<rdiag> error message in dscf

[Thomas_Hofer]

Hello Turbomole users,

I have encountered following error-message while executing dscf in parallel using TURBOMOLE 5.91 on an Opteron architecture.
According to the dscf output the error appears after the 4th step of the iteration:

[...]
<rdiag> error of main call to DSYEVD
dscf ended abnormally

This error is reproducible and I have encountered this problem for two different systems so far.

- Cu(II) plus n H2O (exact number cannot be given); LANL2DZ ECP for Cu, DZP for water atoms, UHF level
- Sulfate plus 33 H2O; DZPD for S and O; DZP for water atoms, RHF level

Switching to Turbomole 5.8 helps to overcome this problem, the systems converge nicely.

Thanks for your help,
Thomas

[uwe]

Hello,

this seems to be a problem with the optimized linear algebra routines from the Lapack package. Strange that those standard routines fail while the Turbomole-own old code from a previous version did work.

This is a problem that you should send to the Turbomole support instead of the user forum.

The optimized linear algebra routines that are provided from Intel (MKL) or AMD (ACML) often have problems with the user limits. Did you check your stack size limit on all nodes as described in http://www.turbo-forum.com/index.php?topic=23.0?

Regards,

Uwe

[abhijit]

I am a beginner in turbomole, Currently, I am using TMolx Interface for processing my Jobs. I am requesting for your kind assistance regarding how to use LANL2DZ function on a Cd based Coordination polymer in Turbomole. Your kind suggestion and help will be highly solicited.