GLSL IR To TGSI Translator Ready To Be Merged Into Mesa

[przemoli]

I know about gl3.txt

but are not there any legal blockers? and are all of them a target for someone?

As a side note, contrary to what the article says, it will hopefully be merged to Mesa master before the merge window closes at the end of next week, i.e. it should make it into Mesa 7.11.

That is very good news indeed. New GLSL compiler, new translator. Question is what is not __new__?

Ohh, small question how did you tested Your code?

[V!NCENT]

Great job, man!

Next up:
Playing Duke Nukem Forever through Wine on Gallium 3D with ATI cards onn Hurd with Gnome 3 on Wayland, while listening to Dr. Dre's Detox album in 2012 (year of Linux desktop).

The future is upon us and the world still hasn't ended, lol xD

[Plombo]

(For anyone not aware, I am the developer of the GLSL IR to TGSI translator.)

Woah. That guy actually delivered? Someone hire him.

OK, that's it - the man gets a gold star!

Yay, kudos to the guy! These ventures are usually "Hey guys, I've just started programming and I'm going to make a Vi clone, it's going to be rad, excellent and have lots of great features! I haven't written any code yet, but I'm about to start. It's going to be great!". But it seems like this guy delivered!

Code reviews are always great (that's one of the things that FOSS is about, right?) so it's going to be interesting to see what happens when the maintainers have a look at it.

The developer feedback has already started, actually.

As a side note, contrary to what the article says, it will hopefully be merged to Mesa master before the merge window closes at the end of next week, i.e. it should make it into Mesa 7.11.

Hey OpenGL3.0 would be so sweet. It's definitely about time!

Agreed.

Benchmark or didn't happen

There might be some performance gains, but if you're expecting a groundbreaking difference, you'll be disappointed. Converting from Mesa->TGSI had no penalty at all because they use almost the exact same opcodes and conventions, and converting from GLSL->TGSI has many of the same lost optimization opportunities as GLSL->Mesa IR did.

But once I finish incorporating the developer feedback to get it merged, I'm planning to work on improving the code generation for the things that ir_to_mesa didn't handle well.

[Azpegath]

This was done by Plombo, a poster here, right? Good work!

Anyway, the GL3.txt status file is always the best source for status updates: http://cgit.freedesktop.org/mesa/mesa/tree/docs/GL3.txt

The major thing missing for OpenGL 3.0, 3.1, 3.2 and 3.3 are GLSL 1.30, 1.40, 1.50 and 3.30, respectively.

The other stuff is mostly already in place. Actually, we're almost there, but GLSL is a bitch.

Aah thanks, that's the one I was referring to in my post.

[TeoLinuX]

Benchmark or didn't happen

I totally agree

[pingufunkybeat]

This was done by Plombo, a poster here, right? Good work!

Anyway, the GL3.txt status file is always the best source for status updates: http://cgit.freedesktop.org/mesa/mesa/tree/docs/GL3.txt

The major thing missing for OpenGL 3.0, 3.1, 3.2 and 3.3 are GLSL 1.30, 1.40, 1.50 and 3.30, respectively.

The other stuff is mostly already in place. Actually, we're almost there, but GLSL is a bitch.

[Azpegath]

I'm sure that if focused OGL 3.0 can be acomplished, but have all legal problems been resolved, or bypassed?

Maybe some small Phoronix article, about state of implementing newest OGL's i Mesa?
Like who is working on it, what have been done, what are milestone targets, legal problems, etc.

+1. Michael, you have a very good grasp about what is going on in the Mesa camp, could you sum up what's to be done. Not only technically I mean, but just like above, also different issues.
I've seen the feature matrix, but I can't find it now except http://wiki.x.org/wiki/RadeonFeature

[przemoli]

Will there be OGL 3.0?

I'm sure that if focused OGL 3.0 can be acomplished, but have all legal problems been resolved, or bypassed?

Maybe some small Phoronix article, about state of implementing newest OGL's i Mesa?
Like who is working on it, what have been done, what are milestone targets, legal problems, etc.

[Azpegath]

Yay, kudos to the guy! These ventures are usually "Hey guys, I've just started programming and I'm going to make a Vi clone, it's going to be rad, excellent and have lots of great features! I haven't written any code yet, but I'm about to start. It's going to be great!". But it seems like this guy delivered!

Code reviews are always great (that's one of the things that FOSS is about, right?) so it's going to be interesting to see what happens when the maintainers have a look at it.

OpenGL3.0 would be so sweet. It's definitely about time!

[RealNC]

Benchmark or didn't happen