Intel Continues Working On The Mesa GLSL Compiler

[phoronix]

Phoronix: Intel Continues Working On The Mesa GLSL Compiler

Beyond working on initial Ivy Bridge support and continuing to fortify the Sandy Bridge support, Intel's OSTC team is still working on improving the GLSL compiler for Mesa...

http://www.phoronix.com/vr.php?view=OTQwMg

[pingufunkybeat]

This is very important. GLSL is the major missing part for OpenGL 3 support in Mesa.

AFAIK, there is minor work remaining for GLSL 1.30 compliance (basically weeks of work remaining, according to http://dri.freedesktop.org/wiki/MissingFunctionality), which is the major missing part for OpenGL 3.0

Most of the stuff needed for OpenGL 3.1 - 3.3 is already there, but the tough part is GLSL 1.40 and 1.50. That's why this work is so important, and it's important that Intel is working on it.

[elanthis]

Just to clarify on pingu's post though, there will still not be OpenGL 3.x support in any version of Mesa in any of the major distributions any time soon due to the patents on texture compression and floating point framebuffers, both of which are required for a lot of apps that depend on OGL 3, so the FOSS drivers will still have problems running many commercial (and even some FOSS) applications long after every other OGL 3.x and 4.x extension is implemented.

Sigh.

[Kano]

Time estimation is about 10-15 min to compile with a fast intel quad core cpu for all needed components (ddx+full mesa) + install a current kernel with Kanotix. 32 bit only is 3-4 min even faster. That would enable of course float textures - if you don't need that you can just install libtxc_dxtn, if you copy+paste the parts from my script you need maybe 2 min to do that, then you have got texture compression...

[pingufunkybeat]

Just to clarify on pingu's post though, there will still not be OpenGL 3.x support in any version of Mesa in any of the major distributions any time soon due to the patents on texture compression and floating point framebuffers, both of which are required for a lot of apps that depend on OGL 3, so the FOSS drivers will still have problems running many commercial (and even some FOSS) applications long after every other OGL 3.x and 4.x extension is implemented.

Correct.

We will go from "we can't do that at all" to "we can do it if you jump through some hoops and it's legally questionable". Or in other words, we'll be in the LAME and FreeType territory.

It's still very important progress, though, and definitely preferable to the current situation.