Re: [h5md-user] Strings again / parallel issue

[Pierre de Buyl]

On Tue, Feb 04, 2014 at 10:38:55AM -0500, Peter Colberg wrote:
> Hi Pierre,
> 
> On Tue, Feb 04, 2014 at 04:05:26PM +0100, Pierre de Buyl wrote:
> > I just discovered a serious annoyance with HDF5 variable-length types. You
> > cannot use them in parallel [1].
> > 
> > This may have slipped through because nobody is doing parallel H5MD right 
> > now,
> > however for large simulations in parallel (with existing code like ESPResSo 
> > or
> > ESPResSo++) this will become a problem really fast.
> > 
> > I became interested in this because I was toying with topologies (not yet 
> > part
> > of H5MD so no problem for now) but all the strings in H5MD are 
> > variable-length
> > currently. I thus started to check the behaviour for our string data.
> 
> Could you describe the data layout of your topology?

In another thread if you wish :-)

> Maybe you could use an enumeration for the datasets in question?
> 
> The variable-length string type is currently used with attributes,
> which cannot be written in parallel. So I would not expect any
> problem for the H5MD specification.

Well, the code that I attach (for real now) writes an attribute in parallel. The
operation is collective (as it must).

Best,

Pierre

> > I attach parallel_string_attribute.f90 (fixed length attribute, works in
> > parallel) and parallel_vl_string_attribute.f90 (vl strings, doesn't work as 
> > far
> > as I can tell).
> 
> The attachments are missing ;-).

parallel_string_attribute.f90
Description: Text document

parallel_vl_string_attribute.f90
Description: Text document