Molcas Forum

MarcoMarazzi

Member

Registered: 2021-04-27

Posts: 5

Running calculations with OpenMolcas/QCMaquis

Hi all,

After some days trying to install OpenMolcas with the QCMAquis interface, it finally compiled successfully.
Nevertheless, when I try to run a NEVPT2 calculation (therefore in principle using Molcas/QCMaquis), I receive always the following message in the output, when attempting to enter the first module from the input file:

.########################.
.# Non-zero return code #.
.########################.

Any ideas?

thanks,
Marco

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Running calculations with OpenMolcas/QCMaquis

Does it happen only for NEVPT2? Have you run the verification ("pymolcas verify qcmaquis")?

MarcoMarazzi

Member

Registered: 2021-04-27

Posts: 5

Re: Running calculations with OpenMolcas/QCMaquis

Hi,

it happens all the time, with OpenMolcas alone, with OpenMolcas/Tinker (also installed), and with OpenMolcas/QCMaquis.

We have also tried to install OpenMolcas/QCMaquis as root or as user, obtaining the same output trouble.
Also, for both installations (as root ans as user), the necessary script $MOLCAS/qcmaquis/bin/qcmaquis.sh was not generated,
although they give different messages when running the verification "pymolcas verify qcmaquis":

In the case of the user installed version, we obtain:

[build]$ pymolcas verify qcmaquis
qcmaquis:007 Failed! (dmrgscf)
************************************************************************
A total of 1 test(s) failed, with 1 critical failure(s).
************************************************************************
Please check the directory:
  test/failed
for the .out/.err files of the failed tests,
and check the submit directory:
  test/tmp
for the working directories of the last run.

While for the root installed version, we get:

[root@nodo00 build]# pymolcas verify qcmaquis
qcmaquis:001 Skipped!
qcmaquis:002 Skipped!
qcmaquis:003 Skipped!
qcmaquis:004 Skipped!
qcmaquis:005 Skipped!
qcmaquis:006 Skipped!
qcmaquis:007 Skipped!
qcmaquis:009 Skipped!
Verification has been completed

Of course, also the NEVPT2 verification then gives error for both installations:

[build]$ pymolcas verify nevpt2
non-existing group: nevpt2
no tests requested, bye!

Concerning the installation:
we have installed the OpenMolcas/QCMaquis program (and NOT the standalone QCMaquis program), by running the command:

FC=gfortran CC=gcc CXX=g++ cmake -DBoost_NO_SYSTEM_PATHS=TRUE -DBOOST_ROOT=/opt/boost-1.58/boost1.58 -DNEVPT2=ON -DDMRG=ON -DLINALG=MKL ../OpenMolcas/

The specs of our system are the following:
OS: CentOS 6.9
Compilers: GNU 8.3.1 devtoolset-8
hdf5 libs: hdf5-1.8.15-patch1-linux-centos6-x86_64-shared.tar
cmake version:  3.19.0-rc3
Python: version 2.x and 3.x where x > or = 7,  including NumPy y SciPy
Boost: 1.58
MKL libs: composer_xe_2013.0.079

Thanks,
Marco

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Running calculations with OpenMolcas/QCMaquis

Also, for both installations (as root ans as user), the necessary script $MOLCAS/qcmaquis/bin/qcmaquis.sh was not generated

I believe that file is not necessary with the new QCMaquis release.

[build]$ pymolcas verify qcmaquis
qcmaquis:007 Failed! (dmrgscf)
************************************************************************
A total of 1 test(s) failed, with 1 critical failure(s).
************************************************************************

That looks "fine". Apparently only 007 failed, but 009 (which is a NEVPT2 test) succeeded. You'd need to look in the *.err/*.out files to find out the reason for the failure.

[root@nodo00 build]# pymolcas verify qcmaquis
qcmaquis:001 Skipped!
qcmaquis:002 Skipped!
qcmaquis:003 Skipped!
qcmaquis:004 Skipped!
qcmaquis:005 Skipped!
qcmaquis:006 Skipped!
qcmaquis:007 Skipped!
qcmaquis:009 Skipped!
Verification has been completed

That happens if qcmaquis support is not compiled. Are you sure you're using the right installation? Double-check the MOLCAS environment variable, and run a bare "pymolcas", it will tell you at the top which installation it will be using.

