Stochastic analysis of cache thrashing Analyse stochastique des fautes de cache

作者: Xavier Gréhant

DOI:

关键词: Parallel computingThrashingCacheComputer science

摘要: Cache misses are central to processor performance. The analysis of memory access patterns is widely used for performance prediction under monotasking. monotasking hypothesis gives a lower bound the cache miss ratio. With multitasking, which rule on mainstream operating systems, additional caused by context switches. We present higher overhead from switches, based stochastic how warms up, i.e. fills up with useful data. observe that as close actual resulting segment covers exact ratio in finite time quantum, potentially representative whole execution. Les fautes de sont determinantes pour la des processeurs. L’analyse d’acces memoire est largement utilisee en l’absence partage du entre processus. Cette hypothese permet d’obtenir une limite inferieure cache. En presence concurrence processus, ce qui le cas dans les systeme d’exploitation habituels, un surplus aux changements contexte. Nous proposons superieure nombre additionnelles dues contexte, partir d’une analyse stochastique facon dont se ”rechauffe”, c’est-a-dire remplit donnees utiles au processus cours. observons que aussi proche valeur reelle. Le obtenu comprend exacte quantum fini, potentiellement representatif totalite l’execution.

参考文章(18)
Xavier Grehant, Sverre Jarp, Lightweight Task Analysis for Cache-Aware Scheduling on Heterogeneous Clusters arXiv: Distributed, Parallel, and Cluster Computing. ,(2009)
John L. Hennessy, David A. Patterson, Computer Architecture: A Quantitative Approach ,(1989)
K. Grimsrud, J. Archibald, R. Frost, B. Nelson, On the accuracy of memory reference models Proceedings of the 7th international conference on Computer performance evaluation : modelling techniques and tools: modelling techniques and tools. pp. 369- 388 ,(1994) , 10.1007/3-540-58021-2_21
Calin CaΒcaval, David A. Padua, Estimating cache misses and locality using stack distances international conference on supercomputing. pp. 150- 159 ,(2003) , 10.1145/782814.782836
Xudong Shi, Feiqi Su, Jih-kwon Peir, Ye Xia, Zhen Yang, CMP cache performance projection ACM SIGARCH Computer Architecture News. ,vol. 35, pp. 13- 20 ,(2007) , 10.1145/1241601.1241607
Fang Liu, Fei Guo, Yan Solihin, Seongbeom Kim, Abdulaziz Eker, Characterizing and modeling the behavior of context switch misses Proceedings of the 17th international conference on Parallel architectures and compilation techniques - PACT '08. pp. 91- 101 ,(2008) , 10.1145/1454115.1454130
Fei Guo, Yan Solihin, An analytical model for cache replacement policy performance ACM SIGMETRICS Performance Evaluation Review. ,vol. 34, pp. 228- 239 ,(2006) , 10.1145/1140103.1140304
M.D. Hill, A.J. Smith, Evaluating associativity in CPU caches IEEE Transactions on Computers. ,vol. 38, pp. 1612- 1630 ,(1989) , 10.1109/12.40842
James Demmel, W. Kahan, Accurate singular values of bidiagonal matrices Siam Journal on Scientific and Statistical Computing. ,vol. 11, pp. 873- 912 ,(1990) , 10.1137/0911052
G. Edward Suh, Srinivas Devadas, Larry Rudolph, Analytical cache models with applications to cache partitioning international conference on supercomputing. pp. 1- 12 ,(2001) , 10.1145/2591635.2667181