Re: [ESPResSo-users] Implementation of boundaries, lb_boundaries.cpp

[Stefan Kesselheim]

Hi Markus, 
thanks for reading my beautiful code :-).

On Feb 4, 2014, at 1:30 PM, "Wink, Markus" <address@hidden> wrote:

> Dear all,
>  
> I have got a question concerning the implementation of boundaries in the 
> Lattice-Boltzmann source code (lb-boundaries.cpp). For that, a flag (stored 
> in the structure “lbfields[k].boundary”) is set to decide, whether the node 
> is a fluid of boundary node.
>  
> As far as I understood the code, the function “calculate_wall_dist” in line 
> 281 calculates the distance of the node to the wall. The if-clause in line 
> 309 compares that distance to the distance dist=1e99 and gives out the 
> minimum. Line 315 ff. finally sets the flag to the structure 
> lbfields.boundary.
>  
> Here are my questions:
> 1)    A flag of 0 means fluid node, doesn’t it? All boundary-nodes are 1 or 
> bigger.

True. Values > 1 mean "this is a wall node" and the number is the index of the 
lbboundary+1.

> 2)    Why do you initialize the variable “dist” with 1e99. I expected 
> something like: compare the distances origin-node to origin-wall; if the 
> first one is smaller, one gets a boundary node, otherwise a fluid-node. Where 
> does that number come from?

We search for the lbboundary with the lowest distance to a particular node. 
1e99 is larger than any other floating number, that we would expect there and 
could also be a constant like LARGEST_FLOAT (I don't think such a thing is 
defined in espresso).
We want to know the minimum distance to any boundary, and a negative value 
indicates that a node is "in" the wall. 1e99 is just a simple coding trick to 
avoid a bit of code. Probably unnecessary.

Did that help?
Cheers
Stefan