[mythtv-users] Mythfrontend idle cpu consumption help

Jim Stichnoth stichnot at gmail.com
Wed Nov 11 17:49:47 UTC 2009


On Wed, Nov 11, 2009 at 8:43 AM, Johnny <jarpublic at gmail.com> wrote:
> Now back to the CPU usage. What is the right approach to understanding
> this or to giving the devs something informative and actionable. I
> have tried strace as was suggested. But that just shows system calls,
> I don't know how helpful that is. I get this over and over again:
>
> clock_gettime(CLOCK_MONOTONIC, {3769, 150476635}) = 0
> read(8, 0x8b3bd00, 4096)                = -1 EAGAIN (Resource
> temporarily unavailable)
> gettimeofday({1257956725, 614018}, NULL) = 0
> clock_gettime(CLOCK_MONOTONIC, {3769, 150674908}) = 0
> read(8, 0x8b3bd00, 4096)                = -1 EAGAIN (Resource
> temporarily unavailable)
> read(17, 0x8b72f60, 4096)               = -1 EAGAIN (Resource
> temporarily unavailable)
> gettimeofday({1257956725, 614218}, NULL) = 0
>
> And when I look at the count (strace -c), it looks like it spends most
> of its CPU time in those read() calls that are failing. However, I
> haven't done this kind of debugging before and I don't want to waste
> time if this isn't even relevent.

That's pretty much what I'm seeing in my reasonably working system.
You might add the "-tt" flag to strace to add timestamps to the
output, and verify that this pattern is repeating about 70 times per
second.

In the meantime, one of the developers has opened ticket #7525
(http://svn.mythtv.org/trac/ticket/7525) which, when implemented,
should greatly improve this behavior.  I'll be eagerly awaiting that
patch.

Jim


More information about the mythtv-users mailing list