On Sun, 20 Jan 2013 15:49:51 -0600 Rob Browning <r...@defaultvalue.org>
wrote:
> Mike Hommey <mh+report...@glandium.org> writes:
>
> > My system is a mid-2011 MBP. I'm leaving it booted for days, and
> > noticed something rather odd, considering the little pommed supposedly
> > does: pommed's cumulative time is awfully high.
>
> I noticed the same thing some while back, and though I haven't really
> had time to investigate, I just poked around a bit an noticed that with
> the default settings:
>
> #define DBUS_TIMEOUT 200
> #define KBD_TIMEOUT 200
> #define POWER_TIMEOUT 200
>
> Running pommed for 10s results (here) in about 0.05s of cpu time:
>
> real 0m10.275s
> user 0m0.028s
> sys 0m0.024s
>
> If you change the value from 200 to 100, you get a notable increase:
>
> real 0m10.272s
> user 0m0.032s
> sys 0m0.164s
>
> And if you change the value to 1000, a notable decrease:
>
> real 0m10.195s
> user 0m0.024s
> sys 0m0.012s
>
> Of course, most of the change is in the sys time, which isn't
> surprising since we're altering the epoll timeouts.
>
> I suppose one initial question would be whether or not pommed really
> needs to be polling everything so frequently, though I'm not familiar
> enough with what it's doing to know.
>
> I'd also wonder what it's doing with the user CPU time, and whether or
> not that could be improved.
>
> It is surprising that pommed would be one of the top cumulative CPU
> consumers on a machine.
>
> Oh, and here's the command I was using to gather the data:
>
> (time pommed/pommed -f > /dev/null&
> pommed_pid=$(ps -C pommed -o pid --no-headers)
> echo $pommed_pid
> sleep 10
> kill "${pommed_pid}")
The same here, raising *_TIMEOUTs from 200 to 1000 has no drawbacks and it
consumes a lot less CPU.
I think it should be patched that way.
Best, Nico
>
> --
> Rob Browning
> rlb @defaultvalue.org and @debian.org
> GPG as of 2011-07-10 E6A9 DA3C C9FD 1FF8 C676 D2C4 C0F0 39E9 ED1B 597A
> GPG as of 2002-11-03 14DD 432F AE39 534D B592 F9A0 25C8 D377 8C7E 73A4
>
>
