There are known and justifiable reason for potential differences in iteration count and run time of the code, potentially related to, but not limited to, random seed and convergence. But the case you demonstrate below is a bit extreme. And past experience running that code makes me dubious of the claim.
Did you only perform 2 tests? Did the computation time for additional tests continue to demonstrate that variability? Are you sure IO variability (for whatever reason) can't account for the difference in runtime, while actual CPU time is the same? Nonetheless I will investigate by running that binary on some of our many datasets to see if I can replicate such an instance. -Zeke > On Jul 2, 2015, at 3:45 AM, Jordi Delgado <jordi...@pic.es> wrote: > > Dear all, > > I'm performing a study about the variability of the execution time of > recon-all. I noticed that the execution of recon-all in Freesurfer 5.3 for > two identical MRIs (two copies of the same MRI), takes different execution > time. > > I'm performing the execution using GPUs capability, but the same is happening > for the CPU-only executions. > > Checking the output logs I found that in for example CA Reg, the number of > the iterations is different, what's the main source of this variability? I > guess that maybe It is related to some random seeds used and the convergence > of the implemented methods. Am I right? > > CA Reg example: > > Execution Time 1 - Subject 1: 00:25:23 > Execution Time 2 - Subject 1: 00:32:14 > > In both cases the computing node was used exclusively used for the recon-all > execution. > > Did you measured the percentage of this variability? Can we extract some > measures to compare the performed work on each execution? I'm thinking on the > number of GPU kernels executed, the number of the iterations of the methods... > > Thank you in advance, > > -- > Jordi Delgado Mengual > PIC (Port d'Informació CientÃfica) > Campus UAB, Edifici D > E-08193 Bellaterra, Barcelona > http://www.pic.es > Avis - Aviso - Legal Notice: http://www.ifae.es/legal.html > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
