yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

Václav Šmilauer a écrit :
> > It looks correct. Usage is restricted to NormalShearInteractions (I
> > guess yes), or can we cast different Interaction types to this?
> Well, ElasticContactInteraction (for instance) derives from
> NormalShearInteraction (which in turn derives from NormalInteraction).
> Eventually _all_ interactions that create normal and shear forces should
> derive from NormalShearInteraction and there will be no need for casts.
>
> Those parent classes store kn and ks as well, therefore the current
> GlobalStiffnessCounter will work for many different types of
> interactions, provided they have defined kn and ks.
>
> I don't know what other (besides ElasticContactInteraction and
> BrefcomContact) interactions we have but it might be useful to add those
> to this class hierarchy as well, progressively.
>
> I also hope that such generic interactions will reduce code duplication
> (for instance, there is unbalancedForce computation in Triaxial, but now
> it can use this function).
Very good. I get it you modified the class hierarchy, or NormalShearI 
has been here for a while and I didn't notice?

> > It made me think the we should keep an updated list of macros used in
> > Yade (YADE_PTR_CAST, FOREACH, REGISTER, etc), since they are either
> > Yade macros, boost macros or any-other-lib macros, it can be a bit
> > complex to read for newcomers.
> If you create a wiki page, I will edit the macros I know something about.
>
> Oh, any reason why global stiffness timestepper is 2 engines (GSCounter
> and GSTimeStepper)? You could have just 1 engine that would compute
> stiffnesses (store them not in physicalActions, but in a vector<Real> or
> such) and then compute the timestep. (The code would also be easier to
> understand -- in 1 place)
Agreed. Two main reasons for current situation :
1. I tried to comply with Yade design - in a stupid way maybe.
2. Global stiffness is (probably more "could" or "should", in fact...) 
also used to control the stress on walls. So the stiffness data is a bit 
more general than just timestep computing.

One more reason to merge : if somebody makes the mistake of setting 
different intervals in counter and timestepper, it could result in a 
strange behaviour, perhaps uninitialized values, etc.

Bruno



--

_______________
Chareyre Bruno
Maitre de conference

Grenoble INP
Laboratoire 3SR - bureau E145
BP 53 - 38041, Grenoble cedex 9 - France
Tél : 33 4 56 52 86 21
Fax : 33 4 76 82 70 43
________________