I'd recommend "pymolcas verify .all" right after "make" finishes, from the same directory, without running "make install"

[build]$ pymolcas verify nevpt2
non-existing group: nevpt2
no tests requested, bye!

That's never supposed to work, there's no "nevpt2" group (i.e., directory inside test/).

MarcoMarazzi

Member

Registered: 2021-04-27

Posts: 5

Re: Running calculations with OpenMolcas/QCMaquis

OK, we'll check that issues.
So, the standalone version of QCMaquis is not necessary?

Thanks,
Marco

MarcoMarazzi

Member

Registered: 2021-04-27

Posts: 5

Re: Running calculations with OpenMolcas/QCMaquis

Hello,

here is the result of "pymolcas verify .all":

[build]$ pymolcas verify .all
standard:013 Failed! (false)
extra:072 Skipped!
extra:398 Skipped!
extra:835 Skipped!
extra:850 Skipped!
extra:851 Skipped!
extra:861 Skipped!
grayzone:832 Failed! (scf)
grayzone:834 Skipped!
************************************************************************
A total of 2 test(s) failed, with 1 critical failure(s).
************************************************************************
Please check the directory:
  test/failed
for the .out/.err files of the failed tests,
and check the submit directory:
  test/tmp
for the working directories of the last run.

Looking at the .out of standard:013 test, we get:

Checking results:
All values match the reference
 
*** symbolic link created: INPORB -> standard__013.GssOrb
--- Start Module: false at Fri Apr 30 09:47:46 2021 ---
 
-- Input validation for false
lxml python module not found, no validation possible
--
The total execution time is limited to 600 seconds.
 
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
 
                                               &FALSE
 
                                   only a single process is used
                       available to each process: 2.0 GB of memory, 1 thread?
                                             pid: 27715
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
 
Command to run:  python3 LJ.py
First argument:  /home/alain/OpenMolcasv21QMC/build/test/tmp/standard__013/standard__013.false.in
Second argument: /home/alain/OpenMolcasv21QMC/build/test/tmp/standard__013/standard__013.false.out
 
++ External program output
--
 
--- Stop Module: false at Fri Apr 30 09:47:46 2021 /rc=-6 ---
 
>>> END DO
 
.########################.
.# Non-zero return code #.

While here ¡s the .err of the same standard:013 test:

File "LJ.py", line 83
    f.write(f'{E}\n')
                   ^
SyntaxError: invalid syntax
 
Program aborted. Backtrace:
#0  0x7f8d1ed11cdd in ???
#1  0x7f8d1ed12807 in ???
#2  0x7f8d1eeb77a8 in ???
#3  0x44720a in ???
#4  0x41a8ae in ???
#5  0x40287a in ???
#6  0x315601ed1f in ???
#7  0x4028ac in ???

Installation and environment variable are fine.

Thanks,
Marco

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Running calculations with OpenMolcas/QCMaquis

I see nothing wrong.

standard:013 uses FALSE (see https://molcas.gitlab.io/OpenMolcas/sph … alse.html), which requires to define an external program. The test includes a python script for that, and apparently your python version is too old or too new. That's just a limitation of the test.

grayzone tests are known to fail in some environments, so grayzone:832 is effectively ignored.

The skipped tests are missing some feature (e.g. tinker or chemps2).

But ".all" does not include "qcmaquis". It's confusing, I know, try running "pymolcas verify --groups" to see a summary.

MarcoMarazzi

Member

Registered: 2021-04-27

Posts: 5

Re: Running calculations with OpenMolcas/QCMaquis

Hi,

after running several tests it seems a local library trouble: the NEVPT2 calculation runs without problems on the front of the cluster, but it does not recognize some gfortran libraries when running on a different machine (where the queuing system sends the calculation). This trouble happens only for OpenMolcas/QCMaquis calculations, while everything is fine for OpenMolcas and OpenMolcas/Tinker calculations. In any case, it is a problem related to our cluster, and not to OpenMolcas/QCMaquis.

Thank you for you help,
Marco