yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

> You should call it something clearer, like scaleDisplacementT() (why
> Max?).
>
>   
Because there was "max" in the original name and I went for the smallest 
change. I agree with the name you suggest.
I fixed the division by zero, thanks for pointing it out.


> Since you added it as virtual function to the Dem3DofGeom base class,
> you MUST also implement in Dem3DofGeom_FacetSphere and
> Dem3DofGeom_WallSphere (not just Dem3DofGeom_SphereSphere).
>
>   
Oh, you are right. It is a problem. I'll think about it.

> Make it return void, not zero. It is more confusing to return
> meaningless zero all the time, rather than nothing.
>
>   
I will make it return the plastic displacement 2*diff (vector) as it is 
computed inside the function. It can be used to compute energy terms 
(btw, a general question : why computing distances for energy? can't we 
simply use cross products between force and disp. vectors?).

> I don't really see why you don't add it as non-virtual method to
> Dem3DofGeom which will call internally slipToDisplacementTMax (which is
> virtual). This way you have to implement it at 3 times.  I don't think
> computing norm() takes that much time (it could return squaredNorm
> anyways, and compute the sqrt only where it is really used).
>   
Looking only the interesting lines, and concatenating nested 
slipTo/scaleDisp functions, we have

Before :

if(trialFs.squaredNorm()>maxFsSq){ 
geom->slipToDisplacementTMax(sqrt(maxFsSq)/phys->ks); 
trialFs*=sqrt(maxFsSq/(trialFs.squaredNorm()));}
Vector3r p1=contPtInTgPlane1(), p2=contPtInTgPlane2();
   Real currDistSq=(p2-p1).squaredNorm();
   Vector3r diff=.5*(sqrt(currDistSq)/displacementTMax-1)*(p2-p1);
   return 2*diff.norm();

= 4 squaredNorm() + 4 sqrt() + 3 divisions (considering norm() = 
squaredNorm() + sqrt())


After :

   Real trialFsSq = trialFs.squaredNorm();
    if(trialFsSq>maxFsSq){
       Real multiplier=sqrt(maxFsSq/trialFsSq);
       geom->scaleDisplacementT(multiplier); trialFs*=multiplier;}
       Vector3r diff=.5*(multiplier-1)*(p2-p1);
       return 2*diff;

= 1 squaredNorm() + 1 sqrt() + 1 division

Does it make sense? :)


Cheers.

Bruno