open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello

I have downloaded and installed the development version of TOMAWAC (trunk). I tried running a test case that works in version 6.0 in scalar but I obtain the following errors from proxim.f:

'SPECTRUM OR SOURCE POINT 1 OUTSIDE DOMAIN'

'NOT ALLOWED IN SCALAR MODE?'

I have attached the steering file, mesh and boundary conditions can anyone tell me what is wrong/ how to fix it please?

Thanks

Fay

Hello,

This should be allowed in scalar mode, I'll check what I did, but you can probably bypass the problem by giving the exact digits, as given by the listing. Actually in parallel we cannot allow a tolerance because otherwise two sub-domains could decide that they have the same point.

With best regards,

Jean-Michel Hervouet

Sorry Jean-Michel I don't understand your reply: when you say give the exact digits what do you mean / where? And what is the listing?

Thanks

Fay

Hello,

OK, I was talking about the keywords :

ABSCISSAE OF SPECTRUM PRINTOUT POINTS

and

ORDINATES OF SPECTRUM PRINTOUT POINTS


but actually you don't have them in your keywords (all my test cases have them), so I guess it is perhaps a loop from 1 to 0 that causes the problem. I'll secure this but in the while you can take subroutine proxim.f in your Fortran file and remove the unwanted stops, or put all the loop 10 between a :

IF(NP.GT.0) THEN
DO 10....
...
10 CONTINUE
ENDIF

I hope it works, I may not be able to test it today.

With best regards,

Jean-Michel Hervouet

Thanks Jean-Michel

I commented out DO 10 .. to 10 CONTINUE and it runs.

Cheers

Fay

Hello,

The definitive correction would be to put IF(NPLEO.GT.0) THEN ... ENDIF around the call to PROXIM in WAC.f, I did this but it still did not work.
Actually there is another thing to correct : the dictionary of Tomawac gives by default 10 SPECTRUM PRINTOUT POINTS with coordinates = 0. (keywords ABSCISSAE OF SPECTRUM PRINTOUT POINTS and following, this is no longer possible as the subroutine PROXIM may complain that they are not in the domain. The default values of abscissae and ordinates must be removed. In fact I discovered that test cases had points with given values that were not 0. and that were out of the domain, so, independently of parallelism, I think it is better to secure this.

Regards,

Jean-Michel Hervouet

Hi

I can now run the development version in scalar but not in parallel. I get a segmentation fault when running in parallel in the following location:

*** ALLOCATION OF USER FILES ***

- STEERING FILE : newtest.casw
- DICTIONARY : tomawacv6p2.dico
- GEOMETRY FILE : big_test_mesh.slf
forrtl: severe (174): SIGSEGV, segmentation fault occurred

Would somebody else with this version of TOMAWAC please try running my test case and let me know if they get the same error?

Thanks

Fay

HEllo,

Well received. I'll look at it asap, but have another one to look at before.

Have a good week-end,

Jean-Michel Hervouet

Hello Fay,

Actually it works on my version. I have just commited on the trunk all Tomawac so that you can get this version. I will not change for a while so that things settle a bit, and I'll do other tests in the while.

With best regards,

Jean-Michel