19 Jan
2007
19 Jan
'07
8:10 p.m.
Hi all!
Florian Schmidt was kind enought to point out that I can increase my
system performance by increasing the priority of both the Jack daemon
and the IRQ that my trusty'ol bebob device is hitched to.
I temporarily gave up on putting the device onto another IRQ and will
leaving that up to the kernel developers to sort out; what I am
interested in right now is how to increase the priority of IRQ 10 on my
system.
Florian suggested simply increasing the priority of the 'IRQ 10'
interrupt handling process; but on my system, there is no interrupt
handling process 'IRQ 10', or for any other IRQ, for that matter.
How can I increase the priority of 'IRQ 10' on my Debian based system?
Help much appreciated.
Carlo
PS: upping JackD priority alone took care of 60% of xruns; thanks Florian!
19 Jan
19 Jan
8:59 p.m.
Carlo Capocasa wrote:
Hi all!
Florian Schmidt was kind enought to point out that I can increase my
system performance by increasing the priority of both the Jack daemon
and the IRQ that my trusty'ol bebob device is hitched to.
I temporarily gave up on putting the device onto another IRQ and will
leaving that up to the kernel developers to sort out; what I am
interested in right now is how to increase the priority of IRQ 10 on my
system.
Florian suggested simply increasing the priority of the 'IRQ 10'
interrupt handling process; but on my system, there is no interrupt
handling process 'IRQ 10', or for any other IRQ, for that matter.
How can I increase the priority of 'IRQ 10' on my Debian based system?
Help much appreciated.
Carlo
PS: upping JackD priority alone took care of 60% of xruns; thanks Florian!
Did you try using the rtirq script?
You can find it on my home server, here
http://www.rncbc.org/jack/rtirq-20070101.tar.gz
I guess you can follow the instructions found on Florian's web page, and
specific for your firewire setup (e.g. substitute "usb" by "ohci1394"
on
RTIRQ_NAME_LIST configuration setting). Ah, installation assumes a LSB
distro and it probably needs change for your particular case.
Feel free to ask if you get puzzled :)
Cheers.
--
rncbc aka Rui Nuno Capela
rncbc(a)rncbc.org
10:10 p.m.
On 1/19/07, Carlo Capocasa <capocasa(a)gmx.net> wrote:
I temporarily gave up on putting the device onto
another IRQ and will
leaving that up to the kernel developers to sort out;
Shared IRQs are a property of the hardware. There's nothing the
kernel can do about it.
what I am
interested in right now is how to increase the priority of IRQ 10 on my
system.
Florian suggested simply increasing the priority of the 'IRQ 10'
interrupt handling process; but on my system, there is no interrupt
handling process 'IRQ 10', or for any other IRQ, for that matter.
Are you sure you are running the -rt kernel? What is the output of
'uname -r' and 'ps auxww | grep IRQ'?
Lee
20 Jan
20 Jan
4:09 a.m.
Carlo Capocasa wrote:
Help much appreciated.
So you get a xrun every minute or so?
Did you make sure you disabled any cpu frequency modifying processes
(powernowd for instance)?
And as someone else mentioned, you are on a low latency kernel, right?
What's the output of uname -r?
I think you should try shutting down services (on debian "sudo
/etc/init.d/some-service-thats-not-crucial stop") you don't need,
including webservers, mysql, cronjobs, network, powersaving, etc. Do is
systematically one at the time, while checking if your problem goes away
or at least gets better. When you found out which service(s) is trigging
the problem you can probably look at the whole system setup (incl
interrupts) again and figure out exactly what causes the trouble. With a
bit of luck you can then fix the real problem and enable the service(s)
again and be fine.
--
peace, love & harmony
Atte
http://www.atte.dk | quintet: http://www.anagrammer.dk
| compositions: http://www.atte.dk/compositions
12:22 p.m.
Carlo Capocasa wrote:
Hi all!
Florian Schmidt was kind enought to point out that I can increase my
system performance by increasing the priority of both the Jack daemon
and the IRQ that my trusty'ol bebob device is hitched to.
I temporarily gave up on putting the device onto another IRQ and will
leaving that up to the kernel developers to sort out; what I am
interested in right now is how to increase the priority of IRQ 10 on my
system.
Florian suggested simply increasing the priority of the 'IRQ 10'
interrupt handling process; but on my system, there is no interrupt
handling process 'IRQ 10', or for any other IRQ, for that matter.
How can I increase the priority of 'IRQ 10' on my Debian based system?
On my FC5 system, I put this in /etc/rc.local :
/usr/bin/chrt -f -p <desired priority> `pidof "IRQ <desired IRQ>"`
Note: those are "back-ticks" encasing the "pidof" command, not single
quotes.
Thus, your command should be something like :
/usr/bin/chrt -f -p 90 `pidof "IRQ 10"`
HTH,
Rick
4:29 p.m.
