Installing ITensor with HDF5 Support

HDF5 is a standardized file format for storing numerical data, making it ideal for storing data such as tensors or tensor networks. Benefits of using HDF5 include:

  • binary-portable files, which can be read on any system (Mac, Linux, etc.)
  • HDF5 files have structure similar to a file system, so that objects stored can be retrieved using their name
  • attributes can be associated to stored data, for example, stating which version of code was using to write the data

Steps to Compile ITensor with HDF5 Support

Step 0 of compiling ITensor with HDF5 is to have the HDF5 library available on your system. It may already be available: one way to check is to see if the command h5cc exists and is in your path. If HDF5 is not installed or you are unsure, use your package manager on Linux, or a package manager such as Homebrew on Mac to install it. Alternatively, you can download and install HDF5 from source similar to many other libraries.

Step 1 is then to determine where your HDF5 libraries and header files are located. A very convenient way to do this is to run the command

h5cc -show

Which shows the commands used to compile a program with HDF5 support. By observing the folder names occurring after the "-L" flag in the output, you can determine the prefix where HDF5 is installed. This prefix is the folder just before the part beginning with "/lib/libhdf5...". So if the output includes "/usr/local/lib/libhdf5_hl.a" then the prefix is "/usr/local".

Step 2 is to modify your file, which is used to configure the ITensor compilation process. In, find the line


then uncomment this line (remove the "#" character) and edit the prefix to be the one where HDF5 is installed on your computer.

Step 3 is to compile, or recompile ITensor fully, so that HDF5 support is built throughout the library:

make clean

If the compilation succeeds, you will have HDF5 support within ITensor.

Back to Install
Back to Main