[gmx-developers] GMX_FLOAT_EPS changed between 4.6 and 5.0

Berk Hess hess at kth.se
Sat Sep 12 22:47:34 CEST 2015


Our definition of eps wasn't according to the standard so we decided to change it to the standard. I realized that this would change the PME table spacing and decided that the sqrt(2) wider table was accurate enough. We used to have a max error of the old eps, which was half the least significant bit, now it's the full last significant bit (which is still causes less error than nearly all other aspects of the MD calculation).
But since that was already changed before the 5.0 release, how could that cause issues now?



On Sep 12, 2015 10:15 PM, David van der Spoel <spoel at xray.bmc.uu.se> wrote:
> Hi, 
> I'm having issues getting a patch to behave in Jenkins, but now it turns 
> out that it may be due to the definition of GMX_FLOAT_EPS 
> in 4.6 we have 
> types/simple.h:#define GMX_FLOAT_EPS    5.96046448E-08 
> From 5.0 we have 
> simple.h:#define GMX_FLOAT_EPS    1.19209290e-07F 
> (twice as big?) 
> This has an impact on the table spacing for the PME corrections. The 
> table spacing now is sqrt 2 less. 
> Since all tests until work there may be other issues too (in particular 
> in my code obviously), but suggestions are welcome. 
> Cheers, 
> -- 
> David van der Spoel, Ph.D., Professor of Biology 
> Dept. of Cell & Molec. Biol., Uppsala University. 
> Box 596, 75124 Uppsala, Sweden. Phone: +46184714205. 
> spoel at xray.bmc.uu.se    http://folding.bmc.uu.se 
> -- 
> Gromacs Developers mailing list 
> * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before posting! 
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists 
> * For (un)subscribe requests visit 
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers or send a mail to gmx-developers-request at gromacs.org. 

More information about the gromacs.org_gmx-developers mailing list