TURBOMOLE v7.6: ricc2_smp calculation produces trash

Installation and usage of TURBOMOLE > Parallel Runs

(1/2) > >>

thda0531:
Hi,

an ADC(2) calculation with cc-pVDZ basis set calculates senseless numbers in PARA_ARCH=SMP
and PARNODES>=2 scheme and continues:

.
.
.
**************************************************************************
* *
* OPTIMIZATION OF THE GROUND STATE CLUSTER AMPLITUDES *
* *
**************************************************************************

Iter. MP2 energy Norm(Omega) Norm(t1) Norm(t2) cpu wall
...........................................................................
1 -1004.728222448856372.6842824962 0.000006583.89878 3.08 0.17
Error in cc_d1_diagnostic:
eomxr,evmxr: 187327269.334471 187327269.334471
eomnr,evmnr: 119874.613636796 -2.495951987227418E-008
eomxi,evmxi: 0.000000000000000E+000 1.087509974514332E-008
...........................................................................
.
.
.

The same job runs properly using TURBOMOLE v7.5.1.
Additionally, v7.6/ricc2_smp and PARNODES=1 runs properly, too!!

Any ideas what the reason could be???
Thanks, Thomas

[Further details can be provided on demand]

uwe:
Hi,

is that a restart case, i.e. are there any restart files in the directory that are taken into account?

I'd also recommend to use the 7.6.1 version and check if the error still shows up there.

thda0531:
Hi,

I'm sorry, but the problems remains with 7.6.1

jamesagreen:
Hi,
I am also having this problem with v7.6.1 - did you ever find a solution?

thda0531:
Hi,

you may try setting this variable:
export MKL_ENABLE_INSTRUCTIONS=SSE4_2

Navigation

[0] Message Index

[#] Next page

Go to full version