Subject: Re: [linux-audio-dev] Re: [DATAPOINT] kernels and latencies
From: Andrew Morton (andrewm_AT_uow.edu.au)
Date: Tue Jul 04 2000 - 07:00:56 EEST
Roger Larsson wrote:
>
> > I don't think mmap002 is a very pointful test. "Don't
> > run mmap002 when you're doing multimedia".
> >
>
> Sadly it is...
> There will be other programs that utilizes mmap that can give the same
> problems...
True, but sys_mmap() is not appearing on the radar screen.
The point the audio guys have been making is that they require
reasonable latency on a *dedicated machine*. That changes the game a
lot.
I'll take a peek at sys_msync(), but I think "don't do that" is a
solution here.
>
> ...
> > 160 milliseconds in sys_msync() -> "don't use sys_msync()...."
>
> What do you use to get this output?
It's based on http://www.uow.edu.au/~andrewm/linux/#timepegs
I've added a lot of new stuff for this activity. I'll get a new set of
patches and docco out later this week.
I think the sys_clone() and sys_exit() numbers may be wrong - the
current instrumentation is probably getting confused by system calls
which are entered once and exitted twice or never..
Pretty soon, I'll ask the audio people to put their money where their
mouthes are, and start testing, analysing and helping to "file off the
high spots".
>
> Note: I have seen X causing 50 ms latencies...
> [aux_write_dev+26/28]
awwwww.. Yuk. What a piggy little function!
That's going to cause network overruns at 100mbits, heaven knows what
else. It needs a redesign.
In the meanwhile, this probably means that people who require low
latency will need to buy an RS232 mouse. That's OK.
This archive was generated by hypermail 2b28 : Tue Jul 04 2000 - 07:19:01 EEST