ATI R200 Documentation Coming Too

[phoronix]

Phoronix: ATI R200 Documentation Coming Too

In the Phoronix Forums following Friday's release of the (NDA-free) ATI R300 3D register information, AMD's John Bridgman has confirmed that he is looking to release documentation going back to the ATI Radeon 8500 era graphics processors. Years ago ATI had released this R200 documentation to the open-source driver developers at the time, but it was encumbered by legal restrictions. John doesn't believe these R200 documents will be of any use to the xf86-video-ati developers for improving the Radeon 8500 to 9250 open-source support, but that it will help assist new open-source developers in beginning to understand graphics programming...

http://www.phoronix.com/vr.php?view=NjM5NA

[Uchikoma]

If I have to say, it appears that they're really pulling out all the stops here when it comes to the OSS effort.

[stan]

Wonderful!!!

Is there any hope of having r128 documentation released? Phoronix Forum member Svartalf mentioned that he has an NDA-ed version (http://www.phoronix.com/forums/showthread.php?t=6470&page=3) of RagePRO docs. Perhaps they can also be released?

[ssmaxss]

It will be great if they will release specifications for PowerPlay, as today support for PowerPlay on my mobility radeon with FOSS drivers is very limited.

[mattst88]

Wonderful!!!

Is there any hope of having r128 documentation released? Phoronix Forum member Svartalf mentioned that he has an NDA-ed version (http://www.phoronix.com/forums/showthread.php?t=6470&page=3) of RagePRO docs. Perhaps they can also be released?

I too would like to see the r128 docs released. I found copies of NDA'd version a long time ago. If I can find them, I would have to imagine that ATI could find them in their archives.

bridgman, if it helps locate the documents in ATI's archives, their filenames are:

R128REGRE.pdf - Rage 128 Register Reference Guide
RR45101C.pdf - Rage 128 Pro Register Reference Guide
rr47003oemp.pdf - Rage Mobility M3 Register Reference Guide (OEM)

[agd5f]

I too would like to see the r128 docs released. I found copies of NDA'd version a long time ago. If I can find them, I would have to imagine that ATI could find them in their archives.

bridgman, if it helps locate the documents in ATI's archives, their filenames are:

R128REGRE.pdf - Rage 128 Register Reference Guide
RR45101C.pdf - Rage 128 Pro Register Reference Guide
rr47003oemp.pdf - Rage Mobility M3 Register Reference Guide (OEM)

We have the PDFs, it's a matter of whether or not we can find the original *editable* versions so we can remove all the confidential language ;)

[Svartalf]

We have the PDFs, it's a matter of whether or not we can find the original *editable* versions so we can remove all the confidential language ;)

Heh...which is why I haven't breached my agreement with ATI over those files... :D

In all honesty, I don't think there's anything TOO problematic at this point in time in those files- but if you can't find the editables, you might want to see if legal will sign off on what was sanitized to the point of the original NDA'd FOSS releases to the Utah-GLX team.

[agd5f]

Heh...which is why I haven't breached my agreement with ATI over those files... :D

In all honesty, I don't think there's anything TOO problematic at this point in time in those files- but if you can't find the editables, you might want to see if legal will sign off on what was sanitized to the point of the original NDA'd FOSS releases to the Utah-GLX team.

It's not the content, but the confidential notices that are written all over the documents. Unfortunately, we can't release the documents with that language all over them.

[domi]

It's not the content, but the confidential notices that are written all over the documents. Unfortunately, we can't release the documents with that language all over them.

It may be possible to directly manipulate the content of the PDF. See CAM::PDF perl module for instance http://search.cpan.org/~cdolan/CAM-PDF-1.12/

HTH

[jnolck]

Me three. I might be wrong, but i remember there was 2d docs for the r128 without NDA's.

I know you guys are interested in pulling in new developers. Wouldn't the best way to do so be to release all docs for all graphics chips you've ever made? Let new developers start from the beginning?

This is not a request, it's just my dream come true. Docs from the mach32 all the way to r600 and beyond. Mach32 being important because programming it was talked about in Richard F Ferraro's book.

[NeoBrain]

I know you guys are interested in pulling in new developers. Wouldn't the best way to do so be to release all docs for all graphics chips you've ever made? Let new developers start from the beginning?

Don't wanna kill your dream, but I think there are more important docs to release apart from the ones of 10 year old graphics chips.

