On Jan 11, 3:35 pm, Martin <[email protected]> wrote:
> Hi,
>
> Thanks for Devel::NYTProf 3.
>
> I have a daemon process which sits in DBI->execute for long periods of
> time because it has called Oracle's dbms_alert.waitany to wait for an
> alert. Once some procedure issues an alert the execute unblocks, we
> find out what alert was issued and then call another procedure to
> obtain work to do. Running Devel::NYTProf on this daemon we get > 95%
> of the time in DBI->execute blocked in waitany :-( - this makes it
> more difficult to home in on the areas needing attention.
>
> I tried setting usecputime=1 but got huge numbers like this:
>
> 111     1       1       2186934044s     4876069979s     
> XXX::Data::Generation::refreshXXXLists
>
> and yet the daemon ran for around 5 minutes for that run.
>
> I looked for different clocks as recommended in usecputime but I only
> have:
>
> /usr/include/bits/time.h:#   define CLOCK_REALTIME              0
> /usr/include/bits/time.h:#   define CLOCK_MONOTONIC             1
> /usr/include/bits/time.h:#   define CLOCK_PROCESS_CPUTIME_ID    2
> /usr/include/bits/time.h:#   define CLOCK_THREAD_CPUTIME_ID     3
>
> and none of these help.
>
> I tried running with start=no and calling enable_profile and
> disable_profile but the problem is that the code outside of the
> waitany unblocking is in various loops and every time enable_profile
> is called a new nytprof.out file is created (losing me all the
> previous data). There is no place I can put the enable_profile/
> disable_profile which is not in a loop.
>
> Is there any way around this?

OOPs - ignore most of this, I was calling finish_profile instead of
disable_profile.

Sorry for wasting your time.

Martin
-- 
You've received this message because you are subscribed to
the Devel::NYTProf Development User group.

Group hosted at:  http://groups.google.com/group/develnytprof-dev
Project hosted at:  http://perl-devel-nytprof.googlecode.com
CPAN distribution:  http://search.cpan.org/dist/Devel-NYTProf

To post, email:  [email protected]
To unsubscribe, email:  [email protected]

Reply via email to