So, there was a lot of fuss a couple of months ago about the problems due to binary X11 drivers on distributions for nVidia and ATi cards, and the problems caused to non-binary users that required Xorg 7.1 to go stable only a while ago for x86 and amd64 architectures.
In that mess, I decided to buy a radeon card, a 9250, to get rid of binary drivers myself.. about €50, that, to be honest, I’d have saved if I knew I was going to experience the current troubles to get a new job.
Anyway, I was pretty okay with the compromises I needed to take, because they were mostly minor things, and I told myself that reporting them might make them being fixed soon. Retrospectively, this was an utopia.
Summarising the bugs I found, there’s the first big one, reported on FreeDesktop bugzilla as bug #7431; basically when you enable EXA, Xv does not work on windows moves.. well actually it performs as badly as X11, which is the reason I started the bug as “fail”; the issue is there for both the computers I have the radeon driver on, and it’s the main reason I’m not using EXA at all (although it performed decently with composite, better than nVidia’s driver). I still had no news about that bug, but oh well, I can wait, after all I don’t have a solution myself, so I can just wait.
The second issue is that when you move an Xv window on the border of the two heads on MergedFB, the image is replaced by the Xv colorkey for the whole window… I was aware of this bug before buying the card, thanks to leio, and I decided it wasn’t much of a bug. Of course it’s annoying when I want to move windows around while debugging, and there’s the problem with VLC going fullscreen and taking a pixel column of the second head, dropping the whole output… not much of a problem, but I was pretty depressed when Donnie laughed at me when I said that bug was a bit annoying.. it’s not always possible to avoid putting a window between the two screens, and in those cases, having this fixed would help. nVidia had no such problem.
The third issue is probably a generic limitations, that might even depend on the card, I’m not sure. On nVidia drivers I could open two Xv-outputting software, like xine/kaffeine and Tvtime, and have them both display at the same time. Right now if I have Kaffeine open, even in the system tray, tvtime simply refuses to start. I think nothing can be done on this at this time tho, so I just shut up on this.
Tonight I found a fourth issue, that’s a bit more annoying, although the fact that I found it only after two months I have the card shows how much it counts for me (almost nothing). While trying to use libcaca’s OpenGL output I got nothing but a black window, on the second head. Joseph told me it should have been working there, but playing around I found the problem happened about at 2⁄3 of the second screen.. the result is that the hardware support is only for the first 2048×2048 pixels, which means a slice of the second head is non-working, considering I have a 2560×1024 right now.
Yes, of course these issues are minor, and the driver works, and I should be happy, and I am, and I’d like to thank all the developers who worked on the radeon driver for their results. But don’t start tell me that’s as good as the nvidia binary drivers, and that the issues are fixed in a timely fashion… there are issues, although minor, you can say it’s perfectly usable, that suites 90% of the users out there, the ones that don’t play 3D games all the day long, but cannot say it’s (yet) flawless.
It is definitely a good idea for who don’t like binary drivers, it’s definitely a good idea for single-headed boxes, it’s definitely a good idea to try it out if you want to hack at it. That said, it’s not free of issues, so you can understand that some people don’t recommend it.. That said, I’ll continue using it and be happy 🙂
You’re right, nobody pays people full-time to work on the open-source Radeon driver, and there aren’t any significant contributions coming in outside of a very small group of people working on it. Also, most people working on it don’t really care about Xv.
“The third issue is probably a generic limitations, that might even depend on the card, I’m not sure.”Xv in the radeon driver is implemented via overlays, as you are aware from the fact that a color key is shown when the output is on both heads. However, this means that because most radeon cards only have one overlay in hardware, only one Xvideo port can be open at a time.For example Xglx gets around of that by implementing xv with using FBO or pbuffers, if passed appropriately on the command line with it’s accel switches.Kaffeine might want to consider closing the Xv port when it’s not needed, so that other apps can use it meanwhile (but then when one does when kaffeine wants to use it too, it can’t).mplayer is a bit sensible and fallbacks to a different video outputting method after saying:Could not find free Xvideo port – maybe another process is already using it.Close all video applications, and try again. If that does not help,see ‘mplayer -vo help’ for other (non-xv) video out drivers.