After making sure that PHDF5 is installed and available for your use, you are ready to try the example programs that are presented in this topic, which demonstrate:

• Parallel file write
• Parallel file read
• Parallel file operations with hyperslabs

Any large HPC system will generally have parallel HDF5 (PHDF5) preinstalled on it as a supported way of accessing its associated parallel file system. Consequently, all that you typically need to do is load a few relevant environment modules to ensure that the right libraries and header files become available to your application.

For example, on TACC's Frontera, where modules for Intel's compilers and MPI libraries are loaded into your environment by default, adding PHDF5 can be as simple as:

$ module load phdf5

If you happen to be on a machine that lacks PHDF5, yet provides access to a parallel file system like Lustre, the PHDF5 software can be installed for individual use as follows:

1. Download the latest release from https://portal.hdfgroup.org/display/support/Downloads. Note that you can also download a system-specific binary distribution if you do not want to compile from source.
2. Unpack the downloaded tar file:

   $ tar -zxf <file-version>.tar.gz

3. Configure and set the installation directory, while enabling parallel HDF5:

   $ ./configure --prefix=<install dir> --enable-parallel

4. Build:

   $ make

5. Check the build:

   $ make check

   Note that some tests might fail or be skipped. Make sure that the necessary libraries are created after the next step (installation). If they are not created, please check the output of make check and debug the issue.
6. Install:

   $ make install

Note that you need to have an MPI compiler that supports MPI-IO and a POSIX-compliant parallel file system to build PHDF5.