Subject: Re: [linux-audio-dev] Lots about latency and disk i/o and JACK...
On Wed, Oct 03, 2001 at 10:50:45PM +1000, Andre Pang wrote:
> > for example, even in 2.4.0, with the low latency patch, it was
> I wonder if Robert Love's kernel pre-emption patches would make
> From what I've heard on the kernel mailing list, people are _very_ happy
Plus, pre-2.4.10 kernels tend to allocate too large caches for programs that
Bye,
Peter Surda (Shurdeek) <shurdeek_AT_panorama.sth.ac.at>, ICQ 10236103, +436505122023
This archive was generated by hypermail 2b28
: Wed Oct 03 2001 - 18:08:42 EEST
From: Peter Surda (shurdeek_AT_panorama.sth.ac.at)
Date: Wed Oct 03 2001 - 18:11:33 EEST
> This is true, but turning on DMA support for the IDE hard disk
> lowers your latencies and CPU usage by several orders of
> magnitude.
ACK. Plus, kernel 2.4.10 (new VM) and preemption patch also help. The only
disadvantage I found is that with some hardware (VIA chipset and sblive) using
DMA produces corrupt sound.
> > possible to cause scheduling delays of 30-50-1000! msecs by hitting
> > the VM and disk subsystems (e.g. a 4-process C++ compile while running
> > Ardour).
Yes, lowlat patch is limited, because it only tries to reschedule things.
However, preemption patch allows it for more system calls to be running
concurrently, hence in (almost) all cases where the program that needs low
latency doesn't get pushed into swap you get very nice latency.
> this better. Has anybody tried them extensively?
Hmm depends. I noticed that with lowlat, scheduler tends to eat lot of CPU
time when load gets higher. This doesn't happen with preempt.
> with them. (Happy == xmms skips once in a blue moon with uptime loads
> of >10, including hitting the VM and disk hard.)
I have similar observations too. I mainly use my box for video, not only
audio.
don't need latency and put those who need (players and X) into swap. This is
way better in 2.4.10.
--
The dark ages were caused by the Y1K problem.