[chriss85]

I'm calling radiation.correct() in a loop a few thousand times and the performance of this function call is gradually getting worse. The other code I have written inside this loop is unaffected. The memory usage doesn't increase. I'm using a self-written radiation model that is a non-grey P1 model. Any idea what the reason might be? Can this be a memory related issue (page fault, needing to move memory on allocation, ...)?
How would one go about finding the reason and fixing it?

[chriss85]

As a follow up on the first post, I noticed that the performance gets good again after increasing the time step. I have absolutely no idea why, but I suppose there is some book keeping (possibly related to objectRegistry?) happening that leads to slowdowns when an equation is solved very often in an iteration that gets cleared on the next time step. This is just a guess, no idea if it's the case, but my benchmarks clearly show much better performance if I increase the time step regularly in the iteration. Naturally, I don't want to have this effect outside of the loop, so I backup the original time step and revert to it like this:

TimeState tSCurrent(runTime);
loop();
runTime.TimeState::operator=(tSCurrent);

As far as I can tell, this completely restores the time state but I would be happy if anyone can confirm this and/or provide an explanation for this effect I'm seeing.