After the discussion in this thread, I decided to have a look on the internet to see what problems other VIA686b boards had. Interestingly, some of the problems sounded a lot like problems we have with the A1. Some x86 boards had random lockups when using bt8x8 based cards. IDE data corruption was also reported when DMA was enabled. There was also mention of crackling in audio from SB Live cards when the system was under load.

Suggestions to deal with these problems were things such as:
- The SB Live doesn't like sharing IRQs, make sure it has its own IRQ (clears up the crackling on some systems)
- Make sure the AGP graphics card has its own IRQ or your system may lockup randomly

Additionally, bios updates for these motherboards gradually solved many of the problems.

Looking at the IRQs, my SB Live card, the Radeon 9000 and the onboard AC'97 all use IRQ 11. The onboard sound isn't a problem as I its disabled (I have one of the A1-XE's with the sound chips missing). To see if it would make a difference, I tried to change the SB Live's IRQ but the Radeon 9000's IRQ changed to match.

So, my question is, can I set my graphics card to have its own IRQ independently of the last PCI slot's IRQ? All 3 PCI slots are in use so I can't just move the sound-card. If I did then the graphics card would share IRQs with a different card, which is not what I'm trying to achieve.

NOTE: My machine works fine. I'd just like to change the IRQs to see if it has any effect.

Hans

Last edited by Hans on 25-Mar-2006 at 12:40 AM.

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Hans

Looks like no-one knows if this is possible. I was hoping that one of the UBoot developers would see this and give some input. Ah well, it's not like it's important.

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Hans

Quote:

Ah well, it's not like it's important.

Yes it -is-.

Have a look at this thread for more info:
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=18235&forum=33

Also, a search for "interrupt" on AW will pull up more interesting recent threads
on this topic.

A great deal of discussion was initiated after June, 2005 with this post #109.

#6

Last edited by number6 on 25-Mar-2006 at 02:57 PM.

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@number6

Thanks for the links. It probably is more important than I originally thought. But for me at least, my machine works fine most of the time, so I'm interested more out of curiosity really. The only real issues I have at the moment are:
- I get occasional lockups with Warp3D applications when I have interrupt=yes set in Picasso96. This may well be due to a bug in the driver and I can work-around this one until the next OS4 release.
- Linux suffers from occasional lockups (using the 2.6.8 dma-fixed kernel) but we already know that Linux has problems that will take a while to fix.
- Linux will lockup within 10 seconds when using my BT878 tv-tuner card with no valid signal when Overlay is diabled. It's fine when using overlay. The TV tuner is still unusable in AOS4.

I thought it would be worth trying to change the IRQ settings to see if it has a positive effect on these problems. For now, I can work around these problems.

Ok, so seeing as others may be having more trouble with their machines than me, does anyone know if we can change the AGP bus' IRQ? Even if it doesn't make a difference in the end, it's work looking in to.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Hans

You may find this thread interesting as well, since it discusses mouse and
radeon in relation to interrupts.(audio/internet is well documented elsewhere)
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=18122&forum=14

#6

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@number6

In that thread you just linked to, you said:

Quote:

I have heard the Teron design PCI interrupt C is shared with Radeon 7000. Do you know whether this is the case with all Radeon or just the 7000 in the Micro?

I guess I just confirmed that as this is exactly what happens with my Radeon 9000. I hope that this is set by the firmware, not the hardware. Otherwise the graphics-card will always share an IRQ with the card in the last slot. That would mean that what I'm asking in this thread is impossible.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Hans

Summary of sorts for audio/internet and interrupts here in post #21.
Also an additional question is raised about "SIL controller" and interrupts.

Edit(Added):In post #3 of this thread I linked to an email response from George
Breese. Here is an informative link which also makes reference to George's work:
http://techreport.com/onearticle.x/2441
Note that writings on this subject have circulated for about 5-6 years and these
writings (acknowledging the issues) come from experts in the field.

I have posted a great deal today on this subject, and I hope that this puts to
rest once and for all any discussion concerning problems with OS4, that -clearly-
originate elsewhere.

#6

Last edited by number6 on 25-Mar-2006 at 06:12 PM.

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@Hans

Sorry, no solution yet. I'm just bumping this thread back into recent topics in the hope that one of the Uboot devs may see it and answer the question. Only a few people are likely to know enough about UBoot to provide an answer and it's possible that they never saw this thread.

