How to run Savu

To run Savu you require a data file (e.g. an hdf5 file) and a process list (a link to process list). After Savu is successfully installed into your conda environment you will have an access to the following commands bellow from your UNIX shell:

Alias	Description	Required input parameters
savu_config	Create or amend process lists
savu	Run single threaded Savu	<data_path> <process_list_path> <output_path>
savu_mpijob_local	Run multi-threaded Savu on your PC	<data_path> <process_list_path> <output_path>
savu_mpi	Run mpi Savu across the cluster	<data_path> <process_list_path> <output_path>
savu_mpi_preview	Run mpi Savu across 1 node	<data_path> <process_list_path> <output_path>

Optional arguments:

short	long	argument	Description
-f	–folder	folder_name	Override the output folder name
-d	–tmp	path_to_folder	Store intermediate files in this (temp) directory
-l	–log	path_to_folder	Store log files in this directory
-v, -q	–verbose, –quiet		Verbosity of output log messages

If Savu has been installed into the module system as at Diamond Light source, then you can enable it with:

>>> module load savu

To run Savu on your local machine (single threaded):

>>> savu  <data_path>  <process_list_path>  <output_folder>  <optional_args>

or multi-threaded:

>>> savu_mpijob_local  <data_path>  <process_list_path>  <output_folder>  <optional_args>

and to run Savu across the cluster (in parallel):

>>> savu_mpi  <data_path>  <process_list_path>  <output_folder>  <optional_args>

Note

Savu produces a hdf5 file for each plugin in the process list. It is recommended, if you are running Savu on a full dataset, to pass the optional argument -d <tmp_dir> where tmp_dir is the temporary directory for a visit.