error when I change the computer (cogz)

[sylvain_serra]

Dear all,

I had installed code_saturne2.0.4 on my new computer (without problem) and I tried to run a simulation that already works on our cluster.

It uses the cogz d3p specifics physics.

I don't use the GUI and take the same subroutine.

The only differencies are the computer and the runcase (that seems to be ok...)

Here is the error I obtain on my terminal. there is no message in the listing.

*********************************

At line 367 of file ../../../ncs-2.0.4/src/cogz/colecd.f90
Fortran runtime error: Index '0' of dimension 1 of array 'compog' below lower bound of 1
Error running the calculation.

Check Kernel log (listing) and error* files for details

*********************************

is somebody could help me?

thank you very much

[Yvan Fournier]

Hello,

You are probably using a different compiler, and have Fortran bounds checking enabled (which is very useful for debugging builds, as it helps find bugs, but is slower for production code).

Is the code built in debug mode ? There have been several fixes relative to bounds checking issues in version 2.0.6 relative to the Rij turbulence model, but not for combustion to my knowledge.

Although I usually recommend running a case on debug builds before production builds to test for bugs, in this case, running with bounds checking disabled is probably the easiest solution.

If you are running a non-debug build, then if your case is small, I can give it a look if you post it, but probably not before 2 weeks. We are preparing to branch 3.0, which is a long-term maintenance release and will be validated more extensively, expecting release next January, so we will be very busy in the near future, and version 2.0 will become "low priority" relative to 3.0.

Best regards,

Yvan

[sylvain_serra]

Thank you Yvan.

Of course, I have another question, how can I disable the bound checking?

I use gfortran 4.6

Thank you very much

Sylvain

[Yvan Fournier]

Hello,

Was the code configured using --enable-debug ?

If this is not the case, bounds checking is probably already disabled, in which case a small test case would be useful (I suspect that on some processors, the number of boundary faces is 0, so arrays of size nfabor are size 0, and bounds checking may case issues with multidimensional arrays having one component in nfabor).

Best regards,

Yvan

[sylvain_serra]

Hello,
I tried with and without --enable-debug and the problem was the same.
The first test I done was on a single processor so it had all the boundaries.

Finally, I added for the configuration of ncs the flag :
FCFLAGS="-O2"

and now it works.

Thank you very much Yvan.

Best regards,

Sylvain