On Thu, 4 Dec 2003, Chris Ison wrote: >I realize replies will be resticted under NDA, but I am unable to refine >my questions until I can find the answers to the following questions in >relation to ATI docs some developers have access too. > >1. Do the ATI docs outline or detail expected PCI access process for the >r200 chipset, and if so, how close does the code match to it.
The ATI docs pretty much document the complete operation of the hardware for all intents and purposes. Not sure how better to answer your question, as it is kindof vague IMHO. >2. Do the docs specify expected process of initializing the card for >both AGP and PCI and how close does the code match this. Again, kindof a vague question IMHO. >3. Are all registers mentioned in the docs also in the code. If you mean, is every single register in the Radeon specifications present in radeon_reg.h and actually used somewhere in the driver, the answer is definitely a resounding no. >4. From the published register information (under the NDI) is the detail >enough to know if some registers need special handling, and if so, is >this being done in the code exactly as detailed in the docs. What registers? Obviously the Radeon docs were used to write the driver, and are consulted by those who have them when fixing bugs and when adding new features, however nobody is a perfect programmer 100% and so there is definitely flaws in the driver, or it would be bug free. Again you're being very vague here. >5. Is there anything in the documentation in reguards to PCI or anything >effecting PCI that is not implemented in the code Yes, it's entirely likely. PCI support was implemented for Alpha initially, and was never tested or contracted to work on anything else. SInce then various people (including yourself IIRC) have debugged different problems and got PCI support working better. There are definitely still bugs in the PCI support however, so that means that there are parts of the driver or kernel module missing code that is needed in order to make the cards function properly. You could interpret that as meaning the driver does not follow the documentation 100%, or it's possible that the docs are missing needed info, or it's much more likely that nobody has been interested enough to just debug and pound on the PCI support and work out any quirks on all hardware supported. >and if so, is it possable to implement without breaching the NDA. Fixing bugs in the driver is always possible without breaching NDA. >6. Is there anything that is not implemented in the code that exists in >the docs. For complete total support of all features on all Radeon hardware? Yes, absolutely. The OSS driver doesn't support anywhere near the complete functionality that the Windows drivers support on Radeon hardware, and probably never will, both from lack of volunteers to do the work, lack of people to fund the work, and various other factors. >If so is the NDA preventing the implementation or has it been >a case of time constraints. Implementation of what specifically? There are some advanced 3D features which are not implemented, which would require permission from ATI to do, but just as importantly would require someone wanting to actually do it, and dedicate time to do it, and that generally only happens when there is money involved, and big business contracts riding on it, such as the R200 3D support funded by The Weather Channel. >I think that covers everything I wish to ask at this stage, and >incase you are wondering, I am willing to help solve the mystery >of the slow pci-gart, and the first place would appear to be >comparing the docs to whats in the code and find any difference >in implementation no matter how small. But since I don't have >access to these docs I need the help of those that do. THe first step to me, would be to oprofile things, and find where the bottleneck is, then try to determine why. I doubt it would be that hard, and I doubt it requires the docs, although having the docs is always better than not having them when troubleshooting. If there are specific registers you don't understand and think might be a problem, or any other code construct, feel free to ask, and also to stick around in #dri-devel. While nobody can provide you the docs except ATI, we can probably clarify some uestions you might have about radeon hardware operation et al. IMHO, it's best to approach this from a direct concrete angle of a specific problem though, rather than a general problem of PCI not working properly and wanting to go over every inch of the driver anc compare register usage to the manual. The latter IMHO is a huge waste of time. -- Mike A. Harris ------------------------------------------------------- This SF.net email is sponsored by OSDN's Audience Survey. Help shape OSDN's sites and tell us what you think. Take this five minute survey and you could win a $250 Gift Certificate. http://www.wrgsurveys.com/2003/osdntech03.php?site=8 _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
