[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [h5md-user] Offset

From: Olaf Lenz
Subject: Re: [h5md-user] Offset
Date: Tue, 10 Sep 2013 19:03:42 +0200
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130811 Thunderbird/17.0.8

On 09/10/2013 06:04 PM, Pierre de Buyl wrote:
> Konrad Hinsen <address@hidden> a écrit :
>> While working on periodic boxes, I hit a problem: I don't 
>> understand what the "offset" value is supposed to contain, nor
>> even why it exists. The periodicity of a system in space is fully 
>> described by three lattice vectors, which are stored in "edges".
> Some people work with [-L/2:L/2], others with [0:L].
> offset is there so that users may dump their position data without 
> operation between memory and file, whatever box convention they use.

I was always wondering about the offset, but I didn't want to start that
discussion. Thanks, Konrad. :-/

The offset is mathematically irrelevant.

I also do not see what the offset is needed for in the file. When
someone writes a h5md file, he will simply output the coordinates where
he has stored them, irrespective whether they are folded or not, and in
what image. The offset is irrelevant.
When the file is read, the reading application will have to fold the
coordinates into the image where it needs them, insofar it needs that at
all. An offset written in the file would not be used.

The only case that I can think of where the offset might be useful is,
when it is somehow guaranteed that the positions are all within the
primary image that is defined by the offset and the edges.
In that case a reader might gain a little performance as he will not
have to fold the coordinates into the primary image.
However, this guarantee is not in the specs.

Therefore, from my point of view, the offset is useless and can be
removed from the specs.


Dr. rer. nat. Olaf Lenz
Institut für Computerphysik, Allmandring 3, D-70569 Stuttgart
Phone: +49-711-685-63607

Attachment: olenz.vcf
Description: Vcard

reply via email to

[Prev in Thread] Current Thread [Next in Thread]