Subject: Re: [linux-audio-user] Anyone using multiface+cardbus?
From: Mark Knecht (markknecht_AT_attbi.com)
Date: Sun Mar 23 2003 - 17:45:52 EET
Ron,
OK, most import item - if it works, then have a beer and make some
music! There is no reason to get overly concerned about this stuff if
the machine does it's job.
I think your machine is using the newer 'APIC' and 'IO-APIC' type
interrupt controllers. I actually got brave and joined the kernel
developer's reflector to ask some questions about this as I was not sure
(and am still not) about IRQ priorities in this hardware configuration.
To be clear, this is hardware that is part of a newer PC architecture
and is NOT the same as most machines out there in the world. (I.e. -
it's been around for a year now, but it's new, and likely to work well
for most PC uses, but less sure about audio applications.) Don't get
worried, as I know a number of people using it on Pro Tools and doing
fine, although many of them had problems getting it to work well in the
beginning.
Also, I do not have a machine with this hardware to do any testing
on, so I cannot be sure about any of this. We'll just have to figure it
out.
My rudimentary understanding of this configuration is that the
hardware priority is lowest to highest, except for the normal,
out-of-order issue about IRQs 3-7. Since those are apparently not
assigned on your box, I hope they don't matter.
This would imply, in your case, that one or your SCSI controllers
trumps your RME, and both trump the ICE1712, which I presume is an
on-board sound chip. (16 is before 17 is before 18 is before 19)
I do not 'think' that this will be a big problem as SCSI controllers
are normally pretty well designed and have good buffering.
I'm a bit more concerned about the mouse at IRQ 12 trumping all of
them. This _has_ been a problem on the normal IRQ setup and _could_ be a
problem for you. (I'm just guessing.)
To test this, I'd suggest starting jack using the RME with a buffer
size of 64 and verbose logging, but no recording. I'd then try moving
the mouse around violently. Lots of mouse movement will cause lots of
mouse interrupts, and you'll get an idea how the system is handling all
this interactive load.
If that works, then I'd do the same thing with Ardour doing playback,
but not recording, and see how mouse movements effect it, and then
finally the same while doing recording and playback.
Also, remember that as PC's get fast and faster, and if the interrupt
handler's are written better and better, then it could get to the point
that IRQ order no longer makes a difference. I don't think we're there
yet, but I do think we're getting closer.
Drop me a note if you need some more specific info.
Cheers,
Mark
On Sun, 2003-03-23 at 07:00, R Parker wrote:
> Hi,
>
> What do you make of my IRQs, they're not consistent
> with what you're describing. Is this some newer bios
> aipc feature?
>
> The pci cards begin at 16 with two scsi controlers
> sharing interupt 16, at 18 we see a third scsi
> controler that isn't the same as the one on 16. It's
> just the same driver.
>
> There's an NVIDIA graphics card in an AGP slot, soon
> to be replaced with a radeon. Is the NVIDIA the "LOC"
> entry?
>
> The scsi HDD controlers, interupt 16, are a lower
> interupt than the rme9652 so I'm assuming they recieve
> priority over the audio card. I'm guessing two HDD
> controlers requesting CPU time slices aren't the most
> optimal scenario.
>
> I am comfortable running jackd with larger buffers
> because I use an external digital consol and monitor
> inputs for studio and control rooms from the consols
> inputs. So latency isn't as big an issue for me as it
> might be for engineers that don't have external
> consols. Reguardless, I'd like to tune this box to its
> optimal performance capability.
>
> If the AGP interupt can't be reordered, then I assume
> the correct strategy is to use "setpci" or the bios to
> adjust latency. Make sense? I'd probably opt for using
> "setpci" because it enables resetting latency on the
> fly--I intend to do some video work.
>
> I'd be curious to know other people's opinions of my
> irq ordering.
>
> bash-2.05b# cat /proc/interrupts
> CPU0 CPU1
> 0: 3729010 3949857 IO-APIC-edge timer
> 1: 8714 9043 IO-APIC-edge keyboard
> 2: 0 0 XT-PIC cascade
> 12: 206994 212633 IO-APIC-edge PS/2 Mouse
> 16: 212646 216114 IO-APIC-level sym53c8xx,
> gdth
> 17: 1234920 1245793 IO-APIC-level rme9652
> 18: 3721 4640 IO-APIC-level sym53c8xx
> 19: 364290 368882 IO-APIC-level ICE1712,
> eth0
> NMI: 0 0
> LOC: 7678904 7678970
> ERR: 0
> MIS: 0
>
> ron
>
> --- Mark Knecht <markknecht_AT_attbi.com> wrote:
> > On Sun, 2003-03-23 at 01:59, Jaakko Prättälä wrote:
> >
> > > What could I do if I can't set irq's in my bios?
> > >
> > Jaakko,
> > Hi. First, what machine are you running and what
> > BIOS does the
> > machine have? Phoenix? Award? Usually you can set
> > IRQ's, but often the
> > settings are hard to find. In my machine it's under
> > Advanced->PCI
> > Configuration. Even though I have this feature, I
> > still recommend that
> > you should:
> >
> > 1) Remove all of your cards EXCEPT graphics, boot
> > and look at interrupt
> > settings. Write them down or save a file somewhere.
> >
> > 2) If you have on-board devices, like sound or LAN,
> > then you should
> > attempt (through BIOS if available) to put them on
> > low-priority
> > interrupts. (3,4,5,6,7)
> >
> > 3) Add a card, such as your HDSP that drives the
> > multi-face, on maybe
> > PCI slot 3, boot the machine, and look at what
> > interrupt each device is
> > on. If you get 9 or 10 for the HDSP, then you're in
> > pretty good shape.
> > If you do not, then again, set the PCI slot to IRQ 9
> > if available, or
> > try a different slot with just the HDSP and no other
> > cards. (Always
> > graphics installed, obviously!)
> >
> > 4) Go through this process, adding a single card,
> > booting, and looking
> > at which IRQ each card has, until you get your
> > important sound card/s on
> > the best IRQs. Again, IRQ order is:
> >
> > 0,1,8,9,10,11,12,13,14,15,3,4,5,6,7
> >
> > 2 does not exist. (Or it's called 'cascade'.)
> >
> > 0,1 & 8 are not available.
> >
> > 9 is best, followed by 10, etc.
> >
> > 10 is as good as 9 if 9 is not used, or only used by
> > acpi support.
> >
> > Networking works fine on 3-7.
> >
> > I hope this helps. Feel free to write directly if
> > you have other
> > questions.
> >
> > Cheers,
> > Mark
> >
> > 3) NOTE: On my machine, BIOS seems to move the USB
> > interrupts around as
> > I chan
> >
>
>
> __________________________________________________
> Do you Yahoo!?
> Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
> http://platinum.yahoo.com
This archive was generated by hypermail 2b28 : Sun Mar 23 2003 - 17:39:48 EET