Re: ext2 performance in 2.5.25 versus 2.4.19pre8aa2

[Andrea Arcangeli]

On Mon, Jul 15, 2002 at 04:06:21PM +1000, Lincoln Dale wrote:
> At 10:30 PM 14/07/2002 -0700, Andrew Morton wrote:
> >Funny thing about your results is the presence of sched_yield(),
> >especially in the copy-from-pagecache-only load.  That test should
> >peg the CPU at 100% and definitely shouldn't be spending time in
> >default_idle.  So who is calling sched_yield()?  I think it has to be
> >your test app?
> >
> >Be aware that the sched_yield() behaviour in 2.5 has changed a lot
> >wrt 2.4.  It has made StarOffice 5.2 completely unusable on a non-idle
> >system, for a start.  (This is a SO problem and not a kernel problem,
> >but it's a lesson).
> 
> my test app uses pthreads (one thread per disk-worker) and 
> pthread_cond_wait in the master task to wait for all workers to finish.
> i'll switch the app to use clone() and sys_futex instead.

unless you call pthread routines during the workload, pthreads cannot be
the reason for a slowdown.

Also I would suggest Andrew to benchmark 2.4.19rc1aa2 against 2.5
instead of plain rc1 just to be sure to compare apples to apples.
(rc1aa2 should also be faster than pre8aa2)

BTW, Lincol, I still have a pending answer for you, about the mmap
slowdown, that's because of reduced readahead mostly, you can tune it
with page-cluster sysctl, it's not only because of the expensive page
faults that mmap I/O implies. I've some revolutionary idea about
replacing readahead, not that it matters for your workload that is
reading physically contigous though.

Andrea