open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I am trying to open a slf file using a modified version of bord.f subroutine. Along subroutine bord.f, the file is supposed to be read several times by calling subroutine lit.f. However, every attempt to call lit.f causes the run to crash. Here are the console output, and the subroutine I use for my case:




I first thought the file was corrupted, but it can be opened and its variables seen with Blue Kenue. Moreover, it is correctly copied into the run temporary folder, and can also be queried within the run. Here it is before and after getting copied into the temporary folder.

drive.google.com/open?id=1SkGbq41ssK9L3byqBjFLqfGjCqtoohRC

Any idea of what might cause of this problem?


Regards,

Sebastian

Hi Sebastian,

If I well understand your problem, you're trying to read a serafin file, on a different grid, thanks to the "BINARY DATA FILE 2" (channel T2DBI2) in the bord.f subroutine, and you have set "BINARY DATA FILE 2 FORMAT = 'SERAFIN '" (with a space at the end), I'm suspicious about fidc='STD' in your bord.f subroutine, and bout the channel 2 you have in the printouts, normally it's the channel 25 for this file... If it's well the case, this file is already open previously by the code.

I think your error message comes from this line:
CALL LIT(FRES,WRES,NBVAR,TITRES,2,'I',fidi,fidc,ISTAT)
You should add a space character after 'I' and so replace the previous line by:
CALL LIT(FRES,WRES,NBVAR,TITRES,2,'I ',fidi,fidc,ISTAT)

Another (simpler) way to read a serafin file is to use the subroutines of the hermes module (get_bnd*, get_mesh*, get_data*)
You can define a bief mesh object dedicated to your mother grid, and store the read data on bief objects (also related to your mother grid).

Hope this helps.

Regards,
Laurent

Thanks for your answer Laurent,

It is indeed weird that the file channel be 2, when the regular channel for the binary data file 2 should be 25 (as stated here).

I did try as you suggested to add a space after "I" when calling LIT subroutine but it did not work. What actually worked was adding the command -fconvert=big-endian in my cfg file. Apparently, what caused the model to crash was that it could not open big endian type files. I attach the file in case somebody has a similar problem.



However, now that the model is running I have another issue, it is very cpu demanding. Do you think using get_bnd, get_mesh or get_data, instead of lit would make it simpler and less demanding?

Regards,

Sebastian

Hi Sebastian,

Really strange that you don't have T2DBI2=25. It seems there is something wrong somewhere... Are you in coupling mode ? I also saw you're using the trunk (no-tagged) version of the code, be careful.
The "BINARY DATA FILE 2 FORMAT" should also be "SERAFIN".

It's also strange that without the "big-endian" compilation flag, you didn't have any problem to read other serafin files (like geometry file for instance) or to write results.

The error message was clear though:
At line 70 of file /home/sebastian/telemac/trunk/sources/utils/hermes/lit.F
Fortran runtime error: Substring out of bounds: upper bound (2) of 'type' exceeds string length (1)
At line 70 of lit.F, type(1:2) is compared to 'R4', that causes an exceed in the size of your string argument provided to lit.F which is of length 1. Other call to lit.F in the telemac system are done with an additional space: 'I '.

What do you mean by "very cpu demanding" ?
It's difficult to say if it will improve cpu performances. Can you give an order of computational times and memory resources with and without your modified subroutine bord.f please ?

Regards,
Laurent

Hi,

It is normal that T2DI2 is not equal to 25.
You should useinstead of
This was changed in the latest version of Telemac-Mascaret.

Hermes might indeed be more efficient as it is using stream access to read the file instead of how it is read using lit.

Hope it helps