[molpro-user] About detailed error message for filled disk space

[Peter Knowles]

The disk space for the 3-external integrals scales as  
number_of_electrons**1 * number_of_basis_functions**3
The minimum memory requirement for (T) scales as  
number_of_basis_functions**3 ; the optimal memory requirement scales  
as the number of 3-external integrals.
Peter
On 6 Dec 2007, at 10:28, Min Seung Kyu wrote:

> Thanks Peter for your reply.
>
> I want to calculate CCSD(T) energies for several conformers.
>
> The basis set size for one of the conformers is following.
>
> --------------------------------------------------------------
> NUMBER OF PRIMITIVE AOS:         883
> NUMBER OF SYMMETRY AOS:          843
> NUMBER OF CONTRACTIONS:          577   ( 577A   )
> NUMBER OF CORE ORBITALS:          20   (  20A   )
> NUMBER OF VALENCE ORBITALS:       93   (  93A   )
> --------------------------------------------------------------
>
> Actually I finnished a calculation for one conformer with slightly  
> smaller size. The basis size for this is
>
> --------------------------------------------------------------
> NUCLEAR CHARGE:                  126
> NUMBER OF PRIMITIVE AOS:         809
> NUMBER OF SYMMETRY AOS:          773
> NUMBER OF CONTRACTIONS:          531   ( 531A   )
> NUMBER OF CORE ORBITALS:          18   (  18A   )
> NUMBER OF VALENCE ORBITALS:       85   (  85A   )
> --------------------------------------------------------------
>
> and it used  218.27 GB disk space for CCSD(T) calculation.
>
> Could you guess required memory for the above case?
>
> Thank you.
>
> Sincerely, Min.
>
> p.s. If you need more informations for that, I can tell you any.
>
> -------------------------------Sig.---------------------------
> Center for Superfunctional Materials, Department of Chemistry,
> Pohang University of Science and Technology (POSTECH)
>
> Mr. Seung Kyu, Min
>
> Ph.D. Candidate
>
> E-Mail: min0220 at postech.ac.kr
> Tel.  : 82-54-279-5858
> ----------------------------------------------
>
>
> On Thu, 6 Dec 2007, Peter Knowles wrote:
>
>> fd=15 means that the error is on file 5, commonly used to store the  
>> transformed 2-external Coulomb integrals.
>> l=30892 is the length of the buffer being written at the failure  
>> point.
>> p=... is the address on the file that you are writing to, in units  
>> of 8-byte words.
>>
>> Unfortunately, most parts of Molpro do not look ahead to calculate  
>> total memory and disk space requirements; furthermore such  
>> calculations are of course not easy to do, and as you don't even  
>> say what kind of method you are using, it's hard to help you with a  
>> prediction.
>>
>> Peter
>>
>> On 6 Dec 2007, at 04:49, Min Seung Kyu wrote:
>>
>>> Dear MOLPRO users,
>>> Could you tell me about this error message in detail?
>>> ------------------------------------------------------------------------------
>>> Write error in iow_direct_write; fd=15, l=30982, p=788964613;  
>>> write returns -1
>>> ------------------------------------------------------------------------------
>>> I know this message is about filled disk quota or space, but I  
>>> want to know the details.
>>> What do fd, l, and p mean?
>>> Are there any messages about required hard disk space for the  
>>> calculation?
>>> Because my quota for scratch directory is limited, I want to know  
>>> the required workspace size exactly.
>>> Thanks alot.
>>> Sincerely, Min.
>>> -------------------------------Sig.---------------------------
>>> Center for Superfunctional Materials, Department of Chemistry,
>>> Pohang University of Science and Technology (POSTECH)
>>> Mr. Seung Kyu / Min
>>> E-Mail: min0220 at postech.ac.kr
>>> Tel.  : 82-54-279-5858
>>> --------------------------------------------------------------
>>
>> --
>> Prof. Peter J. Knowles
>> School of Chemistry, Cardiff University, Main Building, Park Place,  
>> Cardiff CF10 3AT, UK
>> Telephone: +44 29208 79182 Fax: +44 2920874030 Email KnowlesPJ at Cardiff.ac.uk
>> WWW http://www.cardiff.ac.uk/chemy/contactsandpeople/academicstaff/knowles.html
>>
>>
>>
>>
>>
>
>
>
>

--
Prof. Peter J. Knowles
School of Chemistry, Cardiff University, Main Building, Park Place,  
Cardiff CF10 3AT, UK
Telephone: +44 29208 79182 Fax: +44 2920874030 Email KnowlesPJ at Cardiff.ac.uk
WWW http://www.cardiff.ac.uk/chemy/contactsandpeople/academicstaff/knowles.html