open TELEMAC-MASCARET The mathematically superior suite of solvers

Using v7p2r0 on a large coastal domain, I am able to couple successfully T2D-Tomawac and also T2D-Sisyphe. But 3-way coupling has not yet been successful, with an error starting as follows:
forrtl: severe (257): formatted I/O to unit open for unformatted transfers, unit 17
The 'littoral' test case shows that 3-way coupling can work well.
Can anyone advise of pitfalls of which I should be aware in 3-way coupling?
Thanks, Alan Davies

Hello Alan,

Aren't you retired like me? The common pitfall is to read or write on a file with hardcoded unit number, since in coupling these numbers are floating and dynamically recomputed, and should be parameterised, e.g. T2D_FILES(24)%LU for file 24 of Telemac-2D when there is no coupling. If it is not this we would have to try the case with more traceback options, unless you know how to do it.

With best regards,

Jean-Michel Hervouet

Hello Jean-Michel,
Thanks very much for the helpful steer which, hopefully, will resolve matters.
Best wishes, Alan
p.s. As evidenced by your message, there is no such thing as retirement for people like us. The old issues of sediment transport etc. refuse to lie down ..... !

Hello Alan and Jean-Michel,

Using v7p2r1 on a coastal domain, I have the same kind of problem with the 3-way coupling.
I only modified limwac.f so as to read boundary spectra on formatted file 1. So I'm using the NFO1 integer passed to limwac by wac, and no hard-coded unit number.
I have no problem with the 2-way coupling T2D-TOM, and when I add Sisyphe, I have the following error during TOMAWAC time step, within the call to ECRSPE :
forrtl: severe (47): write to READONLY file, unit 17, file WACFO1
It seems that ECRPSE wants to write the spectra (that I do not ask for) on my user formatted file...
The error happens both in scalar and parallel modes, even if I ask for spectral outputs or not.

Is there any problem with this dynamical allocation of unit number ?

Regards,
Laurent Leballeur

Hello,

So this is a case for Tomawac and Thierry F., if Alan or you can post the case in the Tomawac topic I'm sure he will have a look at it.

With best regards,

Jean-Michel Hervouet

Hello,

So this is a case for Tomawac and Thierry F., if Alan or you can post the case in the Tomawac topic I'm sure he will have a look at it.

With best regards,

Jean-Michel Hervouet

Hello,

A way to get around this problem is to comment out the calls to ECRSPE in wac.F
But it's a poor solution...
I can give my case to Thierry by contacting me at
leballeur"at"actimar.fr
02.90.26.20.41

Thank you for your help!

Regards,
Laurent Leballeur

Dear Laurent,

Thank you for your comments. Our problem with 3-way T2D-Tom-Sis coupling at Bangor using v7p2r0 turned out to be rather different from what you have described. It was resolved by relacing the TPXO data base for generation of tidal boundary conditions:
OPTION FOR TIDAL BOUNDARY CONDITIONS = 1
TIDAL DATA BASE = 2
BINARY DATABASE 1 FOR TIDE = '../TPXO/h_tpxo7.2'
BINARY DATABASE 2 FOR TIDE = '../TPXO/u_tpxo7.2'
MINOR CONSTITUENTS INFERENCE = YES
by the use of bord.f instead. Things have been running well since doing this, albeit with only 'static' waves (height, period and direction specified at the open sea boundary), but including bed roughness feedback via Sisyphe (time- and space-varying ks). The only caveat is that this 3-way coupling has only worked so far on 1 node (16 processors) on HPC Wales. With more than one node, it fails (for the time being!).

All the best, Alan

Thank you for your feedback Alan!

If I well understand, the problem occurred when you were using the 2 binary tidal database files (T2DBB1 and T2DBB2) from which the tide was directly computed within T2D. Do you know where the error occurred ? in T2D or TOM, and in which subroutine ? what was the unit 17 in the error message ?
And now, you get tidal forcing in bord.f by reading a formatted data file (T2DFO1). And no specific file for Tomawac and Sisyphe, and everything is ok.
What is your error message when you ask for more than 16 procs ?

For my configuration, I read wave spectral forcing WACFO1 and tidal constituents from an external database T2DHAR. nothing for Sisyphe.
By commenting out the calls to ECRSPE in wac.F, I can run my configuration on at least 56 procs (HPC Datarmor Ifremer), I don't test with more.

Regards,
Laurent

Dear Laurent,

With the tidal database TPXO being used during a 3-way coupling, the error seems to arise in Tomawac through WACLEO:

|forrtl: severe (257): formatted I/O to unit open for unformatted transfers, unit 17, file WACLEO

So this may mean something going wrong in Subroutine ECRSPE, though I cannot see what exactly unit 17 refers to.

I can't say more than this, I'm afraid. If TPXO is replaced by bord,f the problem goes away!

When I try running with more than 1 node with 16 processors (e.g. 64 processors on 4 nodes), the 3-way coupling, without TPXO, simply runs on until I kill the job (giving no useful error message unfortunately). For this reason I am sticking with 1 node for now.

Hope this casts some light?!

Best wishes, Alan