ISSN 0021-3454 (print version)
ISSN 2500-0381 (online version)
Menu

1
Issue
vol 63 / January, 2020
Article

DOI 10.17586/0021-3454- 2017-60-10-961-966

UDC 004.254

THE PROBLEM OF FALSE SPLITTING OF CPU CACHE MEMORY STRINGS IN MULTIPROCESSOR SYSTEMS

M. S. Kosyakov
ITMO University; Department of Computer Technology;


A. O. Kuzichkina
ITMO University, Department of Computation Technologies; Student


Read the full article 

Abstract. The organization of memory in modern multi-core multiprocessor systems causes the problem of false split of CPU cash string addressed in the paper. The false CPUs cache lines split makes tasks scheduled to be run in parallel by several cores, are executed sequentially. The problem is difficult to detect, and its presence can cause a sharp degradation of the system performance. Programming patterns which give rise to the problem are identified. The templates execution time is measured in the presence of false splitting and after applying the methods proposed in the work to solve the problem. The effect of the number of threads running in parallel on the speed-up of programming templates execution and the efficiency of utilization of the system computing resources are analyzed. It is shown experimentally that the false split significantly affects temporal characteristics of multi-threaded applications. Several methods to eliminate the difficulties are proposed.
Keywords: multiprocessor systems, cache memory, false splitting, multi-threaded application

References:
  1. Tsil'ker B.Ya., Orlov S.A. Organizatsiya EVM i sistem (Organization of the Computers and Systems), St. Petersburg, 2011, 688 р. (in Russ.)
  2. Tanenbaum A.S. Computer Networks, 4th edition, Prentice Hall, PTR UpperSaddIe River, New Jersey, 2003. 
  3. Drepper U. What Every Programmer Should Know about Memory, Red Hat, Inc., 2007, vol. 11, 114 p.
  4. Venkataramani G., Hughes C.J., Kumar S., Prvulovic M. Proceedings of the 13th Workshop on Interaction between Compilers and Computer Architecture, 2009, 10 p.
  5. Hennessy J.L., Patterson D.A. Computer Architecture a Quantitative Approach, Fifth edition, Waltham (Massachusetts), Elsevier, 2012, 848 p.
  6. Williams A. C++ Concurrency in Action. Practical Multithreading, 2012, 528 p. ISBN 9781933988771.
  7. http://ark.intel.com/ru/products/64596/Intel-Xeon-Processor-E5-2690-20M-Cache-2_90-Ghz-8_00-Gts-Intel-QPI.
  8. Brendan G. Systems Performance: Enterprise and the Cloud, 1st edition, Upper Saddle River, New Jersey, Prentice Hall,2013, 729 p.