[FLASH-USERS] IO_write/readUserArray
Dean Townsley
Dean.M.Townsley at ua.edu
Tue Mar 6 17:23:57 EST 2012
Hi Massimo,
I'm not sure if anyone has answered your question off-list...
I think that the short answer is that I don't think that those functions
are really intended for broad usage. They were created for a specific
usage case and I think were never properly generalized, and I'm pretty
sure (someone at the flash center can confirm) they are not maintained
for general usage.
I have used the StirTurb setup on a fair number of processors
(hundreds), though not recently (pre flash-4), so they did work with
parallelio at some point. But in my experience installations of
parallel hdf5 are very non-uniform and odd interactions with particular
usages in the code and any given installation are not uncommon. So it
wouldn't be too surprising to me if it worked on one cluster and not
another.
My very vague recollection is that these functions are not cleanly
implemented to work in both serial and parallel hdf5, but I could be
wrong. I just gave up and used parallelio.
I expect that nobody would mind if you wanted to tidy these up a bit. :)
I'm not sure if that helps you, but at least it is some info.
Dean
On 03/02/2012 10:14 AM, Massimo Gaspari wrote:
> Hi guys,
>
>
> I have taken as template IO_writeUserArray.F90 and IO_readUserArray.F90
> of the StirTurb unit. I have 2 problems.
>
>
> 1) They seem to work well with IO hdf5 parallel on a local small cluster.
> However on Pleiades NAS cluster, the code hangs for several minutes
> during the creation of the *chk file.
> This problem drastically increases with the number of processors. Any
> hint or solution for HPC systems?
>
>
> 2) In order to avoid parallel IO, I want to use serial IO.
> However, IO_writeUserArray.F90 and IO_readUserArray.F90
> (and the call to io_h5write/read_generic_*) crashes with HDF5 error:
> Invalid Argument to Routine.
>
>
> Have you ever tried to use IO_write/readUserArray for the standard
> serial IO?
>
>
> Thanks in advance.
>
>
> Max
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20120306/d6500796/attachment-0001.htm>
More information about the flash-users
mailing list