AMD's Hiring Another Open-Source Driver Developer

[elanthis]

So, new Open Source driver developer. That's pretty awesome. We can all agree that's good news, however you cut it.

[runeks]

So, new Open Source driver developer. That's pretty awesome. We can all agree that's good news, however you cut it.

Going from three to four guys is an increase of 33% in the AMD open source driver work force. That's is quite an increase. Definitely good news!

I'm very pleased that AMD is taking this step, and it only further strengthens my commitment to buying AMD products in the future.

[Decatf]

I'd wait for actual results before committing to anything.

[crazycheese]

I'd wait for actual results before committing to anything.

Speaking as someone who commited, opensource is still much closer to VESA than to equivalent of nvidia binary. Sadly, 4 developers are not much, the development is very very slow. You dont commit to opensource if you purchase AMD - either they are not interested or selling more cards cause of opensource driver is not part of the opensource goal.

If the company is to do opensource, they should do it professionally, not the hungry student way. Current opensource drivers will do fine for server environment, maybe with 3D surplus.

In the end, there is crappy, but working 99,5 of the time, nvidia binary, with crappy closed code(and heavily reduced linking possiblities), worser, but near same-day support, hardware and bad company policy; vs way more crappy amd binary, which is fun to watch to be developed in consumer desktop direction instead of leaving it workstation-only as usual, opensource driver with one-third of functionality of binary, no same-day support - serious effort from 3 people really; just like 300. Intel is not usable for 3D, especially in linux(!).

I don't see "Supported Operating Systems:.. Linux" neither on Nvidia, nor on AMD, nor on Intel hardware. Lots of DirectX and Windows stuff everywhere though. Seems the crappier the stuff is, the more people around the globe will buy it - just keep it PR'ed.

[xiando]

Sadly, 4 developers are not much, the development is very very slow.

You forget the RHT (http://finance.yahoo.com/q?s=RHT) drones. Log http://cgit.freedesktop.org/mesa/mes...t=grep&q=r600g shows them doing as much code commits as AMD drones, lately more. Counting them doesn't increase the amount of people that much, but they do important work and should not be forgotten.

[bridgman]

Yes, and all the messages are from you

[crazycheese]

You forget the RHT (http://finance.yahoo.com/q?s=RHT) drones. Log http://cgit.freedesktop.org/mesa/mes...t=grep&q=r600g shows them doing as much code commits as AMD drones, lately more. Counting them doesn't increase the amount of people that much, but they do important work and should not be forgotten.

I don't have RedHat subscription, and though they are known for committing nearly 40% of new kernel code, I have no chance to support them.

But I do buy AMD cards and AMD hardware, so I really want them to know about users as me, the reason I spend the money on AMD (not others) and want them to create correct moneyflow for group of persons as I am.

Currently, by buying AMD hardware I am not supporting opensource driver, although near 94% of people do prefer opensource opengl4 driver, but catalyst(fglrx) at best. Which I really don't want to do. Because I already had nvidia and had closed source driver.

[Kano]

It is impossible to buy a card to support Linux development, all chips are sold to oem and all cards target win users. Basically amd/nvidia could only track the numbers of driver downloads, maybe with help of the distributions which ship prepackaged drivers and would not download from the vendor site - but tracking oss drivers is impossible.

Basically there are lots of Linux users out there who would deserve good drivers as well, this it not completely out of the scope as basically amd+nvidia share lots of (opengl) code between different os. But the layers on top of xorg / kernel drivers are definitly linux specific. Nvidia goes a bit further and even creates drivers for FreeBSD+Solaris, there are no optimized amd drivers for those os.

The biggest drawback with nv cards is currently that you have to use nvidia-settings to configure your screens and can not use xrandr or the nice gui setting variants inside kde4 or gnome for multi monitor support. But you usually even have to start those after you used nvidia-settings, thosewise kde 4.4.5 did not notice the new screen size...

