AMD's R300 Gallium3D Driver Is Looking Good For 2011

[Beiruty]

In free market, here how it works:
1) Buyers vote by their dollars.
2) Buyers buy a product and expect it to work.
3) The selling company keep supporting their product, so their clients keeps buying their future products.
4) No support by the seller, no confidence by the buyer, buyers move on to another product, another platform.

As simple as that.

[smitty3268]

In free market, here how it works:
1) Buyers vote by their dollars.
2) Buyers buy a product and expect it to work.
3) The selling company keep supporting their product, so their clients keeps buying their future products.
4) No support by the seller, no confidence by the buyer, buyers move on to another product, another platform.

As simple as that.

You seem to have forgotten the point where someone gets a product for free and then goes onto the internet to complain about how it's not good enough. Especially the part where they're actually complaining about a 3rd party product and roll it into a criticism of the completely free other product.

[loonyphoenix]

In free market, here how it works:
1) Buyers vote by their dollars.
2) Buyers buy a product and expect it to work.
3) The selling company keep supporting their product, so their clients keeps buying their future products.
4) No support by the seller, no confidence by the buyer, buyers move on to another product, another platform.

As simple as that.

Gah, I hate trolls

I'd like to see what part of r300g you, personally, bought (or paid for) to be justified to talk about buyers.

[mattst88]

That's fine and well, but your target wasn't AMD. It was r300g, which has had about zero-input from AMD directly.

/home/mattst88/projects/mesa/src/gallium/drivers $ git log --no-merges r300/ | grep ^Author | sort | uniq -c
      1 Author: Alex Deucher <alexdeucher@gmail.com>
      7 Author: Brian Paul <brianp@vmware.com>
      4 Author: Chia-I Wu <olv@lunarg.com>
      1 Author: Coleman Kane <Coleman Kane>
     13 Author: Cooper Yuan <cooperyuan@gmail.com>
    544 Author: Corbin Simpson <MostAwesomeDude@gmail.com>
      1 Author: Dave Airlie <airlied@linux.ie>
     50 Author: Dave Airlie <airlied@redhat.com>
      5 Author: Jakob Bornecrantz <jakob@vmware.com>
      1 Author: Jeff Smith <whydoubt@yahoo.com>
      5 Author: Joakim Sindholt <bacn@zhasha.com>
     15 Author: Joakim Sindholt <opensource@zhasha.com>
     16 Author: José Fonseca <jfonseca@vmware.com>
     22 Author: Keith Whitwell <keithw@vmware.com>
      5 Author: Luca Barbieri <luca@luca-barbieri.com>
      1 Author: Łukasz Krotowski <lukasz.krotowski@gmail.com>
      2 Author: Maciej Cencora <maciej@osiris.(none)>
     11 Author: Maciej Cencora <m.cencora@gmail.com>
    528 Author: Marek Olšák <maraeo@gmail.com>
      1 Author: Mathias Fröhlich <mathias.froehlich@gmx.net>
      1 Author: Mathias Fröhlich <Mathias.Froehlich@gmx.net>
      2 Author: Mathias Fröhlich <Mathias.Froehlich@web.de>
      3 Author: Mathias Gottschlag <mathias-go@web.de>
      1 Author: Matt Turner <mattst88@gmail.com>
     15 Author: Michal Krol <michal@vmware.com>
      1 Author: michal <michal@transistor.(none)>
      8 Author: Michel Dänzer <daenzer@vmware.com>
     16 Author: Nicolai Hähnle <nhaehnle@gmail.com>
      3 Author: nobled <nobled@dreamwidth.org>
     20 Author: Roland Scheidegger <sroland@vmware.com>
      1 Author: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
     10 Author: Tom Stellard <tstellar@gmail.com>
     38 Author: Vinson Lee <vlee@vmware.com>
      3 Author: Zack Rusin <zackr@vmware.com>

Of the 1355 total commits, Corbin and Marek account for a full 79.1% of them (and a much larger percentage in terms of actual work done). AMD's Alex Deucher and Cooper Yuan have 14, or 1%.

You're confusing your frustration with r300g and AMD.

Say what you will about AMD, but Alex's post really describes the situation very well.

If you want to complain that AMD's not doing enough to support r300g, fine; but complaining about r300g as if its developers owe you something: not good.

[Beiruty]

Yep, my frustration and anger should be directed to AMD. Nothing but praises for those who contributed to r300g.

AMD is frankly keeping the "secret sauce" for itself. At least let AMD help in the optimization aspects and release a way to use GPU HW for video decode via OpenCL.

[bridgman]

OK, now I'm confused too. You're talking about 300g, which supports only the older GPUs that don't have OpenCL-capable hardware or video decode hardware (other than a legacy MPEG-2 IDCT block).

[Beiruty]

OK, now I'm confused too. You're talking about 300g, which supports only the older GPUs that don't have OpenCL-capable hardware or video decode hardware (other than a legacy MPEG-2 IDCT block).

Bridgman,
Regardless, in windows world even those old GPU have better support for video acceleration and 3d acceleration too.

Newer GPU HW have a lot to be desired in the linux world .

[bridgman]

Sure, but that's not what you said. First you were attacking the 300G project and developers, then you were saying that :

- we were holding back "secret sauce" that would presumably let 3D run faster

- we should be implementing an OpenCL-based video decode solution for that hardware

The "secret sauce" is hundreds of millions of dollars spent on driver development, and 10-15 million lines of code. We can't really distill that down into something that can practically be used in the open source drivers, although we do keep looking for "best practices" which *can* be carried over to the open drivers.

It's like when people ask about the secret to English lawns ("seed and roll for 400 years"). Everyone knows how to do it, but it's hard, slow, expensive work.

[glxextxexlg]

I really wonder why GLSL shaders work faster than the fixed function ARB.. I'm really curious about this weird issue..

[smitty3268]

I really wonder why GLSL shaders work faster than the fixed function ARB.. I'm really curious about this weird issue..

Because modern hardware designs focus on creating lots of shader power, in order to be fast on modern games. The fixed function portion of the hardware is not considered important, and really only even included for backwards compatibility.