Main options

Main options#

The [main] sections is used to select the process and set some basic behaviour. Let’s look at a simple example:

[main]
process = d db -> u ub
batch_size = 100
n_batches = 10
verbosity = debug
random_seed = 1234

Available options are:

  • process: This can one of ppjj, ppee or pptt. Appending the character j \(n\) times will add \(n\) additional final-state jets to the process. Alternatively, a partonic channel can be specified by giving a whitespace-separated list of particles using -> to separate initial- and final-state particles. An example would be d db -> u ub. More than one channel can be specified by giving them as a comma-separated list, e.g. d db -> u ub, d db -> c cb, ...

    • Default: There is no default. This option must be set explicitly, either in the runcard or by using the -p|--process command line argument.

  • batch_size: The number of events processed simultaneously as a so-called batch. This is particularly important for tuning performance on GPU.

    • Default: 1

  • n_batches: The Number of event batches to be generated in the simulation.

    • Default: 0

  • verbosity: The verbosity of Pepper.

    • Default: info

    • Options: error, warn, info, verbose, debug

  • color: The colourfulness of Pepper’s output.

    • Default: automatic

    • Options: automatic, always, never

  • device: The device ID when running on machines equipped with multiple CUDA devices.

    • Default: 0

  • seed: The random seed used in the simulation. (Or rather, the offset from the internally used default random seed.)

    • Default: 0

  • collision_energy: The centre-of-mass collision energy in GeV units.

    • Default: 14000

  • helicity_summing: Set to true to use helicity summing instead of sampling

    • Default: false

  • helicity_integrator: Set to false to disable improved helicity sampling. This Option is ignored when helicities are summed.

    • Default: true

  • helicity_summing_options: Flags that steer the use of symmetries to accelerate helicity summing. Adding the flag c enables caching results of the recursion’s final reduction step, while adding p enables the use of parity symmetry (if possible).

    • Default: cp

    • Options: c, p, cp

  • mu2: The squared scale, setting both the factorisation and renormalisation scales.

    • Default: m_Z^2

    • Options: m_Z^2 (squared Z mass), H_Tp^2 (the squared sum of the transverse Z mass and the transverse jet momenta)

  • fixed_alpha_s: Set to a positive number, e.g. 0.118, to use a fixed strong coupling with that value. If this is not set, a running strong coupling will be determined using the LHAPDF6 library.

    • Default: not set

Specialised batch size options#

The two subsections [main.gpu] and [main.cpu] can be used to set the batch size depending on the device type used, like so:

[main.cpu]
batch_size = 100

[main.gpu]
batch_size = 1024

These batch size settings take precedence over the batch size within [main], if the given device is used. Thus, the same runcard can be used for different architectures.

Contents