This has been a very timely discussion. I just got my son a new mobo, CPU
and RAM, and he's been having a lot of trouble using jack, ardour and hydrogen
until I followed the advice here and tuned his machine up. Thanks all!
On Sat, 20 Jan 2007 12:21:58 -0500
Rick Wright <riwright(a)vt.edu> wrote:
Thus, your command should be something like :
/usr/bin/chrt -f -p 90 `pidof "IRQ 10"`
Here's a novel twist on all of this: my sound and video cards don't seem
to exist on my FC5 system, so I can't run this command!
For example, my audio card is a Delta 1010. lspci -v shows it:
02:0b.0 Multimedia audio controller: VIA Technologies Inc. ICE1712 [Envy24] PCI
Multi-Channel I/O Controller (rev 02)
Subsystem: VIA Technologies Inc. M-Audio Delta 1010
Flags: bus master, medium devsel, latency 64, IRQ 11
I/O ports at df80 [size=32]
I/O ports at dfa0 [size=16]
I/O ports at df60 [size=16]
I/O ports at df00 [size=64]
Capabilities: [80] Power Management version 1
However, it's not in /proc/interrupts, and trying to use chrt as shown
above fails:
[root@xtc ~]# cat /proc/interrupts
CPU0 CPU1
0: 67992 2883426 IO-APIC-edge [........N/ 0] pit
1: 8 0 IO-APIC-edge [........./ 0] i8042
8: 1 0 IO-APIC-edge [........./ 0] rtc
9: 0 1 IO-APIC-level [........./ 0] acpi
12: 96 0 IO-APIC-edge [........./ 0] i8042
14: 6779 89 IO-APIC-edge [........./ 2] ide0
15: 28172 6252 IO-APIC-edge [........./ 2] ide1
17: 2 0 IO-APIC-level [........./ 0] ehci_hcd:usb1
18: 31 0 IO-APIC-level [........./ 0] uhci_hcd:usb2
19: 1276 0 IO-APIC-level [........./ 0] skge
NMI: 0 0
LOC: 2951050 2954633
ERR: 0
MIS: 0
[root@xtc ~]# chrt -f -p 82 `pidof "IRQ 11"`
sched_getscheduler: No such process
failed to get pid 82's policy
Hmmm, where's the process for IRQ 11??
[root@xtc ~]# ps -ef | grep IRQ
root 33 27 0 15:28 ? 00:00:00 [IRQ 9]
root 249 27 0 15:28 ? 00:00:00 [IRQ 8]
root 268 27 0 15:28 ? 00:00:00 [IRQ 12]
root 296 27 0 15:28 ? 00:00:00 [IRQ 14]
root 299 27 0 15:28 ? 00:00:00 [IRQ 15]
root 350 27 0 15:28 ? 00:00:00 [IRQ 1]
root 838 27 0 15:28 ? 00:00:00 [IRQ 17]
root 860 27 0 15:28 ? 00:00:00 [IRQ 18]
root 897 27 0 15:28 ? 00:00:00 [IRQ 19]
root 1133 27 0 15:28 ? 00:00:00 [IRQ 6]
I'm really baffled here! Has anyone seen this? My video card (listed
as having IRQ 18) is equally missing in action as well. Thanks for any
clues sent this way.
--
======================================================================
Joe Hartley - UNIX/network Consultant - jh(a)brainiac.com
Without deviation from the norm, "progress" is not possible. - FZappa
4:41 p.m.
I hate following up on my own post, but...
On Sat, 20 Jan 2007 16:28:56 -0500
Joe Hartley <jh(a)brainiac.com> wrote:
Here's a novel twist on all of this: my sound and
video cards don't seem
to exist on my FC5 system, so I can't run this command!
For example, my audio card is a Delta 1010. lspci -v shows it:
02:0b.0 Multimedia audio controller: VIA Technologies Inc. ICE1712 [Envy24] PCI
Multi-Channel I/O Controller (rev 02)
Subsystem: VIA Technologies Inc. M-Audio Delta 1010
Flags: bus master, medium devsel, latency 64, IRQ 11
However, it's not in /proc/interrupts, and trying to use chrt as shown
above fails:
I did a "service alsasound restart" and it now shows up in /proc/interrupts:
17: 2 0 IO-APIC-level [........./ 0] ehci_hcd:usb1, ICE1712
I'm not thrilled that it's sharing an IRQ, time for some BIOS playing I think.
--
======================================================================
Joe Hartley - UNIX/network Consultant - jh(a)brainiac.com
Without deviation from the norm, "progress" is not possible. - FZappa
5:53 p.m.
Joe Hartley wrote:
This has been a very timely discussion. I just got my
son a new mobo, CPU
and RAM, and he's been having a lot of trouble using jack, ardour and hydrogen
until I followed the advice here and tuned his machine up. Thanks all!
