yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxxxxxx>]

Hi Jérôme,
I don't know, but I guess they were introduced independently by 
different authors.
I actually don't know why we would need to define tolerance at all, 
unless in very specific cases.

For instance I judge that a line like this (ViscoelasticPM.cpp) is not 
only useless but actually wrong:
if (std::abs(cn1) <= Mathr::ZERO_TOLERANCE ) cn1=0;

If it's small let it be small. No need to make it zero. And since cn1 is 
not dimensionless it implies that currently the accuracy depends on a 
specific choice of units...

Bruno




Bruno

On 11/15/18 3:54 PM, Jerome Duriez wrote:
> Hello,
>
> Out of curiosity, why do we have both Mathr::ZERO_TOLERANCE and 
> Mathr::EPSILON in the code [*] ?
>
> I would understand they're both intended to give an expected magnitude 
> of rounding errors. (Note that ZERO_TOLERANCE is hardcoded as 1e-20, 
> while EPSILON is ~ 2e-16 here)
>
> Both are used throughout the code (EPSILON more often ?)
>
>
> Thanks,
>
> Jérôme
>
>
>
> [*] 
> https://github.com/yade/trunk/blob/3269232e4982c1aa527581a200a0224555b09a1e/lib/base/Math.cpp
>
> ------
> Chargé de Recherche / Research Associate
> Irstea, RECOVER
> 3275 route de Cezanne – CS 40061 13182 Aix-en-Provence Cedex 5 FRANCE
> +33 (0)4 42 66 99 21
>
>
> _______________________________________________
> Mailing list: https://launchpad.net/~yade-dev
> Post to     : yade-dev@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~yade-dev
> More help   : https://help.launchpad.net/ListHelp