ATI Catalyst 10.8 For Linux Brings OpenGL ES 2.0

[smitty3268]

opendrivers + opengl renderer in wine = very very choopy graphic, unplayable
opendrivers + GDI = almost ok, but still choppy, cant play on battlenet..

catalyst + GDI = slow, same as open+gdi
cat + opengl = very good performance, almost like native version

That makes me wonder what kind of openGL wine is generating. I don't think the original starcraft had any 3D bits in it, it was all 2D sprites.

[NomadDemon]

yes i doesnt, but opengl renderer set in regedit do this effect

[Dandel]

A defualt Linux system without FGLRX = A real stable Linux system .

A default Linux system with FGLRX = Windows ME .



And I'm not joking.

Yet another FUD argument... I see more than 6 days uptime, as compared to windows ME when i used that where i was crashing left and right.

KDE 4.4.5 (squeeze)+fglrx= crap. Especially when you want to use iceweasel (firefox) you often see black. Even other windows are often with a black boarder. It is really extreme that ati ONLY tests with ubuntu, when compiz mostly works they are happy. KDE 4 could not be called supported, also the composite tests always fail with fglrx. What is ati thinking? When you have got a hd 5 card you will definitely not get oss 3d support with a current distro right now. fglrx has to work, but in such a bad shape it hurts. Basically ati should be sued for such bad drivers - they never worked for KDE 4 correctly as long as i test squeeze. The only thing that was fixed was tesselation with OpenGL.

Really, this is the type of stuff that needs to be reported with the linux feedback url provided on the downloadload page. Also, don't forget to open up the bug report on the unofficial bug reporters website...

unofficial bug reports: (take your pick, these are all related to KDE v4.4.5)
HD5470: Windows contents stop updating after some time
HD5470: Changing brightness from KDE PowerDevil doesn't work
window background corruption
v-sync's regression in catalyst 10.8

LOL, this is hilarious.



Are they going to optimize the Linux part of their drivers now for a useless benchmark that isn't really a benchmark anyway?

btw. XvBA still broken on Evergreen.

The 6% optimization is not really that important at a glance, but it probably was just repairing 3 or 4 lines that where changed, or did not get placed into the new rendering engine.

I expect XvBA to be broken on Evergreen right now. This is taking into consideration that AMD has yet to make an official public release this api. As for Official release, i mean the header files and development library.

[DanaG]

Supports OpenGL ES? I don't get how it possibly can -- there's no libGLESv2.so in the package!

And how the heck did they bring back the lag that they fixed in 10.6?

This driver also gives me a new pane in amdcccle: Workstation ⇒ Deep Bit Depth Pixel. It gives me options to force disabled (this is default), enable only if capable display is present, or force enabled. Even when set to force enabled, I don't see any differences in glxinfo or Xorg log. Then again, the same is true of the Windows driver -- enabling 10-Bit Pixel Format does nothing visible.



A funny string in glesx.so:
ooboghieteipheilahpheengohnashoahveerohpithaewaxae pheepaiquevuhpahmaezosengoinoraeteichyaighaidxiegh oghaichoocheecaesalphoothizhoonaethpohvahniouzaghi ezoquahbaleikohpoaiyooyaueuroothoeefohsah
Try feeding it to espeak... it'll buffer-overflow and crash.

[smitty3268]

Supports OpenGL ES? I don't get how it possibly can -- there's no libGLESv2.so in the package!

The ES API was integrated into OpenGL4.1, so any 4.1 implementation automatically supports ES.

[allquixotic]

Cat 10.8 on OpenSUSE 11.3 brings back the unrendered black boxes issue that was fixed with 10.7

[allquixotic]

The ES API was integrated into OpenGL4.1, so any 4.1 implementation automatically supports ES.

I think his point was more about applications that try to link against libGLESv2.so. Any such apps would have to be told to link against libGL, I guess, but that is probably as simple as making a symlink. Has anyone tried that to see if it works as expected?

[DanaG]

The ES API was integrated into OpenGL4.1, so any 4.1 implementation automatically supports ES.

Still, all the Mesa demos of OpenGL ES, as well as PowerVR's (on ARM), all link against some variant of libGLES. The demos won't run accelerated on fglrx -- there's no ES (1.0 or 2.0) library to point them at!

I've also tried the ES demos on the open-source driver, and got "driver mismatch" between "r600" and "radeon".

[adamk]

Cat 10.8 on OpenSUSE 11.3 brings back the unrendered black boxes issue that was fixed with 10.7

Except that it was never fixed on 10.7:

http://www.phoronix.com/forums/showpost.php?p=139094&postcount=18

Adam

[allquixotic]

Except that it was never fixed on 10.7:

http://www.phoronix.com/forums/showpost.php?p=139094&postcount=18

Adam

Fair enough, but for me at least, it was 100% fixed on 10.7. So maybe my hardware / X server version / kernel config options combination somehow fixed it in 10.7, but even with my options set "just so", it turned up broken again on 10.8.