yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

Don't worry Vaclav, I know it works (and worked), I played with it a lot.
Or, let's say, it worked for what you wanted, but not for what I want. I 
also fully understand that you are afraid of breaking this code because 
it is smart and complex (I had to understand how it WORKS the hard way).

Let me repeat it again, and if you already implemented that, excuse me 
for being an idiot, and please send me a demo script :

I want a constant velGrad like :

1 x 0
0 1 0
0 0 1

and I want to never reach the algorithm limit. I have working code for 
this, using your algorithm.

> > In one sense you were right, it 
> > already works (after a few simplifications here and there though, but it 
> > is only few lines changes).
> >     
> What changes, except for fixing cos?
>
>   
I was reffering to my current code, not commited and not in the diff. 
Hard to explain in few lines. In short, I change some sequences of 
wrap/shear.
> > It needs a few more things in cell so that the renderer will display the 
> > correct thing, but infinite shear works here at the moment (more 
> > difficult to understand with a single diff as in previous one).
> >     
> It did work before as well...
>   
The renderer display a deformed cell (for me at least). I want a fixed 
one, or the period looks like a thin line of spheres after 600% shearing 
and we don't see anything.

> > Appart from that, the only gain is you can "simple shear" a bit longer 
> > before the aabb size limit.
> >     
> Are you sure your enlarge is correct? It must be mathematically equal to
> cos for in-plane shear, since that is derived geometrically. 
The result is exactly equal for planar pure shear, I tested it. But 
there are two different things here.
1. I used the same mathematics as you did to compute non-orthogonal 
distances between parallel planes (some commented lines not included in 
the sent diff, doing with Hnorm and Hortho what you see with Unorm and 
Uortho), I confirmed exact equality in this case.
2. I unshear (or say, unstretch, since it is exactly the same thing) 
from eigen basis of U instead of using the deformed vectors of Hnorm. It 
divides the intensity of box deformations by a factor 3, thats why the 
simulation last longer. You see this part of the code in the diff, but 
the paradox is I don't use that any more. Shearing a bit longer is 
useless when you can shear forever. :)

> The rendering code for Aabb is correct, I think, so you can use that +
> some extreme shear value to at least visually check.
>
>   
I never diplayed Aabb's or wanted to.

Bruno