Anyways, it's still nice to see AMD/ATI keep its word and releases all docs the open source community needs (or even doesn't really need :D)

[SunTzuTech]

Phoronix: ATI R200 Documentation Coming Too

In the Phoronix Forums following Friday's release of the (NDA-free) ATI R300 3D register information, AMD's John Bridgman has confirmed that he is looking to release documentation going back to the ATI Radeon 8500 era graphics processors. Years ago ATI had released this R200 documentation to the open-source driver developers at the time, but it was encumbered by legal restrictions. John doesn't believe these R200 documents will be of any use to the xf86-video-ati developers for improving the Radeon 8500 to 9250 open-source support, but that it will help assist new open-source developers in beginning to understand graphics programming...

http://www.phoronix.com/vr.php?view=NjM5NA


I have a laptop with an IGP340M, and still use it
pretty heavily. If you can get them to release the
programming specs on the mobile R200 chipsets, that
would be a big help, as I'm still seeing one
particuarly nasty bug in Solaris X86/Xorg where
the Xserver goes infinite when I start Gnome. CDE
still works fine.

[mattst88]

Obviously the R600/R700 push is number one priority currently, but where do we stand on the release of R200 and Rage 128 docs?

[TechMage89]

Aren't the Rage 128 docs already out there?

Also, I believe most of the information in the docs has already been put to use in the drivers, so I'm not sure how much it would help for these models.

[bridgman]

I don't think the docs we have would help either of the drivers; the only benefit I see from them is as an introduction for new driver developers... assuming they can *find* an R2xx these days ;)

I believe there may be some features in the chip which aren't fully supported in the drivers but they aren't in those documents either. Given the ongoing pain with AGP there's an argument for writing a similar "so you wanna be a driver developer, huh ?" guide for a more recent chip instead, maybe a 5xx-family PCIE part. I always felt the docco would be more useful if it referenced actual sections in the open source driver rather than just saying "do this".

[bugmenot]

Apropos: I installed Debian Etch on a 600 mhz laptop yesterday, which has an ati chip inside. Some kind of rage foo... and I wasn't able to get 3d (I think because I could not get any DRI, according to the log). I red I had to compile the mach64 DRI driver or something like that, but after 3 hours or so I gave up. I could not compile it , there were always errors.

I asked myself why this did not work out of the box, I mean this is really old. On [1] I read that mach64 isn't build by default, I wanted to ask if that is still the case. And if it is, if one could not change this.

Or *should* 3d work actually out-of-the-box? So is this a a bug in Debian?
I know you have really a lot to do to get R600/700 3d working (I have also a radeonhd 3200, so I really like your work!) but I also like to hear a statement if so old chips are still supported.
(If you want to see something else, just take a look at this. Should be easy compared to complicated chips like the current ones.)
Would be great, if *every* chip produced by ati could still be supported.

Thanks for your hard work and come good to the next year. :)


[1] http://dri.freedesktop.org/wiki/ATIMach64?highlight=%28CategoryHardwareChipset%29

[chithanh]

mach64 drm is not in the upstream kernel and not shipped by all distributions, because it used to be insecure. It has been made secure some time ago but the developers did not push it upstream. On Gentoo, you can install it with VIDEO_CARDS="mach64" in make.conf and emerge x11-drm, on other distributions you may have to download and compile it manually.

[bugmenot]

mach64 drm is not in the upstream kernel and not shipped by all distributions, because it used to be insecure. It has been made secure some time ago but the developers did not push it upstream.

Thanks for the answer. :)

If it is secure now, why can't a dev just push it so that in future distributions it works ootb?
And since this is an ati driver, an ati dev could do this, doesn't he? This would make me really happy! (Could a dev take a look at this, please?)


Just imagine every chip has open source 3d support ootb *dream* :D
Wouldn't this be great?

[User0x45]

the only benefit I see from them is as an introduction for new driver developers... assuming they can *find* an R2xx these days ;)
Hi Bridgman, I am interested in helping debug the r2xx video adapter module I'm using in X.

I would love to read this r2xx doc to help me understand and describe more accurately what's going on with my machine.

What's the timeline for release?

[bridgman]

When we either find an editable copy (no luck yet) or have time to scan and OCR what we have back into something we can edit and release. Rough guess is a month before we have time to start work on it, maybe another month after that. Just guessing though... I was hoping an editable file would have turned up by now.

AFAIK everything covered by the document is already working in the R200 drivers, so there doesn't seem to be much to debug other than the usual AGP problems which we can't help with in our documents anyways -- most of the fixes really need to be in chipset drivers. Do you have an R200-based system that isn't working ?

