[Halld-offline] HDGeant4 update tomorrow, last in a series

Richard Jones richard.t.jones at uconn.edu
Tue Jan 31 14:37:50 EST 2017


I would like to give an update on HDGeant4 at the offline meeting tomorrow.
I plan this to be the last during the roll-out phase of the development
cycle. Construction is now complete to my satisfaction. Processing speed is
now at the level I consider acceptable, about 40% of the event throughput
per thread that HDGeant shows on the same hardware. Of course, using
multiple cores, it easily outdoes the G3 simulation in raw throughput.
Running on 4 cores, it runs 3.95 userCPU seconds per second, with a
resident set of 520 MB total for all 4 cores, around 130 MB per core.

At this point I am satisfied that it is ready for people to start trying
out. There will still be a period until the results look completely
consistent, but I am done with my active development. I will call this
version 1.0. I have several benchmark comparisons that I will show to give
confidence that it is working as it should. There is still one detector
that is yet to be fully "instrumented" and that is the DIRC. I will work on
that at some point, but I think there are other more pressing issues
calling for my attention right now and HDGeant4 is now mature enough to
pass into production.

-Richard Jones
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-offline/attachments/20170131/dc779ef7/attachment.html>

More information about the Halld-offline mailing list