Sherpa on Stampede

Sherpa 1.4.3 is installed on Stampede, in the directory /work/02130/ponyisi/sherpa-intel. It is compiled with the Intel compiler with MPI enabled, so it is able to run on (in principle) hundreds of cores simultaneously while generating Feynman diagrams and process grids.

To run Sherpa, create a directory and an appropriate run card in it:

Run.dat

# The following generates p p -> t t~ tau+ tau-
(run){
  # ATLAS general parameters
  MASS[6]=172.5
  MASS[23]=91.1876
  MASS[24]=80.399
  WIDTH[23]=2.4952
  WIDTH[24]=2.085
  SIN2THETAW=0.23113
  MAX_PROPER_LIFETIME=10.0
  MI_HANDLER=Amisic
  # Events to generate?
  EVENTS = 1000
  # Output fully showered events in HEPMC format, with this filename prefix
  HEPMC2_GENEVENT_OUTPUT = ttll-hi
  # Output "short" (initial parton level) events in HEPMC format, with this filename prefix
  HEPMC2_GENEVENT_OUTPUT = ttll-hi
  # take tau mass into account
  MASSIVE[15] = 1
}(run)

(beam){
  # beams are protons at 7 TeV each
  BEAM_1 = 2212; BEAM_ENERGY_1 = 7000;
  BEAM_2 = 2212; BEAM_ENERGY_2 = 7000;
}(beam)

(processes){
  # collide light quarks/gluons (93) to top, antitop, tau, antitau, and up to two additional light quarks/gluons
  Process 93 93 -> 6[a] -6[b] 15 -15 93{2};
    # on shell top -> W b
    DecayOS 6[a] -> 5 24;
    DecayOS -6[b] -> -5 -24;
    # matrix element/showering match with CKKW prescription
    CKKW sqr(30/E_CMS)
    # enhance rates for final states with seven and eight partons - events will be weighted
    Enhance_Factor 2.0 {7}
    Enhance_Factor 5.0 {8}
    # Factorization/renormalization scales: loose prescription for high multiplicity states
    Scales LOOSE_METS{MU_F2}{MU_R2} {8}
    # Allow larger error for high multiplicity states
    Integration_Error 0.1 {7,8}
  End process;
}(processes)

(selector){
  # Dilepton mass between 7 GeV and CM energy
  Mass 15 -15 7.0 E_CMS
}(selector)

(mi){
  # Multiple interactions on
  MI_HANDLER = Amisic  # None or Amisic
}(mi)

You can then create a SLURM batch submission file and submit:

batch_script

#!/bin/bash
#SBATCH -J my-process
#SBATCH -o my-process.txt
#SBATCH -n 160
#SBATCH -p normal
#SBATCH -t 24:00:00
# launch on 160 cores, you have 24 hours max for job.  In practice it's a hybrid process (1 process per node, 16 threads per process).
ibrun /work/02130/ponyisi/sherpa-intel/bin/Sherpa MPI_COMBINE_PROCS=16