[Users] Carpet: Domain specification fails

[dumsani]

Hi Ian,

Thanks for your reply, the particularly the last portion of it.

I had already figured out that the assertion failure was due to 
requesting CoordBase
to set the domain and at the same time have an aliased multipatch 
function trying
to do the same. Removing the CTGamma/Coordinates thorn lead to a whole 
host of
errors since most CTGamma thorns depend on (inherit) it. Hence I was 
also wondering
which thorn to replace it with that provides a similar implementation. 
Your suggestion
of CartesianCoordinates just provides the answer! I will try it and get 
back with
feedback.

Regards,
Dumsani

On 17/12/2015 12:55, Ian Hinder wrote:
>
> On 14 Dec 2015, at 17:14, dumsani <g14n8326 at campus.ru.ac.za 
> <mailto:g14n8326 at campus.ru.ac.za>> wrote:
>
>> Hi All,
>>
>> I am working with some piece of code based on a multipatch coordinate 
>> system
>> (llama, http://dx.doi.org/10.1103/PhysRevD.83.044045). Currently, I'm 
>> interested
>> in doing a test in a cubical kind of domain, that is, without the 
>> multipatch system.
>> Closely following the qc0-mclachlan.par parameter file (ET_2014_05), 
>> I converted
>> my multipatch parameter file to use the following:
>> Carpet::domain_from_multipatch          = "no"
>> Carpet::domain_from_coordbase         = yes
>> CartGrid3D::type                        = "coordbase"
>>
>> Unlike when using a multipatch sysytem, I've had to add the 
>> ReflectionSymmetry
>> thorn in my thornlist for running on a cubical grid. The simulation 
>> is able to build
>> successfully. But when running, it breaks with the following error 
>> from Carpet:
>> cactus/arrangements/Carpet/Carpet/src/SetupGH.cc 
>> <http://SetupGH.cc>:1744: void Carpet::get_domain_specification(const 
>> cGH*, int, const ivect&, rvect&, rvect&, rvect&): Assertion `not 
>> CCTK_IsFunctionAliased ("MultiPatch_GetDomainSpecification")' failed.
>>
>> Looking at the function that fails and a few ideas I got from the 
>> user mailing list, my naive
>> thought is that this might possibly be an issue of cctk_lsh vs 
>> cctk_ash in Carpet.
>> (e.g. 
>> http://lists.einsteintoolkit.org/pipermail/users/2014-October/003849.html). 
>> But I'm yet
>> to test this idea. The code I'm working with uses cctk_lsh.
>>
>> I would gladly appreciate help from anyone who could have an idea 
>> about this.
>>
>> I'm using ET_2014_05 release of the EinsteinToolkit/Cactus. I have 
>> attached my test
>> parameter file for viewing in case it will be useful in diagnosing 
>> the problem(s).
>>
>> Thank you in advance for your kind assistance.
>
> Hi Dumsani,
>
> The relevant lines in Carpet, in SetupGH.c/get_domain_specification, 
> require that if you set domain_from_coordbase, then you must not have 
> the function MultiPatch_GetDomainSpecification available in an active 
> thorn.  In other words, you can't activate a thorn which provides 
> multipatch at the same time as getting the domain from CoordBase.  You 
> have the Coordinates thorn active, which provides multipatch (or more 
> specifically, it provides the aliased 
> function MultiPatch_GetDomainSpecification), hence this error.  You 
> will need to remove Coordinates from your ActiveThorns.
>
> At that point, it's possible that CTGamma will complain, because I 
> think it requires some basic multipatch functionality.  The thorn 
> CTGamma/CartesianCoordinates is provided for this case, so if you run 
> into problems after deactivating Coordinates, try activating 
> CartesianCoordinates.
>
> I attach an old parameter file which uses CTGamma on a Cartesian grid. 
>  I have not run this with the current version of the code, but it used 
> to work at some point!
>
> -- 
> Ian Hinder
> http://members.aei.mpg.de/ianhin
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.einsteintoolkit.org/pipermail/users/attachments/20151217/c756a4d2/attachment.html