CPU time in the R.I.N.G.S. code

E-mail Print

In the R.I.N.G.S. code numerous analysis can be performed nevertheless CPU time is mainly driven by the ring statistics analysis, in particular by the maximum search depth, ie. the maximum size for a ring which is an input parameter fixed during the analysis .

This has been illustrated in figure [Fig. 1] which represents the evolution of the CPU time as a function of the maximum search depth.


CPU time in the R.I.N.G.S. code
Figure 1: Evolution of the CPU time as a function of the maximum search depth. A single configuration of g-GeS2 at 300K (258 particles) has been analyzed. Furthermore all nodes have been used to initiate the search. (3600 s = 1 h, 86400 s = 24h, 864000 s = 10 days, 592000 s = 1 month). Computations were performed using an Intel Xeon 2.6 GHz processor with 1.5 Go of memory.

 

As shown in figure [Fig. 1] the CPU times increase exponentially with the maximum search depth.

Furthermore we can see in figure [Fig. 1] that the evolution of the CPU time is similar in the cases of King, shortest path and primitive rings criteria (the CPU times for the strong ring criterion, not shown here, are much more important). Therefore the MPI parallelization appears fully justified to performed the ring statistics analysis of network material with a high search depth.

Last Updated on Wednesday, 05 May 2010 22:51  

 Latest version

RINGS Last version logo
A new version of the RINGS code has been released !

 Mailing List

Mailing list
Join the RINGS code mailing list !

 Hosted by

Get rings-code at SourceForge.net. Fast, secure and Free Open Source software downloads