26 Dec
2009
26 Dec
'09
11:24 p.m.
I have just put up my steps (and many thanks Cal!) for compiling a kernel:
http://linuxlive.joshuacorps.org/?p=318
Comments and suggestions solicited. One item: can someone who uses
CCRMA or close enough tell me, what is the RPM package for installing a
complete kernel-building-capable compilation environment including
menuconfig (i.e., working and compatible ncurses etc.)?
J.E.B.
29 Dec
29 Dec
10:24 p.m.
On Sun, Dec 27, 2009 at 9:54 AM, Jonathan E. Brickman
<jeb(a)joshuacorps.org> wrote:
I have just put up my steps (and many thanks Cal!) for
compiling a kernel:
http://linuxlive.joshuacorps.org/?p=318
Hey Thanks for that!
I taught computer science for 20 years at university and I never
compiled a kernel.
And now as a noob on this list just finished compiling my very own
gleaming new kernel.
I will put up some suggestions questions and (small) corrections in
due course (perhaps on the blog directly?)
But for now one question: How does one quantitatively measure that
one kernel is better than another? eg memory footprint, response time
etc?
My current main use case is nted using timidity using whatever is my
stock laptop hardware.
10:54 p.m.
On Tue, Dec 29, 2009 at 7:24 PM, Rustom Mody <rustompmody(a)gmail.com> wrote:
But for now one question: How does one quantitatively measure that
one kernel is better than another? eg memory footprint, response time
etc?
For those of us doing real-time audio I measure it using Jack and
determining with each kernel how low I can set my latencies while
recording 48 channels of audio in one pass to three hard drives.
That's just me though.
Cheers,
Mark
11:19 p.m.
But for now
one question: How does one quantitatively measure that
one kernel is better than another? eg memory footprint, response time
etc?
For those of us doing real-time audio I measure it using Jack and
determining with each kernel how low I can set my latencies while
recording 48 channels of audio in one pass to three hard drives.
That's just me though.
Cheers,
Mark
30 Dec
30 Dec
9:58 a.m.
On Tue, Dec 29, 2009 at 8:19 PM, Jonathan E. Brickman
<jeb(a)joshuacorps.org> wrote:
Mark, I am curious: how low /can/ you set your latencies while
capturing 48 channels simultaneously to three hard drives??? And are
you using a firewire device, or something else?
J.E.B.
OK, so I don't change latency while actually recording. Changing
latency requires restarting Jack and the audio apps. Probably I should
have said I test Ardour recording 48 channels to 3 drives with
different latency settings in Jack.
I use 3 Firewire drives and tell Ardour that they are attached. Ardour
decides what audio data goes on each drive, but roughly speaking it
spreads out the audio data across all three. When I first did this a
few years ago I ran into trouble trying to do this on a single
internal IDE drive, even if it wasn't the main system drive. It might
work better today with newer drivers and SATA drives. Don't know.
Hope this clears things up.
- Mark
But for
now one question: How does one quantitatively measure that
one kernel is better than another? eg memory footprint, response time
etc?
For those of us doing real-time audio I measure it using Jack and
determining with each kernel how low I can set my latencies while
recording 48 channels of audio in one pass to three hard drives.
That's just me though.
Cheers,
Mark
6:35 p.m.
OK, so I don't change latency while actually
recording. Changing
latency requires restarting Jack and the audio apps. Probably I should
have said I test Ardour recording 48 channels to 3 drives with
different latency settings in Jack.
I use 3 Firewire drives and tell Ardour that they are attached. Ardour
decides what audio data goes on each drive, but roughly speaking it
spreads out the audio data across all three. When I first did this a
few years ago I ran into trouble trying to do this on a single
internal IDE drive, even if it wasn't the main system drive. It might
work better today with newer drivers and SATA drives. Don't know.
Hope this clears things up.
- Mark
Yup. I very much wonder, if you would have less struggle sending all of
the streams to one RAID-10 set (that's four drives minimum) of either
internal SATA or eSATA. RAID-5 (three drives minimum) wouldn't help
because it accelerates read, but not write; but RAID-10 effectively
doubles both read and write speeds.
J.E.B.
6:57 p.m.
On Wed, Dec 30, 2009 at 3:35 PM, Jonathan E. Brickman
<jeb(a)joshuacorps.org> wrote:
Yes, probably so. However I have no money to do such things, etc., so
it remains something for a possible future.
One thing I will say about 1394 is it's easy to put out of the way
physically, at the end of a long cable in a closet or backroom, and
thus reducing noise. Doing things inside the chassis has noise, power
and heat issues that might effect some folk's decisions about how to
handle this sort of thing. Using 1394 allows (or allowed with my
previous laptop really) to use the same drives with a desktop setup or
a laptop.
There's a lot of ways to go with stuff like this. I had some 1394a and
1394b drives hanging around at the time so it didn't cost me anything
to go that way. It would surely be different for others.
Cheers,
Mark
OK, so I don't change latency while actually
recording. Changing
latency requires restarting Jack and the audio apps. Probably I should
have said I test Ardour recording 48 channels to 3 drives with
different latency settings in Jack.
I use 3 Firewire drives and tell Ardour that they are attached. Ardour
decides what audio data goes on each drive, but roughly speaking it
spreads out the audio data across all three. When I first did this a
few years ago I ran into trouble trying to do this on a single
internal IDE drive, even if it wasn't the main system drive. It might
work better today with newer drivers and SATA drives. Don't know.
Hope this clears things up.
- Mark
Yup. I very much wonder, if you would have less struggle sending all of
the streams to one RAID-10 set (that's four drives minimum) of either
internal SATA or eSATA. RAID-5 (three drives minimum) wouldn't help
because it accelerates read, but not write; but RAID-10 effectively
doubles both read and write speeds.
J.E.B.
29 Dec
29 Dec
11:17 p.m.
I have just
put up my steps (and many thanks Cal!) for compiling a kernel:
http://linuxlive.joshuacorps.org/?p=318
Hey Thanks for that!
I taught computer science for 20 years at university and I never
compiled a kernel.
And now as a noob on this list just finished compiling my very own
gleaming new kernel.
I will put up some suggestions questions and (small) corrections in
due course (perhaps on the blog directly?)
But for now one question: How does one quantitatively
measure that
one kernel is better than another? eg memory footprint, response time
etc?
My current main use case is nted using timidity using whatever is my
stock laptop hardware.
Welcome to the club :-) I really must give major thanks to Cal, he
brought me back to distribution-independent kernel skills I had lost
years back.
Indeed, quantitative measurement can be interesting, in part because
needs are different. I have 4G RAM, so I don't optimize for size;
memory footprint is not part of my "improvement" quantitivity [*grin*],
I want the speed. I'm using a balance of (a) latency with (b) the
number of simultaneous voices my primary software synth (Yoshimi) can
produce, hold, and reverb. It's a balance, because as I set my latency
requirements lower, the CPU demands on I/O increase, reducing the
available resources for (b).
J.E.B.
5482
days inactive
5486
days old
linux-audio-user@lists.linuxaudio.org
7 comments
3 participants
participants (3)
-
Jonathan E. Brickman -
Mark Knecht -
Rustom Mody