[Hdf-forum] question about 1.10 scheduling, swmr concurrency
koziol at hdfgroup.org
Wed Sep 14 08:31:04 CDT 2011
*sigh* I meant this message to only go to Gabriel. _Please_ don't grab the code in the URL below without talking to me - the files it produces _WILL NOT WORK_ with past/current/future versions of HDF5.
On Sep 14, 2011, at 8:26 AM, Quincey Koziol wrote:
> Hi Gabriel,
> On Sep 12, 2011, at 8:46 AM, Gabriel Corneanu wrote:
>> Hi Quincey,
>>> We're still shooting for a beta release of 1.10.0 around November of this year. I can point you at some alpha >quality code, if you are interested in test-driving what's available now and won't keep any of the files produced.
>> Sorry, I was in holiday.
>> Yes, I would really like to get an impression about it... Of course I
>> won't keep data, I just want to prepare my wrappers...
>> If possible, my email is gabriel.corneanu AT gmail DOT com
> OK, you can use SVN to check out a current version of the SWMR code from this URL:
> The main feature currently enabled is that when the "latest" version of the file format is used and a dataset is extended, the writer can do this without coordination with the reader (as long as they share a POSIX compatible file system - which NFS is _not_). Take a look at the "swmr_writer.c" and "swmr_reader.c" code in the test subdirectory for how to enable and leverage this behavior. Eventually, all metadata modifications will be supported for SWMR access, not just extending datasets.
> Hdf-forum is for HDF software users discussion.
> Hdf-forum at hdfgroup.org
More information about the Hdf-forum