Koen Langendoen and Dirk-Jan Agterkamp. Cache behaviour of lazy functional programs. Proceedings of the Fourth International Workshop on Parallel Implementations of Functional Languages, Aachen, September 1992, Aachener Informatik Berichte 92-19, RWTH Aachen, Germany, 1992. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Efficient Shared-Memory Support for Parallel Graph Reduction - Bennett, Kelly (1996) (Correct)
....This is also present in parallel imperative programs, although it is inevitably application dependent. Previous work has indicated that locality is also present in sequential functional programs: several simulation studies have shown the advantage of using large cache lines with graph reduction [28, 31]. We present simulation results which quantify the opposing effects of locality and false sharing in Section 6. In the remainder of this section we discuss the synchronisation scheme of PGR and determine how consistency can be relaxed in order to reduce line stealing and thereby increase the ....
Koen Langendoen and Dirk-Jan Agterkamp. Cache behaviour of lazy functional programs. Proceedings of the Fourth International Workshop on Parallel Implementations of Functional Languages, Aachen, September 1992, Aachener Informatik Berichte 92-19, RWTH Aachen, Germany, 1992.
Locality and False Sharing in Coherent-Cache Parallel Graph.. - Bennett, Kelly (1993) (1 citation) (Correct)
....using a direct write instruction [9] since the old contents of the line will not be used ffl We assume that infinite sized, fully associative caches are used, so cache capacity effects do not occur. The extra overheads caused by using non fully associative caches is unlikely to be significant [13]. Adopting infinite sized caches allows the main memory unit to be removed entirely and all shared data resides in a cache Although each assumption is invalid on a real machine, they prevent the results from being obscured by other effects. In addition, these assumptions reduce the number of ....
....The parallel evaluation strategy is a modified form of graph reduction in which shared nodes are only ever read, greatly simplifying the coherency problem, and making direct comparisons with our work difficult. A related study of caching effects in a sequential functional language implementation [13] found significant locality which was attributed to the misses caused by the incremental allocation strategy, the LIFO behaviour of the graph reducer stack and the evaluation of suspension (in which multiple fields are read in a burst) Our stacks are not shared and we assume they are dealt with ....
Koen Langendoen and Dirk-Jan Agterkamp. Cache behaviour of lazy functional programs. Proceedings of the Fourth International Workshop on Parallel Implementations of Functional Languages, Aachen, September 1992, Aachener Informatik Berichte 92-19, RWTH Aachen, Germany, 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