[cb88]

would it be possible to realease docs to support the rage fury maxx I know someone that has one just collecting dust due to poor support for it?

Since my rage pro turbo can run openarena fairly playable I reckon the maxx would be worth releasing the docs on...


I kinda doubt it could run compiz though since mine can't not having the texture from pixmap extension (or does it since the rage 128 supports texture compositing it seems)

[agd5f]

would it be possible to realease docs to support the rage fury maxx I know someone that has one just collecting dust due to poor support for it?

Since my rage pro turbo can run openarena fairly playable I reckon the maxx would be worth releasing the docs on...


Unfortunately, I'm not sure the people who know or information about that card are around anymore.


I kinda doubt it could run compiz though since mine can't not having the texture from pixmap extension (or does it since the rage 128 supports texture compositing it seems)

The r128 lacks non-power-of-two textures which means no compiz.

[cb88]

Ah I see thanks for the reply

[highlandsun]

Unfortunately, I'm not sure the people who know or information about that card are around anymore.


The inevitable outcome for all closed-source projects... sad...

I'm grateful that AMD/ATI turned around and saw the light.

[bridgman]

AFAIK a Rage Fury Maxx is basically two Rage 128 chips on a single board with some complicated (and probably still proprietary) software to synchronize the cards and combine the results from the two graphics engines. Getting it to run as a single Rage 128 should be easy; getting it to use both GPUs was a huge job for us and doesn't really seem like something any right-minded individual would even want to consider ;)

Were you talking about getting it running with one GPU or with two GPUs ?

[cb88]

yeah i think it already works in single GPU mode though I'm not sure.. that is pretty pointless though

[Ant P.]

What's the state of V4L support on older cards? I've got a Rage 128 with S-Video in/out & TV tuner, that I gave up on trying to get working some time around 2006. Is it worth trying again?

[Pseus]

When we either find an editable copy (no luck yet) or have time to scan and OCR what we have back into something we can edit and release. Rough guess is a month before we have time to start work on it, maybe another month after that. Just guessing though... I was hoping an editable file would have turned up by now.

AFAIK everything covered by the document is already working in the R200 drivers, so there doesn't seem to be much to debug other than the usual AGP problems which we can't help with in our documents anyways -- most of the fixes really need to be in chipset drivers. Do you have an R200-based system that isn't working ?
Perhaps you can open them with the latest version of Adobe Illustrator? AFAIK, it can edit PDFs. Though I'm not sure that would work with such an old version, it might be worth a try :)

[bridgman]

What's the state of V4L support on older cards? I've got a Rage 128 with S-Video in/out & TV tuner, that I gave up on trying to get working some time around 2006. Is it worth trying again?

I'm not sure if we ever provided documentation for the capture path. The IP issues are a bit different there (scary) and I haven't really looked at them even for the latest GPUs. Most of the work on capture happened before I became involved; maybe agd5f or someone else can comment.

[chithanh]

When we either find an editable copy (no luck yet) or have time to scan and OCR what we have back into something we can edit and release.For OOo 3.0 there exists a PDF import plugin (http://wiki.services.openoffice.org/wiki/Pdf_Import_Extension) from Sun. If the document structure is not too complex, it could at least be sufficient for removing the watermark.

[User0x45]

Hey Bridgman,

a couple of months is fine. I can get going on this r2xx when they finally do come out.Do you have an R200-based system that isn't working ?Yeah, I wrote about it here: http://www.phoronix.com/forums/showthread.php?t=15028

Since posting, I haven't heard anything, or been asked any questions, so I'm kind of sitting tight. I'd love to work through it. Ultimately I could just add to the worlds E-waste...but you know I'm kind of persistant.

Thanks,

[bridgman]

Hmm. At first glance that doesn't look like anything docs would help with. The drm seems to be coming up ok, then the xord driver comes up ok and sets up drm ok but then gets that error message :

drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is 9, (OK)
drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is 9, (OK)
drmGetBusid returned ''
(II) [drm] DRM interface version 1.0
(EE) [drm] Could not set DRM device bus ID.
(EE) RADEON(0): [dri] DRIScreenInit failed. Disabling DRI.

I wouldn't wait for docs; there's probably enough info in the code already to figure this out. If nobody has an idea here then maybe post to the xorg mailing list and see if someone can suggest a good first step for debugging this.

