Error at termination of mrcc

I am still playing a bit with MRCC and ran into a case where I get:

Fatal error in mrcc.
Program will stop.

************************ 2014-08-21 09:56:53 *************************
Error at the termination of mrcc.
**********************************************************************

I have attached the output file and MINP.

What did I do wrong here?

Thanks,
René

PS I compiled with Intel -pOMP using intel 11 compilers on a RH 6.5 system.

PPS I just thought of adding the stderr output from SGE to this email, which may be helpful?:

forrtl: Invalid argument
forrtl: severe (39): error during read, unit 21, file /tmp/574.1.anynode.q/fort.21
Image PC Routine Line Source
mrcc 00000000005BD42D Unknown Unknown Unknown
mrcc 00000000005BBF35 Unknown Unknown Unknown
mrcc 0000000000569920 Unknown Unknown Unknown
mrcc 000000000050D41A Unknown Unknown Unknown
mrcc 000000000050CC10 Unknown Unknown Unknown
mrcc 00000000005287B7 Unknown Unknown Unknown
mrcc 0000000000526A08 Unknown Unknown Unknown
mrcc 00000000004F1FA3 Unknown Unknown Unknown
mrcc 00000000004D5F89 Unknown Unknown Unknown
mrcc 00000000004CFA58 Unknown Unknown Unknown
mrcc 00000000004C7A80 Unknown Unknown Unknown
mrcc 0000000000435B84 Unknown Unknown Unknown
mrcc 0000000000409CAF Unknown Unknown Unknown
mrcc 000000000040414C Unknown Unknown Unknown
libc.so.6 0000003C84E1ED1D Unknown Unknown Unknown
mrcc 0000000000404049 Unknown Unknown Unknown

Dear René,
I couldn't reproduce the error, however, I am using 4 cores (for the moment I have no access to a machine with 12 cores). Could you please try to run the job on 4 cores to see if the problem is related to the number of cores or to the compiler version? Thank you in advance.

It failed at the exact same spot. Then I tried without ovirt=MP2 and it still fails (the CC iterations take quite a bit longer) but in that case it gets to Spin case 22 Alpha:1 Beta:2 with 632700 excitations before it fails.

I just did run a regular CCSD calculation without a problem.

René

Then it is probably a compiler issue. Again, I would suggest upgrading to the last Intel version (14.0.2). We invested much work in adapting our codes to the compiler bugs of that version, and the program is rather stable with it.
You can also try the following things:
1) Increase the memory available to the program.
2) Compile the code without OpenMP parallelization. Then, of course, the program will run on one core.
3) Since you are interested in a full CCSDTQ calculation, you don't need to run the CCSD(T) calculation (where the program crashes). That is, delete rest=2 from the input, then only the CCSDTQ calculation will be performed. Another option is to run a CCSDT calculation first, and then to perform the CCSDTQ calculation with rest=1 in the input.

Dear Mihaly,

I really would like to use threading, so I don't like option 2. I think option 1 is out because of cost, but I was wondering whether you think compiling with gnu compilers could work out. Or would it be possible for you to distribute an OMP binary version with statically linked libraries?

Regarding 3) I started to go down that way starting with a CCSD calculation followed by CCSDT and rest=1. Things seem to be going along (still at the CC iterations stage), but I do get the message from xmrcc (for Si4Li isomers that have are C1, I didn't get it for another one that has Cs symmetry):

The number of active orbitals has been changed!
It is dangerous to restart the program!

Just before mrcc kicks in and goes ahead.

By the way, not sure whether that has anything to do with it, but I compiled the mrcc.2014-07-10 version.

Cheers,
René

PS is the output from the program buffered? It seems that it often takes a while before the output file is update (usually with the output for several CC iterations at the same time).

Dear René,
We have created statically linked binaries, please find them in the download area.
I wouldn't recommend the GNU compiler, the GNU-compiled executables are rather slow.
The output is not buffered, the I/O buffers are flushed in each iteration.