For ati the oss driver is mainly to support "abandoned" cards, when you look at the supported cards list of fglrx you should know what i mean. It's nice when you can use composite effects or very simple 3d games with oss drivers, but do you really buy a $100+ card to get 30% or less speed with oss drivers compared to binary drivers? Also you fully miss openGL 4 support or video accelleration (hey ati, when do "normal" users get h264 l5.1 support?).

The only logical oss solution would be buying one of the cheapest cards, then you could stay with hd 4 series as even older drivers/kernels supported em already. Or do you think you can use all opengl 4 features with oss soon?

[devius]

But I do buy AMD cards and AMD hardware, so I really want them to know about users as me, the reason I spend the money on AMD (not others) and want them to create correct moneyflow for group of persons as I am.

That makes sense. Right now probably all they can do is use statistics. I think the phoronix graphics survey helps in reaching out to manufacturers, so they can see that there is actually a market for linux graphics. I know that there are lots of users who use linux for workstation type work (me included), and with the appearance of more professional CAD and graphics related software this will become a substancial market in the long run.

[runeks]

Currently, by buying AMD hardware I am not supporting opensource driver, although near 94% of people do prefer opensource opengl4 driver, but catalyst(fglrx) at best. Which I really don't want to do. Because I already had nvidia and had closed source driver.

I see your point here, we definitely support the fglrx-driver more than we do the open source driver when we buy an AMD GPU.

But must it be either or? Either we support closed source, or we support open source?

Wouldn't it be more true if we said that when buying an nVidia card, 0% of our investment goes to further open source GPU drivers. And when we buy and AMD card, 3% of our investment goes to support open source drivers? (if we continue on with those 3 OSS devs vs. 100 fglrx devs numbers)

Currently, as I see it, there simply is no way to make an investment in a graphics card that exclusively goes to promote OSS. But there are some investments which, more than others, favor open source software.

On a different, but although similar, note. I've been thinking a lot about setting up an infrastructure that allows us to economically support the developers of open source drivers (or any open source software project for that matter). If we, for example, could find just 1000 people willing to donate $10 each to see an implementation of OpenGL 4 in the radeon drivers, we'd basically be able to pay a developer full time for two months.
I'm aware that the main motivation for writing code like this is not money. But devs have to make money to support themselves like all of us do. Why not enable them to support themselves by coding on the radeon drivers?

I'm imagining some sort of website where someone can propose an open source software project, for example "the implementation, in the open source radeon driver, of OpenGL 4, which performs at minimum 50% the speed of the fglrx-implementation". The wage would be $10,000 - collected by finding 1000 people each willing to give $10 to the project. The project would have associated with it an independent third party, who would be responsible for determining when the goals of the proposed project have been met, and the developer - whom all three parties have agreed on is to take on the project - thus has deserved the wage paid by the the supporters of the project.
I imagine this could speed up development by a large degree. The strange thing is that there seems to be plenty of people who are interested in OSS drivers, but few ways to channel our willingness, and by willingness I include willingness to pay money, into actual code. Maybe a system like this could help to solve this issue?

So a "coder offer" would look like this:

Open source software project proposition

• Name: OpenGL 4 in radeon
• Purpose of the project: To create an implementation of the OpenGL 4 API in the open source Linux "radeon" GPU driver, with, at least, 50% the performance of the corresponding implementation in the fglrx-driver
• Evaluator: John Bridgman (AMD)
• Wage: $10,000 (collected by 1000 people each donating $10)

For this project to begin, we'd have to get agreements from the following parties:

1. The evaluator, that he is responsible for deciding when this project is finished
2. 1000 people who agree to donate $10 (and are willing to accept that the specific evaluator chosen for this project decides when this project is finished)
3. A developer who thinks he can do this and that $10,000 is a fair amount of money to be paid for this project (and who believes that the evaluator is a reasonable guy)

So... anyone willing to implement this and set up a website on codeoffer.com to get this going?

The website would have some similarity to Kickstarter (http://www.kickstarter.com/). The main difference being that anyone can set up a project. Kickstarter only helps developers with a project find sponsors. This website would help sponsors with a project find developers. In fact, any of the three parties in a project (donators, developer, evaluator) should be able to start a new project on this website.