> Something must be wrong on your side. If you take the example frontend under
>
> midas/examples/experiment/frontend.cxx
>
> and let it run to produce dummy events, you get about 90 Hz. This is because we have a
>
> ss_sleep(10);
>
> in the read_trigger_event() routine to throttle things down. If you remove that sleep,
> you get an event rate of about 500'000 Hz. So the framework is really quick.
>
> Probably your routine which looks for a 'lam' takes really long and should be fixed.
>
> Stefan
Sorry if I am going off-topic but, because the ss_sleep function was mentioned here, I
would like to take the chance and report an issue that I am having.
In all my slow control frontends, the CPU usage for each frontend is close to 100%. This
means that each frontend is monopolizing a single core. When I did some profiling, I
noticed that 99% of the time is spent inside the ss_sleep function. Now, I would expect
that the ss_sleep function should not require any CPU usage at all or very little.
So my two questions are:
Is this a bug or a feature?
Would you able to check/reproduce this behavior or do you need additional info from my
side? |