Hi Peter. The codes indeed needs to be re-written. However, in the cluster I would need to assign a compiler such as gcc, gfortran ... Because I would need to convert the new code from my chosen syntax  (e.g C#, C++, F90.....) into the computer language.

Do you believe I could include in my .sh file a few lines at the end to generate the new newcode.exe and re-submit the job? But I may need to request new walltimes after the new code is re-written and recompiled using any "translater" (e.g. gcc).

Maybe the code should be re-written on my end, and then transferred to the cluster every a couple of days using a chronos, and then converted into machine language in the cluster's end, because I am worried about not using the same compiler. Would this be a good path to follow?

Thanks a lot.

Fernando 

Hi Peter. That is sort of the path I am trying to achieve. 

If the code can recompile maybe once a day for a week that would be great. Then I could check the minor improvements over the weekend.

I am already using the make utility to compile two hydrodynamic models (SLIM and FVCOM) written in C++ and F90, repectively. However, I am now trying a piece of code which differs from the hydrodynamic models that I am often using for research in oceanography. Probably, I should  stick to make utility to bring all pieces together as you suggested.

I thought I could have issues in PBS, which is requesting me to inform the wall-time. However, the maximum allowed walltime is more than enough. I was just checking that.

Thanks again :). 

Hopefully I can find some light at the end.

Hi Peter,

My "clumsy question" was indeed missing detailed information. The code would consider an external source for the input data, and this input source is expected to be different every day, and not be an output from the code. Therefore, it differs from the classical problem that you mentioned (explained by Turing), which indeed cannot be solved. Thanks for bringing that up. It showed me that I must always elaborate my questions, and re-read it a bit later in order to check if no information is missing. Moreover, I need to make sure my question is not misleading the reader, because I would then obtain an answer that differs from what I need. 

I have indeed missed informing that the input data would not be sourced as an outcome from the code.

Thanks again.

Fernando