Subject: Re: [linux-audio-dev] 1000HZ ticks for x86 in the 2.5-tree
From: Erik Walthinsen (omega_AT_temple-baptist.com)
Date: Wed Jul 10 2002 - 02:01:16 EEST
On Wed, 10 Jul 2002, Kai Vehmanen wrote:
> Not a problem. Timing sensitive user-space apps benefit from this change
> because with 1kHz ticks, kernel will be making scheduling decisions much
> more often.
The solution to a lot of these scheduling problems (non-RT) is firm
timers:
http://www.cse.ogi.edu/~luca/firm.html
On machines with a sane timer interrupt (anything x86 with an APIC and
just about all other architectures), it allows extremely fine-grained
control. One example is MPlayer, which sleeps for the next frame.
Without firm timers the wakeups occur on the 30ms and 40ms boundaries
(nominally 33.3ms). With firm timers, they occur at 33.3ms *every* time
(excepting long kernel paths).
Something like this really *really* needs to get into the kernel, because
it's useful for everything from soft-RT to video work to the kernel IP
traffic control (many limits are based on HZ there) and user
responsiveness in general in some cases.
Erik Walthinsen <omega_AT_temple-baptist.com> - System Administrator
__
/ \ GStreamer - The only way to stream!
| | M E G A ***** http://gstreamer.net/ *****
_\ /_
This archive was generated by hypermail 2b28 : Wed Jul 10 2002 - 02:07:32 EEST