So after editing the limiting file, I was able to get realtime output of 
a fairly hefty ensemble ( 25% cpu on an AMD3000+ ) at sr=44100/ksmp=1 
down to -b8 -B32. I even got it to work some of the time at -b4 -B16, 
which just barely cracks up. I have Ingo Molnar's patch applied to 
2.6.12, will try more aggressive system tweaking soon to see if I can 
get that -b4 -B16 working nicely.

So it's not -b1 but it's still pretty screaming. =)
Iain


-------------------------------------------------------
This SF.Net email is sponsored by the JBoss Inc.  Get Certified Today
Register for a JBoss Training Course.  Free Certification Exam
for All Training Attendees Through End of 2005. For more info visit:
http://ads.osdn.com/?ad_id=7628&alloc_id=16845&op=click
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

> So after editing the limiting file, I was able to get realtime output of 
> a fairly hefty ensemble ( 25% cpu on an AMD3000+ ) at sr=44100/ksmp=1 
> down to -b8 -B32. I even got it to work some of the time at -b4 -B16, 
> which just barely cracks up. I have Ingo Molnar's patch applied to 
> 2.6.12, will try more aggressive system tweaking soon to see if I can 
> get that -b4 -B16 working nicely.
> 
> So it's not -b1 but it's still pretty screaming. =)
> Iain

Oops, that was supposed to go to the main list. Anyway the corrected 
correction is that -b4 -B16 works with --sched --expression-opt.

Iain


-------------------------------------------------------
This SF.Net email is sponsored by the JBoss Inc.  Get Certified Today
Register for a JBoss Training Course.  Free Certification Exam
for All Training Attendees Through End of 2005. For more info visit:
http://ads.osdn.com/?ad_id=7628&alloc_id=16845&op=click
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Holy smokes, --sched seems to work wonders. My modest ensemble is 
running without breaking up *with message output* at -b2 -B4!

Apparently I am not allowed by the card to set these any lower. Istvan 
is this a hardware limitation? Are we sure about those limits?

So, that's what, a theoretical csound latency of .09 ms? or is 0.18? 
Damn! Mad props to everyone who made this possible. This make routing 
audio through multiple machine layers eminently usable. My girlfriend is 
not going to like that ... ha ha ha!

Iain
( sorry for being so scattered today with these tests. )




-------------------------------------------------------
This SF.Net email is sponsored by the JBoss Inc.  Get Certified Today
Register for a JBoss Training Course.  Free Certification Exam
for All Training Attendees Through End of 2005. For more info visit:
http://ads.osdn.com/?ad_id=7628&alloc_id=16845&op=click
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

> On Saturday 19 November 2005 22:10, Iain Duncan wrote:
> 
>>Holy smokes, --sched seems to work wonders. My modest ensemble is
>>running without breaking up *with message output* at -b2 -B4!
> 
> Are you sure?
> I haven't heard of anybody going that low without breakups in the sound. 
> 
> Does the rtalsa-plugin report these settings or does csound?
> Have you tried some audio loopback latency testing with these settings?
> (Keeping in mind that your adc/dac converters add somewhere between 16 and 64 
> frames latency each way depending on the make of the converters)
> 
> Very curious to hear the results

I'm sure about the latency settings because if I take off --sched, I 
can't get below -b4 -B16, and even that cracks out a little. ( I edited 
out the limits in the alsa module). However, total system latency from 
audioin-through csound->audio out is 1.5 ms, measured with a loopback 
through a hard disk recorder and looking at the resultant wave files. So 
my system is obviously the main latency issue. I have been told this 
should be improvable by going from my 2.6.12 to 2.6.14rt, I will do the 
test again then. Because of the system latency, the effective difference 
between -b2 -B4 and -b4 -B16 seems to be none.

Iain


-------------------------------------------------------
This SF.Net email is sponsored by the JBoss Inc.  Get Certified Today
Register for a JBoss Training Course.  Free Certification Exam
for All Training Attendees Through End of 2005. For more info visit:
http://ads.osdn.com/?ad_id=7628&alloc_id=16845&op=click
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

  >>However, total system latency from
>>audioin-through csound->audio out is 1.5 ms, measured with a loopback
>>through a hard disk recorder and looking at the resultant wave files. So
>>my system is obviously the main latency issue. I have been told this
>>should be improvable by going from my 2.6.12 to 2.6.14rt, I will do the
>>test again then. Because of the system latency, the effective difference
>>between -b2 -B4 and -b4 -B16 seems to be none.
> 
> 1.5ms is about 64 sampleframes. Your buffer is 4 frames so that leaves about 
> 28 samples latency for the hardware each way. There is nothing in software 
> that can change that, that is just build in the electronic circuit. In the 
> some way every digital mixer has channel-latency on its analog inputs. 
> So B16 or B32 are good defaults for live effects I think. And that is still a 
> lot better than windows or OSX can do.
> It is amazing that a general purpose OS can get that close to a hard realtime 
> embedded system.
> 
> G

Thanks for the explanation. I was curious as to how to find that stuff 
out. I have read of people getting hardware latency with linux down to 
0.6 ms, but would that then be one way only? And I suppose it becomes 
card dependent on how many samples of latency the hardware produces. I 
am really pleased with the results though. You are a jack user correct? 
Have you experminted much yet with lowest possible jack latency settings 
user -+rtaudio=jack?

> PS
>  1.5 ms latency is what you get when you move the loudspeaker about 15 inches 
> further away.

Ha ha, I realize it's very minor. My concern ( as I'm sure you know ) is 
that for certain fx the difference between say 1ms and 5ms dramatically 
changes the perception of the effect if you intend to be listening to 
the unprocessed sound source at the same time, as would be the case if 
the user were to use csound on an old laptop as an fx unit plugged into 
an aux bus on a conventional mixer.

Thanks again
Iain


-------------------------------------------------------
This SF.Net email is sponsored by the JBoss Inc.  Get Certified Today
Register for a JBoss Training Course.  Free Certification Exam
for All Training Attendees Through End of 2005. For more info visit:
http://ads.osdn.com/?ad_id=7628&alloc_id=16845&op=click
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net