. Nettles, and I. Subramanian. Improving the performance of SML garbage collection using application-specific virtual memory management. In Conference Record of the 1992.

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