Nouveau Companion 31
Thunderbird decided to go on a quest of reverse engineering TV-OUT. He added a request for testers to our testers wanted page.
jb17some dared to fight the perils of XvMC. Although no life sign has been spotted from him on IRC, his Wiki page got an update. According to this update, he wrote a small program running with behaves like it would be using XvMC. That is, input is put into a buffer, registers are written and the final result is a frame, which could be displayed. Comparing both output and register status, his program generates the same results as the nv driver does under same circumstances. Please note, that this test program does not display anything.
Hm, let's switch back to marcheu and see what he is doing: Seems to be busy writing the Gallium3D framework for older cards. Let's try a bit later then.
After ahuillet's being defeated by the NV3x team in the race to claim A8+A8 (and thus EXA acceleration) working first, he was eager to avoid being called a "looser" for not being able to complete that "simple" task ;)
Seriously though, that task was far from being simple. To accelerate the PictOpt, multiple conditions (e.g. width limits, destination position limits) had to be met. After talks with IronPeter and Marcheu about what could be done with the register combiners ahuillet implemented an ugly hack which after some iterations of tests and patches did work.
A day later, he claimed that 1) everything was working and 2) that he had refactored that hack into something readable. Confirmations about EXA working came in from p0g and AndrewR.
A few days later we got reports that Xv wasn't working on PPC and that there were some issues with font corruption. As marcheu was hunting for a plethora of PPC bugs (see above), it was decided that we would first kill the basic bugs off and if the problems with Xv would still persist at that time, ahuillet would be the one to debug the Xv problems on PPC. However, the fonts problems are high on ahuillet's TODO list.
Darktama did some clean ups of the DRM and took the chance to introduce a 64 bit bug which prevented nouveau to work on x86. Marcheu found the problem and fixed it a few hours later.
jkolb is still working on TTM.
So perhaps we can get a final release date for the 2D driver from marcheu. (Let's try to ask him... oh, ok). Seems he is working on refactoring Xv code into a generic part and a card specific part. That should help us adding shader based Xv code for NV4x and NV5x more easily. In addition, we need to address those PPC bugs mentioned above... Seems it will take a bit more time until we can announce 2D working.
In the name of the Nouveau team, I would like to thank all testers who joined our IRC channel and offered help. Some problems were fixed, some are being worked on. With some problems requiring much iteration from problem to patch to (new) problem, some testers showed real dedication.
stillunknown is looking for 7xx0 card owners who would be willing to run a few tests for him. Furthermore, he is looking for people with dual head (whatever combination of VGA /DVI) who are willing to test randr1.2 / mode setting.
Marcheu needs testers with a NV30 (not later models) as things are a bit different on that card.
And if your are interested in helping Thunderbird and his TV-Out cause, please have a look here. This page will contain help / test requests of other project members too. So please have a look at it, we always need testers!
If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal or Stripe tips are also graciously accepted. Thanks for your support.