[molpro-user] CI restart problem

[Gabriel Freitas]

Dear Andy,

I tried again with the pre-compiled mpp code (patchlevel 12). Now the job
doesn't crash, but the second CI seems to start from scratch, that it, the
number/time/energies of iterations are the same as the first CI (I expected
it to converge in a single iteration, since the CI guess vector is already
converged). It happens in the closed and open-shell cases.

Am I missing something?

Best regards,

Gabriel

2014-09-16 6:50 GMT-03:00 Andy May <MayAJ1 at cardiff.ac.uk>:

> Gabriel,
>
> I tried locally and your input ran fine with latest 2012. I suspect this
> is the same issue that I've just written to Jacek about; if so please
> update your Molpro install.
>
> Best wishes,
>
> Andy
>
> On 07/09/14 17:06, Gabriel Freitas wrote:
>
>> Dear Molpro users/developers,
>>
>> I'm trying to use a converged CI wavefunction (more specifically, the ci
>> vectors) as a guess for another CI calculation.
>>
>> Even working with a closed-shell, very small system in a same geometry,
>> the program crashes asking for an impossibly large amount of memory.
>>
>> One example below. The same output is given even if I change same
>> options in mrci saving/starting.
>>
>> I appreciate very much any help.
>>
>> Regards,
>>
>> Gabriel
>>
>> *INPUT*
>> /memory,60,M
>> nosym
>> R1=1.34344609 ang,
>> geometry={
>>            H;
>>            S,1,R1;}
>> basis=aug-cc-pV(T+d)Z
>> {rhf;wf,18,1,0;}
>> {mrci;core,5;closed,7;wf,18,1,0;save,5400.1,5401.1,1}
>> {mrci;wf,18,1,0;start,5400.1,1}/
>>
>> *OUTPUT*
>> (after converging first ci)
>> / Wavefunction saved on  5400.1
>>
>>   Internal configurations saved on 5401.1/
>>
>> (second CI, after some lines)
>>
>> / Number of p-space configurations:   1
>>
>>   Overlap between present and previous reference vectors
>>                     1
>>           1     1.0000000
>>   insufficient memory available - require     112652603903961664  have
>>                59999239
>>   the request was for real words/
>>
>>
>>
>>
>> _______________________________________________
>> Molpro-user mailing list
>> Molpro-user at molpro.net
>> http://www.molpro.net/mailman/listinfo/molpro-user
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20140916/5def4a33/attachment.html>