The irregular Nouveau-Development companion
Issue for March, 17th
Well, what can we say? We did make two mistakes obviously, which caught us totally off guard:
- We asked for some specific dumps in the last TiNDC
- We added a news entry about a list generated by JussiP, which shows which cards are dumped and which we still miss. That somehow got things rolling. Digg.com took it up and linuxfr.org too, as well as a post on the ubuntu forum. And then our channel got swamped with questions on how to do dumps for us. PQ, Pmdata and marcheu struggled to help them but it was to much, marcheu surrendered first never to be seen again (on that day at least) while pq and pmdata had to give up on their intentions to do more development.
We really don't know what to say but "Thank you!". It amazes us again and again how eager the community is to help us. As of 12.03.2007 we had 426 new dumps for various cards (even 8800 cards).
If you want to participate in the Google SoC please read SoC and the help section further down. Please contact us before 24.03.2007.
Our upload script from gmail to sf.net broke somewhere during the last two weeks. This became obvious, when the above mentioned torrent of dumps started rolling in. Kmeyer who developed the script was heard saying something like "bandwith too small" before he threw the towel. Marcheu picked up the script and got it to work on his system, so after a delay of about 4 days after we noticed the breakage, updates of the dump directory started working again. But still, the dump script didn't work correctly only finding a very small subset of the dumps sent. So kmeyer once again started debugging and saw problems with some different email clients using weird headers. After some more work, the script was finally working again (this is the main reason for the 426 new dumps).
This was important due to the fact that JussiP created a page at http://users.tkk.fi/~jpakkane/ren/ where the status of the dump directory is displayed. No changes there for a long time and a note on the front page saying "too much dumps" seemed somehow suspicious :) But since 15.03.2007 the status page should reflect our real dump status.
Under the flood of dumps renouveau showed some problems: On PCI-e systems the tool is very likely to crash. Bug hunting revealed that the newest NVidia driver remaps some IO registers on the fly, while renouveau still looks for the old and no longer valid mappings. A bug fix by marcheu didn't help and currently we can only ask you to downgrade to driver 8776 or earlier in order to avoid these problems.
Some people reported OpenGL bugs ("Latest OpenGL error: xxxx") which on some system even led to driver errors (Xid: ....) of the NVidia blob. Pmdata, KoalaBR and pq volunteered to fix the tests, so that the OpenGL errors are no longer displayed. jb17bsome researched the reasons for the Xid errors and it seems that in some cases trying to find an object in RAMIN via INSTANCE_MEM() would lead to this error. As of writing, this issue is still unresolved but under investigation by jb17bsome and marcheu.
We now have our entries in the fd.o Bugzilla. If you got problems with our driver in either 2D or 3D please do report them in there. It doesn't hurt, if you do mention your problem in the channel too, but don't expect immediate help or action. The devs will take notice however! You can find Bugzilla here[https://bugs.freedesktop.org/ ].
Airlied has updated the randr1.2 branch to include all the fixes from the master branch. His testing of the new code showed some regressions however, which may be a result from either merging or his own development. Further testing and bug fixing is hampered by a lack of time on airlied's side, but slow progress can be seen.
On 09.03.2007 Nvidia released version 1.99 of the XOrg nv driver. It now offers G80 support. Shortly after the release darktama, Thunderbird and Marcheu checked the source code for new types of information. Work needs to be done to deobfuscate this code and add it to our code base.
On the MmioTrace front all branches were merged into stable, so all your favorite features can now be found there.
Thunderbird and Skinkie are trying to reverse engineer TV-Out on NVidia cards. Some registers for functions like adjusting flickering, overscan, image positioning were found but still a lot of work needs to be done. Thunderbird knows quite a bit about the NVidia chips as he is maintaining and developing nvclock. He does evaluate the BIOS data / tables. On the other hand is Skinkie, who chooses to evaluate the data coming from MmioTrace.
pmdata and jwstolk did encounter serious problems with glxgears on NV15. As soon as the first fifo command was delivered to the hardware the system crashed. They tried to reduce the amount of commands, tried to insert delays (NOPs) into the fifo, but only when the fifo got "NOP" commands the system would work (hm, make that "not crash", obviously a No OPeration won't do much rendering work :) ). Work is still progressing.
We are currently trying to create some command line tools to help newbies create nouveau dumps automatically. A first experimental version should be published soon. You can find a first version here: http://www.ping.de/sites/koala/script/createdump.sh if you try it, let KoalaBR know how it worked for you.
Refer to our status page http://users.tkk.fi/~jpakkane/ren/ for renouveau dumps. Only dumps for cards listed in red are currently needed. We would be interested in SLI dumps though (and please set the Topic line of your email to hint at the fact that this is an SLI dump). G80 owner willing to test patches would be very welcome too.
Still MmioTrace dumps would be very welcome.
And the onslaught of people mentioned at the top really showed that we need some people in the channel who answer newbie questions. Mostly the answers can be found on the Wiki / FAQ.
If some people would like to take part in the Google Summer of Code, please do contact marcheu. We do have some slots available underneath the Xorg umbrella. It would be really nice if you could even present a project you want to do. Some choices can be found on our ToDo page, but feel free to come up with your own ideas. If you don't have any idea, we will certainly be able to supply one or two for starters. As time progresses, we will certainly add a page dedicated to SoC on the wiki.
And finally: Maximi89 would be very interested in translating the TiNDC's and the Wiki to spanish, too. However, he would really like to have some more people willing to help him there. Generally, we really would like to offer localized pages, so don't be shy and go ahead. And if you want to translate the TiNDC's too, just let me know, I am willing to send out my draft a day before I put them online, so that the translation aren't lagging behind.
And seeing that we slowly get a french translation too, I am asking myself, where are my fellow germans???
Ok, that's it again for this time. I hope you enjoyed reading it.