[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Incoherent brem
Pawel,
Yes, this is true. Since all you want is the incoherent generator, you
can easily increase the efficiency of the generator by an order of
magnitude just by reducing the freqMaximum parameter (see beamgen.F) by
an order of magnitude. Reducing this constant has no effect on the
generator except to increase its efficiency, until you start to notice
the warning "freq exceeds freqMaximum" in the output log file from the
simulation. When you start to see this warning more than a couple times
per thousand events then it is time to raise freqMaximum up again.
For the sake of the rest of us that use the coherent bremsstrahlung
generator, I have improved the overall algorithm efficiency by an order
of magnitude or so, and uploaded it to subversion. If you simply check
out the latest version and modify the code in beamgen.F again like you
already did before to only take the incoherent branch, you should get
performance that I think it close to optimal, something like 50%.
If you want to know more about this parameter freqMaximum and how it
affects the simulation, you can read the log comments that I submitted
when I committed the modified beamgen.F to subversion.
Richard Jones
pawel@jlab.org wrote:
> Dear Richard,
>
> The change in the code, as you said, was really simple.
> It turns out that the Monte Carlo efficiency is quite low, about 2.5%.
> Is there a way to increase this efficiency?
>
> Thanks,
> Pawel
>
>