Re: [Gfs-devel] fit_curvature and parabola_simpler

[Stephane Z. (lmm) <za...@lm...>]

Hi Stephane,

You are always correct.

My surface tension was in the range of 1 (SourceTension T 1.026 K) as
I was working with L = 1. I guess I was getting this error for the
cause that you have mentioned. But my maxlevel was 13 (!).

I printed the value of kappa and sigma just before the crash as per
your suggestion. These are as follows kappa = 1001799.187462  sigma =
1.026000

So I changed the value v > 108 and now I found no crash.

Thanks for your useful suggestion. I also noticed that when I run the
same code for SourceTension T 0.046e0 K and PhysicalParams { L =
0.03e0 } there is no crash. Your point is proved once again.

Though it is a small change still I am sending you the patch.

regards,

Arup
(forwarded by Stéphane Zaleski)

Le 18 oct. 2011 à 04:55, Stephane Popinet a écrit :

> Hi Arup,
> 
> Could you please reply to the mailing list and not only to me? thanks.
> 
>> As we are using gfs_domain_face_fraction (kappa->domain, face) there
>> is a possibility of getting v value very high near sharp bend at the
>> interface.
> 
> It all depends on the value of your surface tension coefficient (in
> Gerris units). If surface tension is of order 1 then kappa*sigma will
> only be greater than 1e6 when your spatial resolution is of order 1e-6
> (i.e. a refinement level of log(1e6)/log(2.) ~ 20).
> 
> But 1e6 maybe too small. Could you please check what values you have
> for kappa and sigma when it crashes? If these values are consistent,
> then we will need to increase the threshold from 1e6 to something
> larger.
> 
>> if (v > 1e6) v = 0e6; just before line 922 of poisson.c. This is to
>> neglect surface forces if i have a very high curvature (isolated
>> interface).
> 
> I believe you should just change the 1e6 value to something larger
> e.g. 1e20 (after having checked the above).
> 
> If this works, please send me the corresponding patch and I will
> include it in the main branch.
> 
> cheers
> 
> Stephane
> 
> ------------------------------------------------------------------------------
> All the data continuously generated in your IT infrastructure contains a
> definitive record of customers, application performance, security
> threats, fraudulent activity and more. Splunk takes this data and makes
> sense of it. Business sense. IT sense. Common sense.
> http://p.sf.net/sfu/splunk-d2d-oct
> _______________________________________________
> Gfs-devel mailing list
> Gfs...@li...
> https://lists.sourceforge.net/lists/listinfo/gfs-devel