...

<source_library title="Library">
  <source flux="0.667283950775" name="bn131231198">
    <spectrum escale="MeV">
    <particle name="gamma">
      <power_law emax="200000.0" emin="15.0" gamma="2.0" />
    </particle>
    <celestial_dir dec="-2.42" ra="10.11" />
    </spectrum>
  </source>
</source_library>

This is the command line I am using:

gtobssim infile=sim_model.xml srclist=source_list.txt scfile=/home/giacomov/FermiData/bn131231198/gll_ft2_tr_bn131231198_v00.fit evroot=boost simtime=100000 ltfrac=1 tstart=410157919.083 nevents=no maxtime=3.155e8 use_ac=no emin=30 emax=100000 edisp=no irfs=P8_SOURCE_V4 maxrows=100000 seed=653786

These are the execution times with the flux as reported above, and then N "boost" times that value:

This shows that the algorithm is essentially linearly dependent on the number of events (apart from a short initialization timethe inflection between 50 and 100 might be due to memory management, or just noise):

Profiling

I used this command to profile the execution:

...

This is essentially only checking if the name contains "TimeTick", because in that case it has to generate a new piece of simulated FT2 file. In my case I am using a real FT2 file, so the check is false. Now, as a quick-and-dirty fix, just by substituting that line with:

...

I reduce the execution time to ~1/3 of what they it used to be:

An easy fix would be to cache the name of the source in the FluxSource class, so that it is only computed once.