Re: sched_yield() makes OpenLDAP slow

[jfannin@gmail.com (Joseph Fannin)]

On Thu, Aug 18, 2005 at 02:50:16AM +0200, Bernardo Innocenti wrote:

> The relative timestamp reveals that slapd is spending 50ms
> after yielding.  Meanwhile, GCC is probably being scheduled
> for a whole quantum.
>
> Reading the man-page of sched_yield() it seems this isn't
> the correct behavior:
>
>    Note: If the current process is the only process in the
>    highest priority list at that time, this process will
>    continue to run after a call to sched_yield.

   The behavior of sched_yield changed for 2.6.  I suppose the man
page didn't get updated.

>From linux/Documentation/post-halloween.txt:

| - The behavior of sched_yield() changed a lot.  A task that uses
|   this system call should now expect to sleep for possibly a very
|   long time.  Tasks that do not really desire to give up the
|   processor for a while should probably not make heavy use of this
|   function.  Unfortunately, some GUI programs (like Open Office)
|   do make excessive use of this call and under load their
|   performance is poor.  It seems this new 2.6 behavior is optimal
|   but some user-space applications may need fixing.

    This is pretty much all I know about it; I just thought I'd point
it out.

> I also think OpenLDAP is wrong.  First, it should be calling
> pthread_yield() because slapd is a multithreading process
> and it just wants to run the other threads.  See:

    Is it possible that this problem has been noticed and fixed
already?

--
Joseph Fannin
jfannin@gmail.com

 /* So there I am, in the middle of my `netfilter-is-wonderful'
talk in Sydney, and someone asks `What happens if you try
to enlarge a 64k packet here?'. I think I said something
eloquent like `fuck'. - RR */