open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi Chi-Tuan, (and the rest of the Telemac development team)

There is more to it...

I see more and more people trying to use meshes with sub-meter element sizes in projected coordinate sytems (like PolarStereo, Lambert, etc).

In the Selafin file format specification, the contents of IPARAM(5) has long been defined as a way to indicate a single or double precision file but was poorly documented.

In the latest Telemac User manual (version 6.0) in Appendix 3, Page 1, the use of IPARAM(5) = 1 to indicate a double precision Selafin file is not discussed. (I asked about this sometime ago on the forum but didn't receive an answer.)

So my questions are...

1. Are double precision Selafin files fully supported by all components of the Telemac system?

2. If IPARAM(5) == 1, are all variables double precision or just the coordinates?

3. Should I be modifying Blue Kenue to always write double precision coordinates. Perhaps it could be an option?

4. Alternatively, should we be automatically making use of IPARAM(3) and IPARAM(4). This would also resolve the floating point precision issue.

5. Does FUDAA support IPARAM(3,4,5) ??

Cheers... Martin

P.S. Edited by T2D Moderator
P.S. initially this post was a reply for thread #6523

Hello Martin,

Actually we do use a double precision SERAFIN format in some cases, but to decide whether a file is double precision or not, we look at the title, the strings 73 to 80. It may be 'SERAFIN ' or 'SERAFIND'. This is used e.g. by partel to do domain decomposition. IPARAM(5) is not looked at. In modules like Telemac-2D there is a key-word for saying if a file is single or double. For example GEOMETRY FILE has a "GEOMETRY FILE FORMAT" which may be so far SERAFIN, SERAFIND or MED. You could say we could deduce it from the title but at least when you write a file you must decide single or double. Then on the question of pre- and post-processors, Rubens does not understand double precision for sure, and it was asked for Fudaa but I am not sure it is ready yet.

With best regards,

Jean-Michel

Hi Jean-Michel,

Adding support for double precision data to Blue Kenue is pretty straightforward for me since all internal data is already DP.

Can you comment on points 3&4 in my post...

Lastly, could you post (or send me) an example SERAFIND file so I have something to validate my implementation.

Cheers... Martin