Hi Werner,

It looks like a bug. I've just entered the issue into our database. 

When you have a chance, could you please try the latest 1.8.8 to confirm that 
the issue is still there?

Thank you!
Elena
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Elena Pourmal  The HDF Group  http://hdfgroup.org   
1800 So. Oak St., Suite 203, Champaign IL 61820
217.531.6112
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~



On Feb 15, 2012, at 7:18 AM, Werner Benger wrote:

> Hi,
> 
> I'm reading a file of size ~83MB consisting out of 6294 datasets,
> converting single precision data in the file to double precision in
> memory, so the memory occupancy would be about 145MB of all data loaded
> itself. However, once reading a dataset with H5Dread() I leave the
> data space identifier open, for further usage in case the same dataset
> needs to be read again later. This leads to some memory overhead of
> 388MB, which is twice as large as the data that is actually used in
> memory. When doing an H5Dclose() on the dataset identifier just after
> the H5Dread() call, this memory overhead does not appear.
> 
> Is there a way to free this evidently HDF5-internal memory that is
> used when reading a dataset? I'd like to keep the dataset identifier
> available for further usage, but this memory overhead per read is
> killing memory performance. Calling
> 
> H5garbage_collect(void)
> 
> doesn't help. I'm using HDF5 version 1.8.4-snap17 , was there any
> change on such memory management behavior in more recent versions?
> 
> 
>       Werner
> 
> -- 
> ___________________________________________________________________________
> Dr. Werner Benger                Visualization Research
> Laboratory for Creative Arts and Technology (LCAT)
> Center for Computation & Technology at Louisiana State University (CCT/LSU)
> 211 Johnston Hall, Baton Rouge, Louisiana 70803
> Tel.: +1 225 578 4809                        Fax.: +1 225 578-5362

_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]
http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org

Reply via email to