[64studio-devel] 0.9.5 is out

Lee Revell rlrevell at joe-job.com
Tue Oct 24 20:49:41 UTC 2006


On Tue, 2006-10-24 at 22:43 +0200, Quentin Harley wrote:
> This is the way we learn...
> 
> Do you know what the original motivation for the high res timers were?  I'd
> like to know out of curiosity sake.
> 
> Thanks for the feedback
> 

High res timers mean that when a process asks to sleep for X amount of
time, it will sleep for exactly that amount of time (to within hardware
limits, usually 20 microseconds or so).

With standard timers, the process will sleep for somewhere between X and
X + 1/HZ seconds.  1/HZ can be 1ms, 2.5ms, or 10ms depending on the
setting of CONFIG_HZ in kernel config.

It's useful for high precision timing (stuff like sending MIDI time code
wants this), but for maximum benefit the app should be written to be HRT
aware.

IOW, it's a useful feature, but this early in it's development I would
go for stability over new bells & whistles.

Anyway the point is moot because it's disabled in the 64studio config
anyway.

> Quentin
> 
> -----Original Message-----
> From: Lee Revell [mailto:rlrevell at joe-job.com] 
> Sent: 24 October 2006 15:50
> To: Quentin Harley
> Cc: 64studio-devel at 64studio.com
> Subject: RE: [64studio-devel] 0.9.5 is out
> 
> On Tue, 2006-10-24 at 08:52 +0200, Quentin Harley wrote:
> > High res should be a selectable feature, if it is not standard.
> > 
> > High res timers has been working fine on my system, and I even saw some
> (be
> > it psychological) improvement on low latency handling for live work. Less
> > xruns for the heavy stuff.
> > 
> 
> High res timers have nothing to do with xruns.  The effect is certainly
> psychological.
> 
> > We should not let all users be tied down to make the distro "standard".
> > I know of a big Seattle-based software company that does just that.
> > 
> > Is it possible to let the installer check the CPU and install the relevant
> > kernel?
> > 
> 
> There's a big difference between tying everyone down and shipping an
> experimental unstable feature.
> 
> There's no way to know a priori whether high res timers will work on a
> given system, there are so many that have buggy LAPIC and other issues.
> 
> > Regards,
> > Quentin
> > 
> > 
> > -----Original Message-----
> > From: 64studio-devel-bounces at 64studio.com
> > [mailto:64studio-devel-bounces at 64studio.com] On Behalf Of Lee Revell
> > Sent: 24 October 2006 02:39
> > To: joel silvestre
> > Cc: 64studio-devel at 64studio.com
> > Subject: Re: [64studio-devel] 0.9.5 is out
> > 
> > On Tue, 2006-10-24 at 02:11 +0200, joel silvestre wrote:
> > > Hi,
> > > The new 2.6.18 kernel doesn't boot on my Intel core duo laptop.
> > > So I reinstall 64studio 0.94 and upgrade to 0.95 from the CD with
> > > synaptic.
> > > Is it a good or a bad idea?
> > > 
> > > Jackd -V returns 0.100.0 and synaptic says 0.101.1-1.64studio is
> > > installed. Where is the truth?
> > > 
> > > Are the 2.6.18 kernel sources available? I think the kernel hang because
> > > of the timer. Disabling both hpet and hrt should help as it does for a
> > > vanilla 2.6.18-rt7.
> > 
> > Does the 0.9.5 kernel enable high res timers?  I don't think it should.
> > It's too new and experimental a feature.
> > 
> > Lee
> > 
> > 
> > _______________________________________________
> > 64studio-devel mailing list
> > 64studio-devel at 64studio.com
> > http://lists.64studio.com/mailman/listinfo/64studio-devel
> > 
> > 
> 
> 
> 




More information about the 64studio-devel mailing list