Without tracks?

Can we make a spill an event?

That's the solution I think, and a good thing to do anyway. It's a days work (means doing the EventAction properly).

I've implemented the event structure in my branch, a worthwhile thing to do anyway. Running now 100 events over just 3 RF cavities. Still get about factor 5 slowdown compared to G4MICE. Notice optimize flags aren't set, let's try that next.

Do you think that Simon would mind setting up something like:

http://speed.pypy.org/

since the code is linked to at the bottom?

We can ask... (Simon added as watcher)

ps:
Adding -O2 or -O3 gave factor 2 in speed, so still missing factor 2 somewhere. Doubt this is the persistency (getting 0.5 k data per primary), but don't know what else has changed. gprof next.