Usage

Yadmark mimics the same inputs needed to run yadism, namely a theory card, an observable card and also the name of a pdf set.

Both the theory and observable card can be gnerated authomatically from a default: the former with banana, the latter with something similar to generate_observable() provided sandbox.py.

In addition to run yadmark you need to specify the external program you would benchmark against. To do so, you will have to initialise a class of type yadmark.benchmark.runner. In the following section we describe some available runners which are the most useful example.

The minimal setup of the input cards must contain:

minimal theory input runcard

Name	Type	default	description
PTO	int	[required]	order of perturbation theory: 0 = LO, …

minimal observable input runcard

Name	Type	default	description
observable_names	dict	[required]	lsit of SF to be evaluated a the corresponding (x, Q2) ex.: ['F2light': [{x: 0.1, Q2: 90} ], ... ]
interpolation_xgrid	lsit(float)	[required]	the interpolation grid
interpolation_polynomial_degree	int	[required]	polynomial degree of the interpolating function
interpolation_is_log	bool	[required]	use logarithmic interpolation?
prDIS	str	[required]	DIS current: EM, NC, CC
ProjectileDIS	str	[required]	DIS projectile

The output of yadmark will be stored in data/benchmark.db inside a Pandas.DataFrame table. You can then use the navigator app to inspect your database.

Available Runners

In benchmarks/runners we provide a list of established benchmarks

• sandbox.py:

  • it is used to provide the boilerplate needed for a basic run, in order to make a quick run for debugging purpose, but still fully managed and registered by the yadmark machinery and then available in the navigator

• apfel_bench.py:

  • it is used by the corresponding workflow to run the established benchmarks against APFEL. The complete run of this script will benchmark yadism against all the compatible APFEL features.

  • the necessary python bindings are provided by the APFEL itself

• qcdnum_bench.py:

  • it is used by the corresponding workflow to run the established benchmarks against QCDNUM. The complete run of this script will benchmark yadism against all the compatible QCDNUM features.

  • the necessary python bindings are provided by us externally

• xspace_bench_bench.py:

  • it is used by the corresponding workflow to run the established benchmarks against xspace-bench. The complete run of this script will benchmark yadism against all the compatible xspace-bench features.

  • the necessary python bindings are provided by us externally

• apfelpy_bench.py:

  • it is used by the corresponding workflow to run the established benchmarks against APFEL++. The complete run of this script will benchmark yadism against all the compatible APFEL++ features.

  • the necessary python bindings are provided by the APFEL++ itself

All of them are examples useful to understand how to use the yadmark package for benchmarking.