[CP2K:1992] Re: Reftraj ensemble

Ondrej Marsalek ondrej.... at gmail.com
Wed Apr 15 12:20:01 UTC 2009


On Wed, Apr 15, 2009 at 13:50, Teodoro Laino <teodor... at gmail.com> wrote:
>
> Or even better.. instead of waiting.. if you already have an idea how to
> patch the indexing behavior just send us the
> patch.. I will be more than happy to apply it to the CVS.

I understand that. Several problems unfortunately. I can't allocate
any time to it at the moment. I would have to get decent knowledge of
the source code, which I don't think I have. It is probably not a
simple file numbering issue, seeing that the cube file data are
actually different and that there are two frames calculated at the
beginning instead of just one, as I would expect. Overall, this looks
to be so deep inside CP2K, that I don't even try.

Hopefully, it will catch the attention of someone with better
knowledge of the code.

On a slightly unrelated note, are there any plans for an issue tracker?

Cheers,
Ondrej


> Regards,
> Teo
>
> Ondrej Marsalek wrote:
>> On Wed, Apr 15, 2009 at 13:18, Teodoro Laino <teodor... at gmail.com> wrote:
>>
>>> Hi Ondrej
>>>
>>>
>>> Just a comment about point (1).
>>> Coordinates are needed to setup the environment. You need them.
>>> Instead of providing the &COORD section you can provide them using the
>>> same xyz file.
>>>
>>
>> Sure, I get that. The point was that in principle they could be obtained
>> from the reftraj without the need for an explicit copy. But that is
>> actually quite insignificant compared to an additional problem:
>>
>> When I do remove the i=0 frame as I said and also want cube file from
>> the run, I get them for different frames than position, and ener file
>> lines. When I want to start at frame 5000 and go with stride 100, I get
>> the correct frame numbers in ener and pos, that is 5000, 5100, 5200, ...
>> The cube file numbers, on the other hand, are: 5000, 5001, 5101, 5201,
>> ... Looking at the coordinates in the cube files, they really seem to be
>> different from the frames in the pos file.
>>
>> Overall, the whole reftraj frame numbering business is very confusing.
>> If anyone is interested, I will create a small testcase that
>> demonstrates this.
>>
>> Ondrej
>>
>>
>>
>>> Cheers
>>> Teo
>>>
>>> ---------------------------------------------
>>> Teodoro Laino
>>> Zurich Switzerland
>>>
>>> Contact info:
>>> Tel.:     http://www.jajah.com/Teo
>>> E-mail: teo... at laino.eu
>>>              teodor... at gmail.com
>>> ---------------------------------------------
>>>
>>> On 15 Apr 2009, at 12:53, Ondrej Marsalek <ondrej.... at gmail.com>
>>> wrote:
>>>
>>>
>>>> Hi,
>>>>
>>>> I am trying to use the reftraj ensemble and there are two things that
>>>> I find slightly inconvenient.
>>>>
>>>> 1) You need to provide coordinates in the input file, although they
>>>> are not needed, as you get them from the reference trajectory. They
>>>> are probably (at least I hope) ignored. If they are, it would be nice
>>>> say that in the input manual at the reftraj entry. It would be even
>>>> better not to require them at all and use whatever is the first frame
>>>> requested instead.
>>>>
>>>> 2) If you want to get consistent step numbers everywhere, you need to
>>>> remove the frame with i=0 from CP2K xyz trajectory output that you use
>>>> as a reference trajectory. Seems to be due to some mixing of
>>>> zero-based and one-based indexing.
>>>>
>>>> As usual with my messages here, nothing critical, but I think it is
>>>> worth mentioning. Either for people trying to make sense of the
>>>> numbers (like me a while ago) or for a developer who is really bored
>>>> ;-)
>>>>
>>>> Best,
>>>> Ondrej
>>>>
>>>>
>>
>> >
>>
>
>
> >
>



More information about the CP2K-user mailing list