ridft probelm of 'actual -r'

[fantastic2012]

I am doing QMMM simulations using gromacs interfaced with turbomole. For the QM methods, I choose B3LYP/DFT with RI approximation.

However, it does not work. I get the clues from rdgrad.log file:


data group $actual step is not empty
due to the abend of ridft


check reason for abend ...

use the command  'actual -r'  to get rid of that

MODTRACE: no modules on stack

  CONTRL dead = actual step


I tried to run "actual -r", and it showed to me this: ridft step seems to have been in serious trouble. This command only replaces "$actual step      ridft"  with "$last step      ridft" in the control file.
But it will still fail if I submit the job again, and the "$last step      ridft" gets back to "$actual step      ridft" in the control file.

How should I deal with such a problem? By the way, how faster is ridft than dft without RI approximation?  Thanks very much!

[Arnim]

Hi,

before re-submitting you should check, what problem in the ridft calculation has occurred.
In your case, rdgrad refuses to run, because the ridft step has aborted for some reason. Maybe just the number of SCF cycles is insufficient.

Cheers,

Arnim

[Chris Bright]

Hi,

before re-submitting you should check, what problem in the ridft calculation has occurred.
In your case, rdgrad refuses to run, because the ridft step has aborted for some reason. Maybe just the number of SCF cycles is insufficient.

Cheers,

Arnim

But is it possible to use RI-JK and MARI-J with B3LYP in parallel mode?
It seems it's not yet implemented.
I don't know if I have an old version or if this is the status-of-art...

Thanks.

Vasco

[Arnim]

Hi,

that is available in the Global Arrays version. Use export PARA_ARCH=SMP.

Regards,

Arnim

[golden]

Hi,

that is available in the Global Arrays version. Use export PARA_ARCH=SMP.

Regards,

Arnim

Wow, I didn't know about that, does it mean RI method can be implemented with any functional now?