open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,
I am running Telemac3d v6p3r1 in parallel mode on a Windows box, simulating a simple tidal embayment with one open boundary.
Starting from zero velocities I spin up the simulation for a few days and produce a restart file. Then I continue the simulation for another few days using the restart file. All behaves as expected.
BUT, if in the second simulation I tell it to create another restart file it fails at startup. The stop appears to occur when it tries to read the restart file created in the previous simulation :
TITLE OF PREVIOUS COMPUTATION: Spinup

NAME: ELEVATION Z UNIT: M
NAME: VELOCITY U UNIT: M/S
NAME: VELOCITY V UNIT: M/S
NAME: VELOCITY W UNIT: M/S
NAME: U ADVECTION UNIT: M/S
NAME: V ADVECTION UNIT: M/S
NAME: W ADVECTION UNIT: M/S
NAME: DM1 UNIT:
NAME: DHHN UNIT: M
NAME: UCONVC UNIT: M/S
NAME: VCONVC UNIT: M/S
NAME: UD UNIT: M/S
NAME: VD UNIT: M/S
NAME: WD UNIT: M/S

SUITE_SERAFIN : READ OF RECORD 1

TIME OF RECORD: 172800.0 S


The relevant parts of the cas files are :
spinup cas file :

RESTART MODE = YES
RESTART FILE = 'day2.slf'

the continuation cas file

COMPUTATION CONTINUED = YES
PREVIOUS COMPUTATION FILE = 'day2.slf'
PREVIOUS COMPUTATION FILE FORMAT = SERAFIND
....

RESTART MODE = YES
RESTART FILE = 'day5.slf'

Any advice on how to fix this would be very welcome. I have attached the two cas files.
Thanks
James

Hello,

I do not see what the error is in your listing, so it is difficult to say, I would just suspect some mismatch between simple and double precision SERAFIN format files. Otherwise I would need the files of the case that crashes to see what happens.

Anyway this restart mode is not mandatory, you can also remove it and use simple continuations (COMPUTATION CONTINUED = YES and PREVIOUS COMPUTATION FILE=..., also in double precision if you want).
The restart mode is to get continuations without any difference in digits (compared to a longer run without restart), it was asked by researchers that work on data assimilation and in that case even the initial guess for solvers must be stored in theory, and all the operations must be done in the same order, we have to store the depth, because retrieving it from the mesh creates truncation errors, etc.).

With best regards,

Jean-Michel Hervouet

Thanks for your prompt reply.
My apologies, I should have read the documentation more carefully - I assumed that using the RESTART MODE keyword was the only way to start a simulation from previous results.
Using the previous standard results file instead, as you suggested, does exactly what I want.
Regards
James