Performance Work Coming Up For Mesa 7.11

[airlied]

That may be valid points, but the pro-ABI people also have valid points. You can't call people idiots just because you disagree with them. Name calling means it's you who doesn't have arguments.

The pro-ABI people have 0 valid points. Since if they were valid points we would rationally take them on board. They can't see past the short-term issue they are trying to solve, hence show a distinct lack of ability to learn, hence idiots.

Dave.

[RealNC]

The pro-ABI people have 0 valid points. Since if they were valid points we would rationally take them on board. They can't see past the short-term issue they are trying to solve, hence show a distinct lack of ability to learn, hence idiots.

Dave.

Of course "short term" in this case means 20 years... and counting. Yeah, very "short term"

[KDesk]

The pro-ABI people have 0 valid points. Since if they were valid points we would rationally take them on board. They can't see past the short-term issue they are trying to solve, hence show a distinct lack of ability to learn, hence idiots.

Dave.

By the way you write, it is easy to see what kind of person you are. Insulting others won't make you smarter or be right.

[sandain]

By the way you write, it is easy to see what kind of person you are. Insulting others won't make you smarter or be right.

In this case he is right (he is the expert here).

The insults stem from the fact that this ABI topic has been discussed ad nauseum here and elsewhere. Certain individuals, who have failed to educate themselves, continue to beat this dead horse. As always, dealing with these people gets old, and tempers run short, hence the insults.

For those people obsessed with a stable ABI in Linux, please read up on the issue at hand. If you still feel that Linux needs a stable ABI, feel free to fork, because I doubt you will change the mind of the development community at large.

[RealNC]

In this case he is right (he is the expert here).

The insults stem from the fact that this ABI topic has been discussed ad nauseum here and elsewhere. Certain individuals, who have failed to educate themselves, continue to beat this dead horse.

We did get educated. We simply don't agree and have a differing opinion.

If that makes me an idiot, so be it and right back at you.

[?John?]

Robustness or backwards compatibility? Make your choice!

Notice that every piece of code out there that tends to suck horribly at what it does (be it Winblow$ or X - no offense to X.Org developers intended because they're doing really great under the circumstances) is so damn broken mostly because it has to maintain some sort of backwards compatibility.

The problem with that is pretty obvious - if you have to keep the interfaces stable at all cost, then you can't really fix anything and you always inevitably end up with an inconsistent pile of junk held together only by a crapload of duct tape. Ask yourself if you were still willing to keep using GNU/Linux if it really went down that road at some point in the future.

The thing is you can't just sit down and try to design a good, future-proof API because stuff just keeps changing and a new idea that turns everything upside-down can come at any time. It's called evolution and it's based on the "adapt or die" principle so you can either break backwards compatibility and properly redesign the whole thing or try hacking it up somehow and make a complete mess of it in the process.

Many people don't understand this and prefer to sacrifice the means for the ends - in another words they're saying "I don't mind you crapped all over my floor as long as it stays slippery". They just don't realize that with blobs they're really shooting their own feet in the long run. One of the perfectly valid reasons why is this: it gives hardware manufacturers the ability to force you into buying new stuff you would otherwise neither need nor want (because what you already have still works 100% and is more than adequate for intended purposes) by simply killing support for stuff you're using any time they see fit.

As far as I'm concerned, everything I need works at least well enough (and keeps getting better almost by the day) and there's not much more than that to play on GNU/Linux anyway so I'll gladly sacrifice OpenGL features and performance for the sake of lightning fast GUI (I just can't stand fglrx slugishnes!), robustness and security provided by libre drivers any day.

That means BIG THANKS TO EVERYONE INVOLVED because you managed to make at least one person perfectly happy with the results of your hard work.

[V!NCENT]

OK... Let's assume there's a stable API. Is AMD and nVidia going to suddenly rewrite the X.org implementation? No? Next...

[pejakm]

An interesting optimization for r600 is here!

http://cgit.freedesktop.org/mesa/mesa/commit/?id=8c631cfeae29b5236928f759e222aa35e6e4984c

This commit which was made last night really improved my framerate so you can really "feel" it. For games as well as for applications (kwin effects feel a bit more smooth, stellarium almost tripled framerate, etc.)

Just wanna say: THANKS!!!

[chrisr]

Yeah, I know what you mean.

Just wanna say: THANKS!!!

Personally, I've been sitting in a little puddle of happiness since this commit went in .

[kernelOfTruth]

whoa !

wicked - how fast compiz got (I'm always running "low" - the lowest performance state so it's quite noticable)

thanks Marek & all the other devs involved