I am a little bit sceptical about this project.
1) It doesn't seem to implement shaders and shaders are likely to make TitaniumGL bloody slow if such support is ever added. llvmpipe is shaders-only. There is no algorithm specifically optimized for some fixed-function pipeline configuration in llvmpipe.
2) I guess TitaniumGL does not strive for OpenGL correctness, which makes it not a viable permanent replacement for any GL driver.
3) TitaniumGL may be using X11 to accelerate some operations.
Regarding the Phoronix article:
A) Different compositing managers are used for TitaniumGL and llvmpipe (why?).
B) There is the nouveau DDX with TitaniumGL, but only Vesa with llvmpipe (why?).
C) The applications can take different actions for each driver, because they see that TitaniumGL has only GL 1.4 and llvmpipe has GL 2.1 and a half of 3.x features.
All in all, TitaniumGL is a nice project and has its purpose, but I don't think its code would be any useful to Mesa because of the points (1) and (2).
1) It doesn't seem to implement shaders and shaders are likely to make TitaniumGL bloody slow if such support is ever added. llvmpipe is shaders-only. There is no algorithm specifically optimized for some fixed-function pipeline configuration in llvmpipe.
2) I guess TitaniumGL does not strive for OpenGL correctness, which makes it not a viable permanent replacement for any GL driver.
3) TitaniumGL may be using X11 to accelerate some operations.
Regarding the Phoronix article:
A) Different compositing managers are used for TitaniumGL and llvmpipe (why?).
B) There is the nouveau DDX with TitaniumGL, but only Vesa with llvmpipe (why?).
C) The applications can take different actions for each driver, because they see that TitaniumGL has only GL 1.4 and llvmpipe has GL 2.1 and a half of 3.x features.
All in all, TitaniumGL is a nice project and has its purpose, but I don't think its code would be any useful to Mesa because of the points (1) and (2).
Comment