yade-dev team mailing list archive
-
yade-dev team
-
Mailing list archive
-
Message #00936
Re: [Yade-users] Be carefull with "normalize"
> a simple solution to avoid this (without modifying [mini]wm3) is to
> change the user units.
> I think yade computation can be done with any system of units as they
> are consistent.
> For example, one can use centimeters instead of meters and deca-
> Newtons instead of Newtons and so on.
> But we must pay particular attention to the parameter values, eg. Kn
> that was N/m becomes daN/cm.
> That is exactly what JJ Moreau was doing for his very high-precision
> computations on old apple.
> I think that people that perform MD simulations (with atom-sized
> particles) do the same.
Want repeat this one? http://www.cnn.com/TECH/space/9909/30/mars.metric/
Seriously, for myself I will never use anything else than basic SI units
since it is very easy to do mistake. Dimensionless quantitites don't scale
with units anyway (like strains or angles). I would prefer to fix that in
Wm3 (or Wm4, for that matter) or switch to another lib -- but I didn't find
anything quite suitable. Have a look at e.g.
http://www.ohloh.net/tags/3d/math.
And I still think that the tolerance value has nothing to do with
precision. NUmber of decimal places is the same for 1e-20 and for 1e20. But
the tolerance must scale with that, that is the only problem, as I see it.
How anout making ZERO_TOLERANCE and let the user set it to some reasonable
value according to the particle size, timestep, etc?
Vaclav
Follow ups
References