So, final call. Does anyone know the answer to my question regarding the graphics-card and IRQs?

Hans

Last edited by Hans on 28-Mar-2006 at 02:51 PM.

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Hans

I'm not very sure about this issue, but here's my take anyway (mostly thinking aloud, here)...

Quote:

Looking at the IRQs, my SB Live card, the Radeon 9000 and the onboard AC'97 all use IRQ 11

In my case (default IRQ settings: IRQ A/B/C/D = 9/10/11/7), pciscan says that my Silicon0680 controller uses IRQ #9 (I have it in the first 33 MHz PCI slot) and that my SB Live! uses IRQ #11 (it resides in the third PCI slot). At the same time, pciscan says that my RADEON 9000 uses IRQ #11, which, besides the SB Live!, is used for VIA AC97.
The A1 works just fine.
I'd say that we have similar setups (if not identical).
I have no idea whether it is possible to change the AGP slot's IRQ, so I can't help you out.

saimo

Last edited by saimo on 28-Mar-2006 at 03:46 PM.

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@saimo

Excuse my ignorance, but my understanding was that 686B handles PCI/AGP interrupts
and passes them to Articia S as NMI (non-maskable interrupt).
Then Articia S, in turn, passes the NMI to PPC with considerable latency, which
might explain some of the issues people have.
Please correct me if I am wrong about this.

#6

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@number6

I'm even more ignorant than you (as you can see, I have now modified the post you answered to (#9) as thoughts cleared up)
This is the first time I hear about those details... thanks for providing them!

This is how the "mapping" of IRQs looks like on my machine:

  Device:           0x0660 (Articia S Host Bridge)
  Interrupt:        Line 0x00, Pin 0x01

  Device:           0x0661 (Articia S PCI Bridge)
  Interrupt:        Line 0xFF, Pin 0x00

  Device:           0x9200 (3c905C-TX/TX-M [Tornado])
  Interrupt:        Line 0x07, Pin 0x01

  Device:           0x0686 (VT82C686 [Apollo Super South])
  Interrupt:        Line 0x00, Pin 0x00

  Device:           0x0571 (VT82C586A/B/VT82C686/A/B/VT823x/A/C PIPC Bus Master IDE)
  Interrupt:        Line 0xFF, Pin 0x00

  Device:           0x3038 (VT82xxxxx UHCI USB 1.1 Controller)
  Interrupt:        Line 0x05, Pin 0x04

  Device:           0x3038 (VT82xxxxx UHCI USB 1.1 Controller)
  Interrupt:        Line 0x05, Pin 0x04

  Device:           0x3057 (VT82C686 [Apollo Super ACPI])
  Interrupt:        Line 0x00, Pin 0x00

  Device:           0x3058 (VT82C686 AC97 Audio Controller)
  Interrupt:        Line 0x0B, Pin 0x03

  Device:           0x3068 (AC'97 Modem Controller)
  Interrupt:        Line 0x0B, Pin 0x03

  Device:           0x0680 (PCI0680 Ultra ATA-133 Host Controller)
  Interrupt:        Line 0x09, Pin 0x01

  Device:           0x0002 (SB Live! EMU10k1)
  Interrupt:        Line 0x0B, Pin 0x01

  Device:           0x7002 (SB Live! MIDI/Game Port)
  Interrupt:        Line 0x00, Pin 0x00

  Device:           0x4966 (Radeon RV250 If [Radeon 9000])
  Interrupt:        Line 0x0B, Pin 0x01

  Device:           0x496E (Radeon RV250 [Radeon 9000] (Secondary))
  Interrupt:        Line 0xFF, Pin 0x00

Last edited by saimo on 28-Mar-2006 at 04:04 PM.

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@saimo

AFAIK a device in the AGP slot of the A1 can use the interrupt lines IRQA and IRQB, so it should be possible to use another interrupt. In theory you could force the graphics card and the soundcard to use its own interrupt and share the other available interrupt lines between the rest of the PCI devices (all PCI interrupt lines have to be level triggered, so this is possible).

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Geri

Quote:

AFAIK a device in the AGP slot of the A1 can use the interrupt lines IRQA and IRQB

Uhm... if so, how is it possible that my RADEON 9000 uses the IRQ #11 (which I have assigned to IRQC)? Maybe my reading of pciscan's output is mistaken (i.e. Line != IRQ number)?

