HDF5 Output Format - ALF — Algorithms for Lattice Fermions

Structure of the HDF5 files produced by ALF simulations compiled with the HDF5 flag.

Output Files¶

File	Content
`data.h5`	All observable bins (scalar, equal-time, time-displaced) in a single compressed file
`confout_<N>.h5`	Auxiliary field configuration for MPI rank N (for restarts)
`info`	Human-readable run summary (same as without HDF5)

data.h5 Structure¶

The file is organized into HDF5 groups, one per observable. Each group contains the bin data as datasets. The group name matches the observable name (e.g. Ener_scal, Green_eq, SpinZ_tau).

Simulation parameters from the parameters file are stored as HDF5 attributes on the root group.

Analyzing HDF5 Output¶

$ALF_DIR/Analysis/ana_hdf5.out              # all observables
$ALF_DIR/Analysis/ana_hdf5.out SpinZ_eq     # specific observable

Results are written to a res/ subdirectory.

Reading with Python¶

import h5py

with h5py.File('data.h5', 'r') as f:
    # List all observables
    print(list(f.keys()))

    # Read a dataset
    ener = f['Ener_scal'][:]
    print(ener.shape)  # (NBin, ...)

Compression¶

When ALF is compiled with -DHDF5_ZLIB, output data is compressed using zlib (deflate). This significantly reduces file sizes, especially for large lattices with time-displaced observables. Compression is transparent — tools like h5py decompress automatically.

If HDF5 was built without zlib support, a warning is printed during configuration:

“Warning: HDF5 installed without compression capabilities. The output files will not be compressed!”