profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/epourmal/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.
epourmal The HDF Group Champaign, IL USA hdfgroup.org

epourmal/ECP-ST-CAR-PUBLIC 0

The Exascale Computing Project Software Technologies Capability Assessment Report - Public Version

epourmal/hdf5 0

Official HDF5® Library Repository

jhendersonHDF/hdf5 0

Official HDF5® Library Repository

Pull request review commentHDFGroup/hdf5

H5Ztrans: (feature) Improved H5Z_xform_noop() and H5Z_xform_create() …

 test_write_filtered_dataset_interleaved_write(void)     return; } +/*+ * Tests parallel write of transformed and filtered data+ * in the case where only one process is writing to a+ * particular chunk in the operation. Normally, a data+ * transform function will cause the parallel library to+ * break to independent I/O and this isn't allowed when+ * there are filters in the pipeline. However, in this+ * case the parallel library recognizes that the used+ * data transform function "x" is the same as not applying+ * the transform function. Therefore it does not apply+ * the transform function resulting in not breaking to+ * independent I/O.+ *

Thank you! We are currently testing on HPC systems now and will accept the change.

jwsblokland

comment created time in 3 days

PullRequestReviewEvent

issue commentLLNL/H5Z-ZFP

memory errors with H5Z-ZFP on 1.12.0 with free() but not H5MM_xfree()

I don't think there is any calling sequence into public APIs. Maybe the info below could help?

There is a way to find out from libhdf5.settingsfile or libhdf5.a file if memory allocation sanity checks are on or off.

libhdf5.settings file in the installation directory (where libhdf5.a(so) is) will contain at the very end

Features
---------
...
   Memory allocation sanity checks: no
            Function stack tracing: no
                  Use file locking: best-effort
         Strict file format checks: no
      Optimization instrumentation: no

One can use strings command on libhdf5.a file (we jam configuration summary into the library file)

strings libhdf5.a | grep "Memory"
   Memory allocation sanity checks: no  

When sanity checks are not disabled (for example, in the current HDF5 develop branch), one will get

Features
---------
...
   Memory allocation sanity checks: yes
            Function stack tracing: no
                  Use file locking: best-effort
         Strict file format checks: yes
      Optimization instrumentation: no
strings libhdf5.a | grep "Memory"
   Memory allocation sanity checks: yes

Also, installed H5pubconh.h will contain

H5pubconf.h:#define H5_MEMORY_ALLOC_SANITY_CHECK 1

markcmiller86

comment created time in 8 days

Pull request review commentHDFGroup/hdf5

H5Ztrans: (feature) Improved H5Z_xform_noop() and H5Z_xform_create() …

 test_write_filtered_dataset_interleaved_write(void)     return; } +/*+ * Tests parallel write of transformed and filtered data+ * in the case where only one process is writing to a+ * particular chunk in the operation. Normally, a data+ * transform function will cause the parallel library to+ * break to independent I/O and this isn't allowed when+ * there are filters in the pipeline. However, in this+ * case the parallel library recognizes that the used+ * data transform function "x" is the same as not applying+ * the transform function. Therefore it does not apply+ * the transform function resulting in not breaking to+ * independent I/O.+ *

Please add the comment to the RELEASE.txt file (similar to what you have here) explaining the change. It should go under Bug Fixed since HDF5 ... / Library. Thank you!

jwsblokland

comment created time in 8 days

PullRequestReviewEvent

pull request commentHDFGroup/hdf5

Add high-level set/get for `unsigned long long` attributes.

Looks good! One more thing: Could you please update RELEASE.txt file (see section New Features and High-Level API) with the new APIs? We will update HDF5 documentation. Thank you!

alessandrofelder

comment created time in 8 days

PullRequestReviewEvent

pull request commentHDFGroup/hdf5

Adding support for parallel tools based on the 3rd party library mpiFileUtils (libMFU)

@rawarren Maybe I missed it... We need to check in User's Guide in both docx and PDF formats to the h5docs repo https://github.com/HDFGroup/hdf5doc/tree/master/RFCs/HDF5/tools(just create a new directory. Thank you!

rawarren

comment created time in 9 days

issue commentHDFGroup/hdf5

reproduction notice vs COPYING

Thank you for reporting! We will fix the headers.

chombourger

comment created time in 9 days

issue openedPyTables/PyTables

HDF5 User Group meeting in October 2021

Dear PyTables developers and community members,

I would like to invite you to give a talk on PyTables and with HDF5 at the next HDF5 Users Group (HUG) meeting, which will take place October 12-15, 2021.

Sessions will be online daily between 9am and 1:30pm Central Daylight Time. The format for talks is a 15-minute presentation or demo plus 5 minutes for discussion. If you accept this invitation, please submit a short abstract. Please feel free to talk about any other experience with HDF5!

If you have any questions about submission please contact me at epourmal@hdfgroup.org. Best regards, Elena Pourmal on behalf of the HUG organizers

created time in 16 days

PullRequestReviewEvent
PullRequestReviewEvent

issue commentHDFGroup/hdf5

Branch logic

Please see https://github.com/HDFGroup/hdf5/blob/develop/doc/branches-explained.md

Master should be ignored at this time.

woutdenolf

comment created time in a month

pull request commentHDFGroup/hdf5

H5fget_name_f len_trim fix

Please provide testing code that shows the problem before the fix and passes after the fix. Thank you!

jhaiduce

comment created time in 2 months