Feedback wanted on new hyperslab selection functions

Hdf-forum archives
1

Hi all,
I've attached an RFC document describing some new API functions that improve how hyperslab selections can be created and operated on in HDF5. Please take a look and let me know what you think.

Also, as a teaser, the changes to implement these in the HDF5 library are showing nice performance improvements for some popular application usage patterns, so they should be good for more than just the new functionality. :slight_smile:

Regards,
Quincey

Microsoft Word - RFC_Enhancement_Hyperslab_Selection-1.4.docx.pdf (329 KB)

ATT00001.htm (396 Bytes)

2

Quincey,

Do you have a place already set up where you are storing this doc? If not, would it be OK if I make in https://svn.hdfgroup.uiuc.edu/hdf5doc/trunk/RFCs/HDF5_Library a directory called something like EnhancedHyperslabSelection for this RFC?

Thanks,

Mark

路路路

From: Hdf-forum [mailto:hdf-forum-bounces@lists.hdfgroup.org] On Behalf Of Quincey Koziol
Sent: Friday, June 13, 2014 10:03 PM
To: Hdf-forum@lists.hdfgroup.org
Subject: [Hdf-forum] Feedback wanted on new hyperslab selection functions

Hi all,
I've attached an RFC document describing some new API functions that improve how hyperslab selections can be created and operated on in HDF5. Please take a look and let me know what you think.

Also, as a teaser, the changes to implement these in the HDF5 library are showing nice performance improvements for some popular application usage patterns, so they should be good for more than just the new functionality. :slight_smile:

Regards,
Quincey

3

Hi all,
I've attached an RFC document describing some new API functions that
improve how hyperslab selections can be created and operated on in HDF5.
聽聽Please take a look and let me know what you think.

API looks fine. I cannot draw to mind a case where I would need such functionality, or where the "dataset copy + hyperslab selection" bit was a significant -- or detectable -- portion of runtime overhead, but there are lots more HDF5 workloads out there than *I* regularly encounter.

some kind of funny extra formatting on top of page 4:
a)
b)

3 Interface

Also, as a teaser, the changes to implement these in the HDF5 library
are showing nice performance improvements for some popular application
usage patterns, so they should be good for more than just the new
functionality. :slight_smile:

you are saying that in order to make hyperslab selection more flexible for these special cases, you've improved performance for the common cases? sorcery!

==rob

路路路

On 06/13/2014 10:02 PM, Quincey Koziol wrote:

--
Rob Latham
Mathematics and Computer Science Division
Argonne National Lab, IL USA