On Sat, 20 Jan 2007 12:21:58 -0500
Rick Wright <riwright(a)vt.edu> wrote:
You can check for the pid with:
ps -Leo pid,pri,rtprio,cmd | grep IRQ
The first column is the pid, the third is the rtprio, which is what gets
set by the "chrt" command I posted previously. This is a good check
that you are able to set the "rtprio" value.
Once you have that pid, do:
/usr/bin/chrt -f -p 82 <pid>
The back-ticks and quotes are only required when this command is put in
a script.
Thus, your command should be something like :
/usr/bin/chrt -f -p 90 `pidof "IRQ 10"`
Here's a novel twist on all of this: my sound and video cards don't seem
to exist on my FC5 system, so I can't run this command!
For example, my audio card is a Delta 1010. lspci -v shows it:
02:0b.0 Multimedia audio controller: VIA Technologies Inc. ICE1712 [Envy24] PCI
Multi-Channel I/O Controller (rev 02)
Subsystem: VIA Technologies Inc. M-Audio Delta 1010
Flags: bus master, medium devsel, latency 64, IRQ 11
I/O ports at df80 [size=32]
I/O ports at dfa0 [size=16]
I/O ports at df60 [size=16]
I/O ports at df00 [size=64]
Capabilities: [80] Power Management version 1
However, it's not in /proc/interrupts, and trying to use chrt as shown
above fails:
[root@xtc ~]# cat /proc/interrupts
CPU0 CPU1
0: 67992 2883426 IO-APIC-edge [........N/ 0] pit
1: 8 0 IO-APIC-edge [........./ 0] i8042
8: 1 0 IO-APIC-edge [........./ 0] rtc
9: 0 1 IO-APIC-level [........./ 0] acpi
12: 96 0 IO-APIC-edge [........./ 0] i8042
14: 6779 89 IO-APIC-edge [........./ 2] ide0
15: 28172 6252 IO-APIC-edge [........./ 2] ide1
17: 2 0 IO-APIC-level [........./ 0] ehci_hcd:usb1
18: 31 0 IO-APIC-level [........./ 0] uhci_hcd:usb2
19: 1276 0 IO-APIC-level [........./ 0] skge
NMI: 0 0
LOC: 2951050 2954633
ERR: 0
MIS: 0
[root@xtc ~]# chrt -f -p 82 `pidof "IRQ 11"`
sched_getscheduler: No such process
failed to get pid 82's policy
Hmmm, where's the process for IRQ 11??
[root@xtc ~]# ps -ef | grep IRQ
root 33 27 0 15:28 ? 00:00:00 [IRQ 9]
root 249 27 0 15:28 ? 00:00:00 [IRQ 8]
root 268 27 0 15:28 ? 00:00:00 [IRQ 12]
root 296 27 0 15:28 ? 00:00:00 [IRQ 14]
root 299 27 0 15:28 ? 00:00:00 [IRQ 15]
root 350 27 0 15:28 ? 00:00:00 [IRQ 1]
root 838 27 0 15:28 ? 00:00:00 [IRQ 17]
root 860 27 0 15:28 ? 00:00:00 [IRQ 18]
root 897 27 0 15:28 ? 00:00:00 [IRQ 19]
root 1133 27 0 15:28 ? 00:00:00 [IRQ 6]
I'm really baffled here! Has anyone seen this? My video card (listed
as having IRQ 18) is equally missing in action as well. Thanks for any
clues sent this way.
24 Jan
24 Jan
8:56 a.m.
New subject: [linux-audio-user] Re: Tuning Jack: Please help me prioritize IRQ 10
Duh, that was really silly. Sorry to bother you guys about something
that turns out to be so trivial, it really seemed like legit issue to me...
Apparently, I was using a non-patched kernel thinking it was a realtime
kernel.
Carlo
Carlo Capocasa wrote:
Hi all!
Florian Schmidt was kind enought to point out that I can increase my
system performance by increasing the priority of both the Jack daemon
and the IRQ that my trusty'ol bebob device is hitched to.
I temporarily gave up on putting the device onto another IRQ and will
leaving that up to the kernel developers to sort out; what I am
interested in right now is how to increase the priority of IRQ 10 on my
system.
Florian suggested simply increasing the priority of the 'IRQ 10'
interrupt handling process; but on my system, there is no interrupt
handling process 'IRQ 10', or for any other IRQ, for that matter.
How can I increase the priority of 'IRQ 10' on my Debian based system?
Help much appreciated.
Carlo
PS: upping JackD priority alone took care of 60% of xruns; thanks Florian!
6334
days inactive
6339
days old
linux-audio-user@lists.linuxaudio.org
8 comments
6 participants
participants (6)
-
Atte André Jensen -
Carlo Capocasa -
Joe Hartley -
Lee Revell -
Rick Wright -
Rui Nuno Capela