Some XDC update

Stefan Dösinger stefandoesinger at
Mon Apr 21 10:00:04 CDT 2008


Here's a short summary of the wine-related things that were going on on XDC:

-> Mouse input: Daniel Stone will take care of that. He says DGA would give us 
what we want, but recommends not to use it. He is working on extending 
XInput(or making a new XI2) which can report relative mouse movements. The X 
people want to phase out DGA. Some games use it for relative input, so 
they'll make a DGA emulation layered on the new XInput stuff. He will need 
debugging / testing help from us since Wine will most likely be the only App 
making use of that directly. Also we should not hold our breath, he has some 
other items on his TODO list, but the end of 2008 is a plausible ETA.

-> Tablets: Config issue is known and being worked on

-> Graphics, general: The Mesa and ATI people are willing to help out with our 
issues. I talked to Aaron Plattner and Pierre ? from Nvidia, and 
they know the issues and worked on some extensions for that internally some 
time ago, but then decided that they're not going to implement that for some 
reason they didn't tell me(I hope they re-consider that now that someone is 
asking for it though). The ATI guys warned me that it might take a while 
until any new extensions are available in production drivers, but that's  
something we have to put up with I guess. Also both the ATI and Mesa people  
like the idea of using our conformance tests for their driver testing and  
debugging. The Nvidia guys offered help with the multithreading crashes and  
the 32 vs 40 varying issue.
(I also talked to Ben Byer from Apple, but he's not in charge of the graphics 

-> Graphics: Flat shading: Picking that as a first case for working on an 
extension. Should be fairly easy to implement in Mesa since all cards have a 
specific register for exactly that issue.

-> Graphics, multithreading: general agreement that what we are doing is 
correct, and driver bugs need fixing. It looks like the crash in fglrx will be 
fixed soon(my test app worked in Matthew's development driver). For the 
Nvidia issues I'll file a bugreport as soon as I know if the glxSwapBuffers 
crash is a driver or Gentoo bug)

-> Graphics, sRGB blending: Everyone believes me that our emulation is 
slow("pow() - uh oh, don't use that, that's not parallelizeable"[Keith 
Packard]). The nvidia people are aware of the issue, but a fix is also 
unlikely. The ATI people aren't sure if their dx9 hardware supports sRGB 
writing, maybe the windows driver is emulating things as well - have to check

-> Graphics, GLSL issues: MOVA and LIT, possibly extensions for that.

-> Graphics, Hacks: Both the ATI and Nvidia developers warned me that their 
Windows drivers have a huge per-application bug workaround database. Aaron 
was pretty impressed that Wine doesn't have any app-specific hacks. I am 
afraid we will need them, as some apps seem to expect different behavior that 
our d3d9 tests get. I think the best way is to fetch some information about 
the Windows driver hacks and write tests which trigger them to show 
that "hl2.exe" gets different driver behavior than "d3d9_test.exe"(or in 
whatever way things work)

More information about the wine-devel mailing list