For option "b", the OpenGL side already has some work done: http://www.cs.toronto.edu/~andreslc/xen-gl/ The D3D side is, as everything involving Microsoft, more problematic. Wine is your frind, and if you say its "too buggy", you can't get anything better at the OSS word, and they are doing a heck of a job if you ask me.
Some time ago I read about making 3D work on virtual machines, and the guy said there were two options: Virtualize the hardware or virtualize the API, course the second is easier, and thats what VMGL does, its intercepts OGL calls and forward them to the server. So the VirtualCard can be anything, and it will just forward things between host and guest. First option is a lot harder, as you need to write the FULL OGL and D3D implementation/Driver and virtualize all the calls to teh PCI/AGP/PCIE card, and, for now, i don't see any "real" benefit on doing it. So my vote? Go with the virtualized API and let people play ;) On 4/8/07, Alexey Eremenko <[EMAIL PROTECTED]> wrote: > hi all ! > > since there are many questions - so I would like to write : "why there > won't be a 3D support anytime soon" > > The problem: > To get a 3D support, you have to have a virtual 3D hardware and > several drivers - 1 for OpenGL and 1 for Direct3D. > > Possible solutions: > a. Emulate a real 3D card such as GeForce - it solves the drivers > problem automatically - but this card is closed-source hardware, so we > can't simply emulate it, + it will be damn slow. > This is very very difficult to do technically, so probably, it won't happen. > > b. Use virtualized hardware, such as Virtualbox 3D accelerator - and > write 3D drivers for that one. Remember - both OpenGL and Direct3D > drivers are needed for several guest OSes. > > The OpenGL problem can be resolved theoretically easier - we could run > virtualized-Mesa3D on top of host's OpenGL and we are done. > The DirectX problem is much more difficult - we will have to translate > DirectX instructions into OpenGL instructions, like Wine does - but > the problem with Wine is that it's too buggy and many games won't run. > So it must be debugged along the way. > This will allow to reach 99% of native speed, so this is very fast solution. > > solution b is Very Difficult task as well, but more realistic than a. > > Innotek also don't see the need to work on 3D virtualization. VMware > were working on it since about 2005, and already have BETA product > that supports 3D, called VMware Fusion. > > Innotek seems no to have any plans to even start such a project, so my > take is that it won't happen at least until 2010... > > -- > -Alexey Eremenko "Technologov" > > _______________________________________________ > vbox-users mailing list > [email protected] > http://vbox.innotek.de/mailman/listinfo/vbox-users > -- As coisas são como elas são porque é assim que as fizemos ser. _______________________________________________ vbox-users mailing list [email protected] http://vbox.innotek.de/mailman/listinfo/vbox-users
