On Wed, 2005-02-02 at 09:55 +0100, Bengt Gördén wrote:
I'm using my laptop to make some music. At home I
planning to buy a
computer with enough power to host a lot of music software. But when
I'm away from home I'd like to use my laptop. The thing is that
according to the specification on my laptop it would be enough power
to run lots of demanding software. But there is, for example,
Rosegarden and I have problem with xruns when I try to run that. I
have an ACER Ferrari 3000 (a damned hot bastard that makes me wonder
if the ACER community doesn't have a fertility problem if they really
have this thing in their lap ;-) that has an Athlon-XP-M 1800+. I had
a lousy HD in it that i changed but that only boosted the memory-to-HD
performance. I seem to have other problems. The graphic is one
thing. I have done all the things for real time performance and other
stuff to get down the latency but still there is performance
problem. It seems to be connected to jack but I'm not really sure. It
would be interesting to here about other people who have set up
laptops for music production.
If there's already a site or compiled list of systems and how they
perform I would be delighted to be pointed to it. But if it's not it
would be nice if we could start a thread with how laptops are
performing as music stations.
First, there was also a pretty bad ACPI related latency bug fixed in
kernel 2.6.10. Try a new kernel before you assume that you have the
ACPI/SMM problem.
If that does not help, read on...
The ACPI/SMM Latency Problem
----------------------------
While testing the RT preemption patches we discovered that many, many
laptops are unsuitable for low latency audio work. Most of this info
came straight from Alan Cox. Search the LKML archives for more info.
The laptops in question have a BIOS that implements ACPI via SMM. SMM
is used for things like talking to the battery and the fan.
Unfortunately SMM events completely short-circuit the OS, and can block
*all* interrupts for a millisecond or more. By definition there is
nothing the OS can do about it, because the whole point of SMM is that
it doesn't need the OS to know about it and vice versa.
According to AC this is a known broken design (in addition to audio, it
causes these systems to lose timer ticks with HZ=1000), but by his
estimate more than half of laptops have the problem.
There is no 100% reliable way to tell if you have the ACPI/SMM problem.
If your xruns correspond to the fan speed changing, that's a dead
giveaway. Disabling ACPI in the kernel can help but will not eliminate
the problem completely because some ACPI events are initiated by the
hardware.
If you need really low latency, in the mllisecond range, this is a
showstopper! So I think the best approach is for the Linux audio
community to create a "whitelist" of known good laptops.
Lee