Originally posted by HokTar
View Post
Announcement
Collapse
No announcement yet.
Progress On The ATI R600g Gallium3D Driver
Collapse
X
-
One think I'd be interested to know is : do the devs find Gallium easier to make a driver than classic Mesa or not?
Since you have now some experience in writing two drivers both in Gallium and in Classic Mesa for the same devices, do you find Gallium3D a real improvement, or basically just a different architecture?
Comment
-
So what about bugs/regressions should one file bug reports for r600g already? I just tried latest git master and now vdrift eats all my memory. It looks like:
Code:commit 66e4cb1cd5a55402606a09417349d2be8b009e89 Author: Jerome Glisse <[email protected]> Date: Mon Aug 30 17:56:59 2010 -0400 r600g: avoid dynamic allocation of states
Comment
-
Originally posted by rvdboom View PostOne think I'd be interested to know is : do the devs find Gallium easier to make a driver than classic Mesa or not?
Since you have now some experience in writing two drivers both in Gallium and in Classic Mesa for the same devices, do you find Gallium3D a real improvement, or basically just a different architecture?
Where gallium gets messy is in the corner cases where nobody has defined stuff and you have to figure out what should be done in state tracker vs pipe driver vs winsys, but that stuff is mostly getting easier as more people develop drivers.
The other thing that helps now and makes writing drivers easier is piglit as once you have the basic driver structure you can use piglit to target things to do and also to stop regressions sneaking in as you go.
Dave.
Comment
-
Originally posted by monraaf View PostSo what about bugs/regressions should one file bug reports for r600g already? I just tried latest git master and now vdrift eats all my memory. It looks like:
Code:commit 66e4cb1cd5a55402606a09417349d2be8b009e89 Author: Jerome Glisse <[email protected]> Date: Mon Aug 30 17:56:59 2010 -0400 r600g: avoid dynamic allocation of states
Dave.
Comment
Comment