Quote:

so it should be possible to use another interrupt

Unfortunately, unless I'm missing something, U-Boot does not seem to offer a way to do this.

Quote:

In theory you could force the graphics card and the soundcard to use its own interrupt and share the other available interrupt lines between the rest of the PCI devices (all PCI interrupt lines have to be level triggered, so this is possible).

Well, at least in my case, this would be very easy: all I'd need is setting IRQC number to 10... but would it be safe?

saimo

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@saimo

In several older threads where we used either kdebug or sashimiPPC, we ran tests
and discovered many "delayed" or "missing" IRQs.
It is my understanding that NMIs also cannot be delayed.(VIA site mentions this)
Perhaps when something else needs "time" available that is -not- available,
we see these "delayed" and "missing" IRQs.
It is important to note that depending on the user's system and perhaps, more
importantly, -what- resources are being used at the time...one person can see
a problem, while the next does not.

Also a concurrent discussion we were having about interrupts is here.

#6

Last edited by number6 on 28-Mar-2006 at 06:00 PM.

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@saimo

Quote:

Uhm... if so, how is it possible that my RADEON 9000 uses the IRQ #11 (which I have assigned to IRQC)? Maybe my reading of pciscan's output is mistaken (i.e. Line != IRQ number)?

You're not mistaken. My graphics-card uses IRQC as well. I get the same IRQ numbers using kinfocenter in Linux so you're reading them correctly.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@number6

Quote:

It is my understanding that NMIs also cannot be delayed.

Well, it does make sense

Quote:

Perhaps when something else needs "time" available that is -not- available, we see these "delayed" and "missing" IRQs.

I guess that the Friedens and other AOS developers (Stephane Guillard, for example) should have quite a lot of information about this issue... I'd be pretty surprised if not, given that AOS runs quite well

Quote:

It is important to note that depending on the user's system and perhaps, more importantly, -what- resources are being used at the time...one person can see a problem, while the next does not.

Yes, it's rooted in the nature of this issue.
However, not rarely I happen to play shoucast/mp3 streams while browsing and doing all sorts of things (sometimes also compiling...), so many of the resources involved (audio, video & network) are working at a steady and significant rate, and still I cannot think of an apparently reasonless lockup despite the shared IRQ numbers... (not that I complain about it ).

Quote:

Also a concurrent discussion we were having about interrupts is here.

I'll have a look at it, thanks!

saimo

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@Hans

Quote:

You're not mistaken. My graphics-card uses IRQC as well. I get the same IRQ numbers using kinfocenter in Linux so you're reading them correctly.

Thanks for the feedback... but it makes me wonder: what if I set IRQC to 10? Would it change for just the card in the corresponding PCI slot or also for the RADEON in the APG one?

saimo

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@saimo

Quote:

saimo wrote: @Hans Quote: You're not mistaken. My graphics-card uses IRQC as well. I get the same IRQ numbers using kinfocenter in Linux so you're reading them correctly. Thanks for the feedback... but it makes me wonder: what if I set IRQC to 10? Would it change for just the card in the corresponding PCI slot or also for the RADEON in the APG one?

I've already tried this and the AGP IRQ changes as well. This is why I'm asking if it's possible to change the AGP IRQ independently. Not having a card plugged into slot C would work, but I'm currently using all my PCI slots so that's not an option.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Hans

Quote:

I've already tried this and the AGP IRQ changes as well. This is why I'm asking if it's possible to change the AGP IRQ independently.

Ah, right... you even explained it in your first post... sorry for getting circular

saimo

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@saimo

Quote:

Uhm... if so, how is it possible that my RADEON 9000 uses the IRQ #11 (which I have assigned to IRQC)? Maybe my reading of pciscan's output is mistaken (i.e. Line != IRQ number)?

Well, AFAIK interrupt lines are normally routed different for every AGP/PCI slot. For example:

PCI0 PCI1 PCI2
IRQA IRQD IRQC
IRQB IRQA IRQD
IRQC IRQB IRQA
IRQD IRQC IRQB

So IMHO it is should be perfectly possible that the interrupt line pin does not always match the used IRQ number.

Please correct me, if I'm wrong (I'm not so familiar with the interrupt handling on the A1).

Quote:

Well, at least in my case, this would be very easy: all I'd need is setting IRQC number to 10... but would it be safe?

I would say yes, because PCI interrupts are level sensitive/triggered.

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -