The HDF5 1.10.5 release is now available for download from:
https://www.hdfgroup.org/downloads/hdf5/
The user documentation and source code (only) can be obtained from the Support Portal:
https://portal.hdfgroup.org/display/support
This release contains several important new features and changes:
-
- Support for OpenMPI was added. For known problems and issues please see OpenMPI Build Issues.
- To better support OpenMPI, all MPI-1 API calls were replaced by MPI-2 equivalents.
- Support was added for cross compiling with CMake on HPC machines.
- A new option was added to enable/disable using pread/pwrite in place of read/write in the sec2, log, and core VFDs:
Autotools: --enable-preadwrite
CMake:HDF5_ENABLE_PREADWRITE
- The ability to minimize dataset object headers was added. New APIs (for C, Fortran, and Java) were introduced to support this feature. See:
H5Fget_dset_no_attrs_hint
,H5Fset_dset_no_attrs_hint
,H5Pget_dset_no_attrs_hint
andH5Pset_dset_no_attrs_hint
. - New chunk query functions (for C) were added. See:
H5Dget_chunk_info
,H5Dget_chunk_info_by_coord
, andH5Dget_num_chunks
. - The macros for the
H5Oget_info
andH5Ovisit
APIs were removed and theH5Oget_info*1
andH5Ovisit*1
APIs were copied toH5Oget_info*
andH5Ovisit*
. As an example,H5Oget_info
andH5Oget_info1
are identical in this release. API compatibility for maintenance releases was broken in HDF5 1.10.3 (and 1.10.4) when these macros were introduced, and this change corrects that. - To improve performance, the default behavior changed when reading a contiguous dataset collectively in its entirety by all processes.
- Several bugs were fixed, including CVE vulnerability issues, as well as hanging issues with collective metadata reads during chunked dataset I/O, found in netCDF-4.
For details regarding these and other changes, please see the Release Notes. Also see Software Changes from Release to Release.