Custom Code Generation

The PIConGPU code generation works by encoding all parameters into a JSON representation, which is combined with a template to render the final configuration.

You can supply a custom template to modify the generated code, e.g. to enable additional output.

Note

It is even possible to use the JSON data during the generation. To get started read the explanation of the translation process, and especially the section on the templating engine *Mustache*.

To see an example of the JSON data that is used either generate an input set with PICMI and examine the generated pypicongpu.json or simply have a look at the example in the documentation.

Step-by-Step Guide

Create a copy of the template: cp -r $PICSRC/share/picongpu/pypicongpu/template my_template_dir

Change whatever you need in the template

e.g. vim my_template_dir/etc/picongpu/N.cfg.mustache

find pypicongpu_output_with_newlines and insert:

--openPMD.period 10
--openPMD.file simData
--openPMD.ext bp
--checkpoint.backend openPMD
--checkpoint.period 100
--checkpoint.restart.backend openPMD

supply your template dir in your PICMI script:

# other setup...
sim = picmi.Simulation(
    time_step_size=9.65531e-14,
    max_steps=1024,
    solver=solver,
    # sets custom template dir
    picongpu_template_dir="my_template_dir")

sim.write_input_file("generated_input")

run PICMI script
inspect generated files

e.g. less generated_input/etc/picongpu/N.cfg now contains the output added above

Warning

It is highly discouraged to incorporate editing generated PIConGPU input files after generation – just because it is very easy to make mistakes this way. Try to use the process outlined here.