[Wyatt]

I think I've run into this problem twice before. As I recall, it has something to do with the different DRM implementations (Kernel, X.org and ATi each seem to have one... :/ ). Looking at my own setup, disabling CONFIG_DRM in .config and installing X.org's libdrm lets things work in relative harmony. Sorry if you've tried that already, but I figure it doesn't hurt to share tips anyway. :)

Cheers,
Wyatt

PS: Tried to just PM this, but that's disabled. Apologies for the slight derail.

[agd5f]

Hey Bridgman,

a couple of months is fine. I can get going on this r2xx when they finally do come out.Yeah, I wrote about it here: http://www.phoronix.com/forums/showthread.php?t=15028

Since posting, I haven't heard anything, or been asked any questions, so I'm kind of sitting tight. I'd love to work through it. Ultimately I could just add to the worlds E-waste...but you know I'm kind of persistant.

Thanks,

I think it's an issue with the VIA agpgart driver. I don't think it's anything radeon docs would help with.

[User0x45]

Hi Wyatt,

disabling CONFIG_DRM in .config and installing X.org's libdrm lets things work in relative harmony. Sorry if you've tried that already, but I figure it doesn't hurt to share tips anyway. :)


No, I haven't tried this. Thanks, I'll try.

[kurukafa]

When we either find an editable copy (no luck yet) or have time to scan and OCR what we have back into something we can edit and release. Rough guess is a month before we have time to start work on it, maybe another month after that. Just guessing though... I was hoping an editable file would have turned up by now.

AFAIK everything covered by the document is already working in the R200 drivers, so there doesn't seem to be much to debug other than the usual AGP problems which we can't help with in our documents anyways -- most of the fixes really need to be in chipset drivers. Do you have an R200-based system that isn't working ?

Hi bridgman, any updates on this issue? Are we going to be able to develop for 9200 chipset without any obligations. ;-)

[bridgman]

Hi kurukafa; no update yet, probably next week before we get a chance to spend any time on this.

I just want to stress that all the programming info in the doc is already included in the current open source drivers, so this will not allow any more development or bug fixing. The R200 driver was written using this documentation, and the Xorg developers probably know the R200 better than we do these days. The sole purpose for releasing the doc is as an "interesting read" for someone who is just learning about GPUs.

The upside is that it might give a better understanding of how R200-class GPUs work, but the downside is that the understanding will be for a GPU which is now 5 generations old and which doesn't have much in common with currently shipping parts.

[kurukafa]

Hi kurukafa; no update yet, probably next week before we get a chance to spend any time on this.

I just want to stress that all the programming info in the doc is already included in the current open source drivers, so this will not allow any more development or bug fixing. The R200 driver was written using this documentation, and the Xorg developers probably know the R200 better than we do these days. The sole purpose for releasing the doc is as an "interesting read" for someone who is just learning about GPUs.

The upside is that it might give a better understanding of how R200-class GPUs work, but the downside is that the understanding will be for a GPU which is now 5 generations old and which doesn't have much in common with currently shipping parts.

Thanks for the update. I will still be waiting for the document, as it is not always fun to reverse engineer the open source driver. :-)

[tuxdriver]

Originally Posted by bridgman

...assuming they can *find* an R2xx these days

There are a lot of PCs still running (Linux) with old 9250,9200SE,9100 and 8500 cards.
You can even buy brand new 9200SE and 9250 cards in stores in some countries (shocking!)

They are still good enough to run Ubuntu with wobbly windows and other compositing effects :)

[leef]

Newegg still sells six 9250 cards. Only a year and a half ago, the r200 was the fastest 3D available for opensource linux. The r200s are generally silent (no fan) and low power.

[kurukafa]

Hi kurukafa; no update yet, probably next week before we get a chance to spend any time on this.

I just want to stress that all the programming info in the doc is already included in the current open source drivers, so this will not allow any more development or bug fixing. The R200 driver was written using this documentation, and the Xorg developers probably know the R200 better than we do these days. The sole purpose for releasing the doc is as an "interesting read" for someone who is just learning about GPUs.

The upside is that it might give a better understanding of how R200-class GPUs work, but the downside is that the understanding will be for a GPU which is now 5 generations old and which doesn't have much in common with currently shipping parts.

Rumors are wide-spread that the documents will be released REAL soon (in the orders of days). Any denials or otherwise supportive information regarding the issue will highly be appreciated.

Thanks.

[bridgman]

Nope, not much change here. I guess you can call that a denial.

[kurukafa]

Nope, not much change here. I guess you can call that a denial.

Rumors... :-(

[bridgman]

Yeah... sorry.