performance in pd - audiobuf, etc

[Karl MacMillan]

On Mon, 11 Dec 2000, d d wrote:

> Thanks Karl,
> I think I may need some clarification if you have the time
> 
> Is there a way to force PD to use the ASIO drivers for my Echo/Layla in 
> win2K (layla drivers are beta in win2K) and forsake the windows sound 
> driver?
> 

This will require extra code in PD.  Also, there is some sort of SDK that
requires licensing.

> So with Linux (+ lowlatency kernel) you can actually get 5ms latency?
> That is basically the timing I am looking for.
> But I've never used Linux, so unless my linux friends help me, I'd be 
> stranded with this as a solution :)
> 

People claim this and I am about to start testing this.  In my experience
I can get latency that is very low but I don't have a hard number.  It is
low enough that when you mix the original signal with the signal going
through linux it is not perceivable as latency, but rather as filtering.

> What you are saying is that Linux w/ RME hardware is the only way to get 
> around 3-9ms latency with PD?
> 

No - it is just a good multi-channel card that works under linux.

> What about linux w/ multiple soundblasters or something?
> 

There are sync problems with multiple cards - you can search the PD
archives for some good posts about this.

> Is it linux + lowlatency kernel that gets 5ms? or that bundled with RME 
> hardware?
> 

It is the low overhead of the linux driver model combined with the low
scheduling latency of the lowlatency kernel - it is mostly independent
of hardware (though the hardware does introduce some latency itself).

Karl

> Thanks so much!
> 
> Also, very good to hear you mention MSP/motu which I was also considering as 
> an alternative. So that fabled 6ms isn't really attainable, interesting. 
> Good thing I know this now, as I have no mac's near me to test this on.
> 
> Thanks again everyone for their help,
> I really don't want to buy a kyma :)
> I'd much rather use PD, if I can get 5ms out of it, I will be ecstatic
> 
> -daniel
> 
> >Really the only way to fix this problem is to adapt PD to use ASIO
> >drivers.  This was mentioned recently on the list, but the problem with
> >the normal windows sound drivers is that they go through a kernel mixer
> >(called kmixer - this allows more than one app to access a sound card
> >simultaneously) that adds a minimum of 30ms.  It seems that a lot of pro
> >cards circumvent this with their ASIO drivers.
> >
> >The best solution is linux, but then you don't get to use your layla
> >because the manufacturer is not nice.  Of course, if you are willing to
> >buy a macintosh you could just be an RME hammerfall card instead, install
> >linux + low-latency kernel, and enjoy 5ms latency regardless of disk or
> >cpu load.  As for the Macintosh latency I think that the 6ms measurement
> >is not a real measure of end to end latency - I have never seen the
> >Max/MSP machines that I have access to (G3s with MOTU 2408) get anything
> >near that because of the complete inadequacy of the MacOS in all regards
> >(sorry about that comment - I couldn't stop myself).
> >
> >Karl
> >
> >On Mon, 11 Dec 2000, Michael Lechasseur wrote:
> >
> > >
> > > 	Perhaps you could dual boot to Linux?  I don't have any latency data 
> >for PD
> > > on Win2000, but if you'd like to write a latency test patch, I could run 
> >it
> > > on my system and send you the results?
> > >
> > > 	I've wanted to test this since applying the kernel low-latency patch, 
> >and
> > > have never gotten around to it.
> > >
> > > 	I have an Athlon 600, 64MB Ram, Redhat 6.2, Kernel 2.2.16 with the
> > > low-latency patch.
> > >
> > > > -----Original Message-----
> > > > From: d d [mailto:denglertdp at hotmail.com]
> > > > Sent: Monday, December 11, 2000 1:02 PM
> > > > To: pd-list at iem.kug.ac.at
> > > > Subject: performance in pd - audiobuf, etc
> > > >
> > > >
> > > > Hello, I'm new to PD, and have been using it sporadically for
> > > > several weeks.
> > > > It is extremely interesting, seems to do the things I can't do with my
> > > > NordModular. My thanks to all who made it possible!
> > > >
> > > > I know you probably don't want to hear this question again, but I was
> > > > wondering if there was a way to get real time audio processing
> > > > performance
> > > > out of PD. (windows 2000 pent III 500mHz, awe64, echo layla 20bit)
> > > >
> > > > I've tried adjusting all the command params, disabling devices, 
> >changing
> > > > audiobuf, etc
> > > >
> > > > Is there anyway to knock the latency to hardware like levels on
> > > > my system?
> > > > Even if the answer is different hardware, OS, etc
> > > > I'm considering a Mac purchase because cycling74 claims w/
> > > > certain hardware
> > > > a 6ms latency can be had.
> > > > Is anything this low possible with PD?
> > > > I'd much rather stay within windows.
> > > >
> > > > Sorry for the bother with this question
> > > >
> > > > I'd like to have this ultimate control over processing/creating
> > > > that PD can
> > > > give, but I also need hardware like latency In/Out
> > > > I don't want to spend money on a Kyma, because it seems logical that a 
> >PC
> > > > can do this sort of thing these days (what with the nifty fast chips 
> >and
> > > > all)
> > > >
> > > > Thanks all who read this far, sorry for the dull question
> > > > -daniel
> > > > _________________________________________________________________
> > > > Get your FREE download of MSN Explorer at http://explorer.msn.com
> > > >
> > >
> > >
> >
> >--
> >_____________________________________________________
> >| Karl W. MacMillan                                 |
> >| Computer Music Department                         |
> >| Peabody Institute of the Johns Hopkins University |
> >| karlmac at peabody.jhu.edu                           |
> >| www.peabody.jhu.edu/~karlmac                      |
> >-----------------------------------------------------
> >
> >
> 
> _________________________________________________________________
> Get your FREE download of MSN Explorer at http://explorer.msn.com
> 
> 

-- 
_____________________________________________________
| Karl W. MacMillan                                 |
| Computer Music Department                         |
| Peabody Institute of the Johns Hopkins University |
| karlmac at peabody.jhu.edu                           |
| www.peabody.jhu.edu/~karlmac                      |
-----------------------------------------------------