Ryan Gordon Criticizes Open-Source Drivers Again

[Sidicas]

I am happy that you thank the open driver devs because I am thankful for their work, too
But on the other hand you blame amd for not supporting you enough. But you should know that most of the opensource-radeon-driver developers (or at least the most active) are payed by amd. So Amd cares about you but you dont know it ^^

And why do you want to play with such old hardware, are you so poor? I mean I like netoboks and notebooks which habe good drivers with fast 2d acceleration and good video-encoding support but why do you want so hard to play with that old office-notebook?

Correct me if I'm wrong, but when I look through the features that Gallium3D supports, I look across and see that even on the new hardware they only support up to OpenGL 2.1.. Which is the same OpenGL level my existing hardware has.. So why buy new hardware if I can't use the features anyway? I've already been in that situation, and don't really see much purpose to go there again..

But you're right the mobo on this laptop died last year and I replaced it for $40 when I could have just bought a new laptop.. It's lots cheaper than buying a new laptop, but I do plan to buy a new laptop eventually.. I just really don't need faster hardware than what I have now in this laptop.. I use my laptop to do what I need to do, and nothing more... I have tons of other computers for other purposes and appropriate hardware to match what I use them for, some of which are high end PCs but I still use this laptop a lot.. I don't like to keep one or two PCs for everything because I never know when my motherboard is going to suddenly die, and I never know if somebody might steal my luggage. I keep a lot of PCs, and if one goes down or gets stolen, I just pick up another and it doesn't even ruin my day..

Only reason why I mention this laptop so much is because it's the only one I've got running open source drivers (and relevant to the thread)... And it's running open source drivers primarily because there is no other/better option for it.

Edit: My Mobility X700 can still beat the snot out of the best of Intel's graphics solutions that are out today... 5 years later... It might be old, but it's certainly not obsolete.

[Luke_Wolf]

I see games in the same category as movies: CONTENT. If I were to buy a disk with a movie and it wouldn't play on linux because it was a Windows executable I woulnd't ask the producer when the Linux version will be ready, I would ask them why it's not a mpg, avi, mkv or whatever other format that I already have a player for.
The same for a game, what I'm interested in a game is the graphicks, sound, maybe some music and the scripts that are required to allow interaction. Not the engine and the hacks they added to the engine so that they get a few extra FPS. When the gaming industry will be out of it's infancy they will be able to separate the engine from content and when you will buy a game you will only get the content, the engine required to play it will be packaged by your distribution.
And until they reach that point I will prefer to install those binary blobs full of hacks under wine than pretend they are native applications. They may be in the ELF format, but they are still follow the Windos model of distributing software that I don't even want to work on Linux, Linux already has a much better model and it shouldn't be broken just because "that's how we do it on Windows and we're used to it".

P.S. Yeah, I'm one of those "zealots" some of you were talking earlier.

+1 Although I already suggested basically the same thing 4 posts above yours. The only real solution to this problem is open source engines, Distros can package the software but eventually it will break, If instead we're running engines and dropping in content, the games will last forever so long as the engine remains up to date. On top of this if the game engine is opensource you don't have to worry about putting in the effort to making it crossplatform, because guess what? It'll do it itself.

[blackiwid]

It appears I will have no choice but to continue using open source drivers for the life of the laptop. So I have to give a shout out to all the great people who are still working on improving drivers for this hardware (YAY!! r300g team!!!).. Just letting you know that your work is greatly appreciated as both AMD and HP don't care at all about releasing decent drivers for my hardware under any platform..

I am happy that you thank the open driver devs because I am thankful for their work, too
But on the other hand you blame amd for not supporting you enough. But you should know that most of the opensource-radeon-driver developers (or at least the most active) are payed by amd. So Amd cares about you but you dont know it ^^

And why do you want to play with such old hardware, are you so poor? I mean I like netoboks and notebooks which habe good drivers with fast 2d acceleration and good video-encoding support but why do you want so hard to play with that old office-notebook?

[Sidicas]

also i found the OEM's releasing drivers to be a disservice the final customer buying the final product. The OEM's typically would only release a couple versions of the original drivers. I personally was thankful I found the mod omega drivers which reenabled support in the cataylst driver so it would have support for the mobility drivers. But I dont think its something I should have had to do.

You mean under Windows, right? I didn't know about those mod omega drivers.. Interesting.. Since HP didn't release updated drivers for the Mobility X700, it seemed like every Windows game after 2008 or so had some graphics problems.. So I moved gaming over to another PC and put this laptop under Linux.. Only to run into limited OpenGL support which didn't really matter much to me until now... and some stability problems which I easily avoided by just not doing those things that made xserver crash (ie: not running certain deinterlace modes on HD videos and not dragging windowed 3D apps too far off the screen). It's been a rough 5 years but it hasn't been unbearable..

[bridgman]

We stopped working on older hardware with the Catalyst driver a few years ago. We didn't stop working on older hardware, just moved the work from the Catalyst stack to the open source stack.

Getting support into the open drivers also makes it possible for other developers to do great things with the code, from MostAwesomeDude writing the first Gallium3D driver for ATI hardware through to all the feature and performance improvements Marek and others have been making.

