Molcas Forum

molcas-user

Member

Registered: 2019-06-17

Posts: 6

failed RASSI calculation with memory problem?

Hi,

I am running RASSI calculation, and it always terminated with error information below:

      Basis set specifications :
      Symmetry species         a    b
      Basis functions          351  348

...

()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()

                                               &RASSI

                                   only a single process is used
                       available to each process: 16 GB of memory, 1 thread?
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()

---------------------------------------------------------------------------------------------
  Nr.    Label          Type            Offset          Length     Atime          Address
---------------------------------------------------------------------------------------------
  1     IP_DUM          REAL           8330644             8           2        [0x4ceefa0]
  2     IP_SDUM         SNGL          16661272             4           3        [0x4ceef60]
  3     IP_IDUM         INTE           8330640             8           4        [0x4ceef80]
  4     NILPT           REAL           8330316             8           5        [0x4cee560]
  5     INILPT          INTE           8330688             8           6        [0x4cef100]
  6     JBNUM           INTE                 0  -4043148383658980352           7        [(nil)]
---------------------------------------------------------------------------------------------
Maximal available memory for Molcas = 15999999964
C_GetMem Calling parameters: ('JBNUM  ','ALLO   ','INTE   ',-2,8717978488897403264)
MMA failed to allocate a memory block.
--- Stop Module: rassi at Mon Jun 17 16:34:03 2019 /rc=_RC_MEMORY_ERROR_ ---
*** files: xmldump

I have tried different value for the memory (up to32 GB), but it always failed with same error. I have used the same files and there was no problem in Molcas 8.1 before.

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: failed RASSI calculation with memory problem?

Can you share the input?

molcas-user

Member

Registered: 2019-06-17

Posts: 6

Re: failed RASSI calculation with memory problem?

Hi Ignacio,

The input file is enclosed below:

&GATEWAY
title=cab16
coord=CAB16.xyz
basis = ANO-RCC-VDZ
ricd
&SEWARD

>>> COPY $CurrDir/singlet.gs.1.JobIph JOB001
>>> COPY $CurrDir/singlet.gs.2.JobIph JOB002
>>> COPY $CurrDir/triplet.gs.1.JobIph JOB003
>>> COPY $CurrDir/triplet.gs.2.JobIph JOB004
>>> COPY $CurrDir/quintet.gs.1.JobIph JOB005
>>> COPY $CurrDir/quintet.gs.2.JobIph JOB006
>>> COPY $CurrDir/singlet.1.JobIph JOB007
>>> COPY $CurrDir/singlet.2.JobIph JOB008
>>> COPY $CurrDir/triplet.1.JobIph JOB009
>>> COPY $CurrDir/triplet.2.JobIph JOB010
>>> COPY $CurrDir/quintet.1.JobIph JOB011
>>> COPY $CurrDir/quintet.2.JobIph JOB012
&RASSI
spin

molcas-user

Member

Registered: 2019-06-17

Posts: 6

Re: failed RASSI calculation with memory problem?

Can you share the input?

I have also tried to decrease the number of states in the calculation, in the initial calculations there are 820 states, and I tried calculations with 760, 740 and 730 states, and they all failed. it works when the number down to 721 states, but I am a little confused that the memory scale with the numbers of states.

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: failed RASSI calculation with memory problem?

The memory requirements increase with the number of states, of course, but it looks like your problem is something else, since the reported size is nonsense and that particular matrix should be quite small. It might be that some JOBIPH file is wrong or corrupted. Can you share the full input so we can try to reproduce it?

molcas-user

Member

Registered: 2019-06-17

Posts: 6

Re: failed RASSI calculation with memory problem?

Hi Ignacio,

Thanks very much for your reply, you are right, the problem is something else rather than memory request since I have tried a calculation on a similar complex with more state (larger than 840), and it is working. I will check the files and see if I could find something strange and I will keep you updated.

Have a nice day

The memory requirements increase with the number of states, of course, but it looks like your problem is something else, since the reported size is nonsense and that particular matrix should be quite small. It might be that some JOBIPH file is wrong or corrupted. Can you share the full input so we can try to reproduce it?

molcas-user

Member

Registered: 2019-06-17

Posts: 6

Re: failed RASSI calculation with memory problem?

The memory requirements increase with the number of states, of course, but it looks like your problem is something else, since the reported size is nonsense and that particular matrix should be quite small. It might be that some JOBIPH file is wrong or corrupted. Can you share the full input so we can try to reproduce it?

Found difference in the format of the output file, in the singlet excited state calculations, the format looks like a old fashion in previous version of Molcas:

      Final state energy(ies):
      ------------------------

::    RASSCF root number  1 Total energy:  -3928.47976246
::    RASSCF root number  2 Total energy:  -3928.46881976
::    RASSCF root number  3 Total energy:  -3928.43452932
::    RASSCF root number  4 Total energy:  -3928.42652849
::    RASSCF root number  5 Total energy:  -3928.41286099
::    RASSCF root number  6 Total energy:  -3928.40277617

for other excited state calculations (triplet and quintet), the format is different:

     Final state energy(ies):
      ------------------------

      root     nonrelativistic        mass-velocity       Darwin-contact         relativistic
                   energy                term                  term                 energy
        1       -3928.48383246           0.00000000           0.00000000       -3928.48383246
        2       -3928.47635017           0.00000000           0.00000000       -3928.47635017
        3       -3928.44756926           0.00000000           0.00000000       -3928.44756926
        4       -3928.43921036           0.00000000           0.00000000       -3928.43921036
        5       -3928.42834712           0.00000000           0.00000000       -3928.42834712
        6       -3928.42258187           0.00000000           0.00000000       -3928.42258187

I don't know if it matters, but it is strange that the output format are different as the calculations are submitted at the same time although they are calculated separated. I will rerun some calculations and see if it works.

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: failed RASSI calculation with memory problem?

That's a bug in the printout that has been recently fixed and should not affect the results.

molcas-user

Member

Registered: 2019-06-17

Posts: 6

Re: failed RASSI calculation with memory problem?

That's a bug in the printout that has been recently fixed and should not affect the results.

Thanks, now it is clear!