Constructs hdf5 caches of correlator data from raw MILC code data.
Currently supports building from allhisq and FNAL-HISQ campaign outputs.

Installation and running

Can be installed from source with pip install .
Main executables: build_allhisq.py and build_fnalhisq.py,
build.yaml - Example input
build.sh - Example driver script

yaml input

src_root - Location of tar files to process.
stage_root - Temporary staging location of tars being processed.
extract_root - Location where extracted data is written (temporary).
ave_root - Location where average correlator data is written (temporary).
h5name - Name/location of target hdf5 cache.
log - Name/location of output log from build process.
T - Time extent of correlator data [integer].
concurrent - Number of parallel processes used to parse data [int]. Set concurrent: 0 to run in serial mode.
nsrc - Number of time sources per configuration [int]. For FNAL-HISQ data set nsrc: 1.
stream - Label of the Monte Carlo stream, e.g. a004740 -> stream: a.
start - Starting configuration number e.g. 4740.
end - Ending configuration number (included).
delta - Spacing between sequential configuration numbers [int].

Notes

The FNAL-HISQ data has data for all time sources in a single text file; the tar files are indexed by a configuration tag only. Set ave_root = extract_root and nsrc: 1 in the input yaml.

The allhisq data is indexed by configuration number and time source. First the numerical data is extracted into extract_root, then it is averaged over time sources and written to ave_root, and finally written to cache.