Anyways, there have already been a few dozen threads about support for older hardware in Catalyst vs open drivers - we should probably get this thread back to talking about Ryan's post and related topics

[Pickle]

I wasn't aware that AMD was still working on stuff related to older hardware, I figured they were only concerned about their latest hardware..

they are not working on older hardware, you will find that your hardware is only supported up to a certain catalyst version. Oh and that catalyst version is going to work with older version of Xorg, so that means if you run that latest version of your distro the last supported catalyst is useless. Your only option then is the opensource driver.
Now its not a totally bad thing, i use the radeon driver and it works. It would be nice some day if it came at least close the closed driver's speed.

also i found the OEM's releasing drivers to be a disservice the final customer buying the final product. The OEM's typically would only release a couple versions of the original drivers. I personally was thankful I found the mod omega drivers which reenabled support in the cataylst driver so it would have support for the mobility drivers. But I dont think its something I should have had to do.

[bridgman]

The Catalyst driver for Linux is maybe 20 million lines of code, much of it shared with Windows and other OSes. Since the shared code also covers OSes with robust DRM requirements (even though that code isn't used for Linux), it's not real practical to open it up. The open source drivers are much smaller (a few hundred thousand lines vs tens of millions of lines) which makes them a lot easier to maintain.

Our initial plans did include sanitizing and opening up some proprietary driver code and we spent a year or so of part time work trying to do that (initial plan was "tcore" and the bottom end of the proprietary OpenGL stack) but in the end our conclusion was that writing new Linux-specific code was a better use of time than sanitizing cross-OS code for use on Linux and similar OSes.

[Sidicas]

Just checking, you do know that some of the developers working on the open source drivers (including a lot of the kernel driver work which made support of higher GL levels possible) are full-time AMD employees, right ?

I wasn't aware that AMD was still working on stuff related to older hardware, I figured they were only concerned about their latest hardware.. Why don't they just open source the Catalyst driver? The benchmarks show it performs better than the current open source driver, so it seems AMD is more committed to the Catalyst driver than the open source driver.

Make no mistake, I'd run whatever runs best, whether it's open source or not. It's just that I've always felt that owning this Mobility X700, ATI really didn't put much commitment into the Linux drivers.. I remember when the first Linux Catalyst drivers came out, everybody was raving about them, but the Mobility X700 users were left out of all the fun...

I was forced to use the open source driver which was buggy in that it crashed Xserver from time to time and had incomplete OpenGL support for almost 5 years.. People are saying the r300g finally fixed everything up.. I suppose getting decent drivers 5 years after buying the hardware is better than never getting it at all..

[bridgman]

AMD does not support my graphics chip on any platform. Under Windows, they pass the buck on to the manuf. of the laptop (HP) to provide the drivers....

Our OEM customers ask us to *release* the drivers through them so that they can do their own system-specific qualification testing. We provide the drivers to our customers.

Under Linux, right now I'm running the Mesa "radeon" driver, which only supports OpenGL 1.4, even though this hardware can run OpenGL 2.0 games under Windows. I've heard that the Gallium3D drivers now supports OpenGL 2.0 on this hardware, so I'm really looking forward to trying it out. For now, a couple of the Linux games that I got aren't rendering properly using the "radeon" driver, but they do render properly using the software Mesa driver. I've got my fingers crossed and am hoping that when I update my OS, Gallium3D will fix everything for me..

For clarity, Gallium3D is "just" a new HW driver interface layer within Mesa (albeit a newer and more versatile one), so you will still be using mesa, just a more up-to-date version. Note that adding higher levels of GL support required changes in both mesa/gallium3D and in the kernel driver's memory manager (among other things).

It appears I will have no choice but to continue using open source drivers for the life of the laptop.

You say that like it's a bad thing

Just checking, you do know that some of the developers working on the open source drivers (including a lot of the kernel driver work which made support of higher GL levels possible) are full-time AMD employees, right ?

[Sidicas]

Oh, let me guess. You're one of those ultra-conservative types who only uses the previous version of Debian Stable or RHEL? Because I can't think of what current distro you might be running that doesn't ship r300g by default.

It's two things..

The first, I keep hearing that the Gallium3D driver isn't as fast as the Mesa classic driver in all situations.. I haven't really seen any benchmarks comparing them since June of last year.. Is it faster now? I just got Phoronix Test Suite this morning and am going to be running some benchmarks on r300c to compare to everybody else's benchmarks running r300g... I haven't heard anything at all.. All I know is that it supports more features than r300c..

The second, is that I don't want to move from KDE3 to KDE 4 until KDE 4.7 (or newer) gets into Debian Testing.. There's a lot of things in KDE3 that are still missing in KDE4 such as the ability to mix your desktop background on-the-fly which I like to have on my travel laptop because it shows which areas of the world are currently in sunlight (updated every 10 mins).. It really helps when it comes to dealing with the jet-lag.

KDE4 has Globe, but it's a separate application. It calculates sunlight like kworldclock used to, but it renders it to the display instead of saving it to an image and even if it did save it to an image, I'd have to figure out how to automate it.. kworldclock got turned into a cute little widget, and now it won't export sunlight map images for Desktop wallpapers (as far as I know).

There are a few other things in KDE4 that aren't quite as polished as they were in KDE3 yet, but they're getting there.. I do plan to upgrade that laptop to Debian testing once KDE 4.7 gets there. Even if I do make the jump to KDE4, I'd probably run it without all the KWIN effects because it's old hardware.

It's not that I want to run old software, it's that some of the newer software loses features, so I just don't run it.. If I could figure out how to get my system configured the way it is now on KDE4 or even Gnome, I would. I've got KDE4 and Debian Testing running in a Virtual Machine and am trying to find other ways to get it configured the way I want.

If I could figure out how to run the Gallium3D driver with KDE3 (on Debian oldstable) without too much effort in recompiling, I'm fine with that also..

And I do run Ubuntu 11.04 and other newer distros,, it's just on different (newer) hardware, but those PCs have different purposes / uses