Yes it would be nice if X.org could use OpenGL directly for it's display and composition, but to date, nobody has made this possible. Is it wrong for a business to make it so? Since when does developing software for GNU products mean that they aren't allowed to do it privately? If Novell is developing XGL behind closed doors, and paying the developers to build it... Where's the problem?
To view parent comment, click here.
To read all comments associated with this story, please click here.
Member since:2005-07-06
First, there may be more X 3D drivers than you realize. All of the cards supported by Mesa can be converted to Xegl with moderate amounts of work. The current chip families supported by Mesa (and also X since X uses Mesa) are: fb, ffb, gamma, i810, i830, i915, mach64, mga, r128, r200, r300, radeon, s3v, savage, sis, tdfx, trident, unichrome. This covers most of the common low end 3D hardware. Note that fb is in the list, Mesa (and Xegl) will work on dumb framebuffer and entirely implement OpenGL in software.
The second group is high end hardware with no open source drivers. Nvidia and ATI for example. Xegl treats OpenGL as a device driver. These two vendors would probably take their Windows OpenGL implementation, port them to Linux and then add the 14 EGL entry points.
So it is possible to get Xegl coverage for most of the 3D hardware out there. It's just going to take some coding and evangelism work.
But the X driver developers have chosen to chase the 2D EXA bandaid. This will probably consume their efforts for a year or two. After that they may start looking at Xegl again. The cards getting EXA support are the same ones as are in the Mesa support list so we aren't picking up any new hardware support with EXA (I think there is an exception this, ajax implemented EXA on number9 for the five people left in the world with number9 hardware).
At the end of all of this you just have to ask, what did we gain by chasing EXA for a couple of years if in the end we are going with Xegl anyway?
Member since:
2005-07-14
I don't want to criticize you Mr. Smirl because you know far more about this than I do (actually you are a hero of mine for you write-up on the subject), but after reading your "State of the Linux Desktop" on the matter and looking into myself I see a different problem for Linux's desktop than you do.
Rayiner is accurate in saying that the driver work involved in fully finishing EXA was equivalent to the work needed to get the Mesa drivers in shape for XGL. The developers just chose the EXA route.
It seems to me that the real problem after messing with many cards and systems is that most of the display drivers for Linux suck. It seems we only have two drivers that can actually do the job that OSX and Vista does- the closed Nvidia drivers and the ATI 92xx's. In your write-up on the State of the Linux Desktop you seemed to shrug off this problem (if I remember you suggested that people without should just go buy a cheap 9250 or do without) but it seems to me to be the central roadblock to a nice desktop!
Most people do not have a ATI 92xx card and most people won't buy one just to use Linux. Also many laptops do not have cards that can be replaced. Many people have Nvidia cards, but it seems that many open source developers would rather keep the Linux Desktop primative if the other alternative is to build it on top of Nvidia's drivers. All the other Linux drivers are crap (cept maybe Sis?)- they were created for the Linux Server where just displaying ANYTHING is good enough.
So you say "that the driver work involved in fully finishing EXA was equivalent to the work needed to get the Mesa drivers in shape for XGL" but I don't know if I agree. You admit that EXA is a bandaid- it does not matter if not every card supports it. Those people will just be left out of what is planned for EXA, which so far to me seems to only be eye candy. The old Xserver is still there for those without an Nvidia or 92xx card. But if the Xserver moved over to Xegl the display drivers (all of them) would HAVE TO support Xgl before a major distro could use that as a default or no display at all right? If I am wrong, please tell me...you are the one that would know!
To get the drivers ready (for non 92xx and Nvidia cards) it seems to me that an entire overhaul of all of them is needed. What will that take- three, four years? And thats just if everyone who can helps. But how long is it taking to move the few drivers that aren't crap over to EXA? Well the ATI 92xx is already there and the closed Nvidia driver will be done soon now that Xorg 7 is out.
Vista and OSX will stay years ahead because they have decent drivers. Yet Desktop Linux (which itself is a side effect of the Linux server) and its crap open source drivers are years behind on that count.
Do you really think that overhauling every driver out there to support Xegl would really take less time than to hack the three or so drivers that are currently decent enough to accerate the desktop to use EXA? It seems to me this is the problem EXA solves with its band-aid approach- it gives the Linux community some way to say "we have an accerated desktop" without lying (even though it only applies to maybe 30% of computers) while the big job of remaking all the other drivers that are currently crap can go on over the next five years.
It seemed that in your write up you do not focus on the driver issue...but it seems to be at the heart of the problem to me. Most Linux display drivers suck compared to what Windows and OSX has....would EXA or Xgl really change that?
Please correct me if I am wrong- you ARE the authority on this as far as I am concerned.