can we look forward to further improvements to the r300-r500 line?

[pedepy]

lately ive run into issues playing certain games under wine - the general response has been 'open source drivers only support very basic 3D - and most games will fail to be playable'.

what im wondering is, with all the working seemingly being directed at improving r600 and r700 3D support, is this all there is to the 3D in r500 for the open source drivers ? or is there active development under way to bring more features to the drivers.

i think its especially important since i have no other alternative ...

[bridgman]

I think you can expect to see continued improvements for 3xx-5xx, both in terms of GLSL support (nha has been working on that) plus moving to Gallium and picking up other GL 2.x functions (MostAwesomeDude and others have been working on that).

The main reasons for having Richard work on adding GLSL to 6xx/7xx rather than 3xx-5xx were (a) the shader compiler code for 6xx/7xx was still "fresh" in his head (the 3xx-5xx shader compiler is totally different), (b) other people were already looking into GLSL on 3xx-5xx, plus (c) he really wanted to do it

[pedepy]

is the implementation of these new functions dependent on this new 'gallium' architecture, or could they be added to the driver and then later one somehow made to work 'with' the new architecture ?

also, it's nice to see work continue on r600 and r700 and all that, but afaik those cards are still supported by catalyst; imo it would be best to first bring unsupported cards up to speed with catalyst and then concentrate efforts on cards as they get deprecated by ati, or are at at least expected to be next in line to that faith.

and just as im clear about this, were all of those features that im lacking at the moment, like GLSL, already supported by fglrx when it still ran my card, about a year ago ?

[Pfanne]

afaik opengl 2.1 and 3 features do not depend on gallium, but they do depend on kms and gem...
and these two features only got usable "recently", so until now there wasnt much work on newer opengl features.

[bridgman]

Yep, KMS and GEM/TTM are what really make the next round of OpenGL support possible.

The situation with Gallium3D is complicated, in the sense that Gallium3D allows a bit more of the driver code to be shared between different families and vendors (which is good) but there aren't a lot of different drivers to share with yet (which is ungood).

There is a solid consensus among the developers that once they reach the point where it's less effort to support a new feature on Gallium3D they probably won't bother with classic mesa for that feature, but nobody is quite sure where that point is and it does not seem to have been reached yet.

GLSL support is mostly in the shader compiler and that code is currently shared between the classic and Gallium3D drivers anyways.

Probably the best way to put it is that there's a good chance all the stuff you want will get done in the classic mesa driver, but nobody is going to commit to that. In the meantime the developers are going to keep pushing along the "path of least resistance", which is often the classic mesa driver.

[forum1793]

imo it would be best to first bring unsupported cards up to speed with catalyst and then concentrate efforts on cards as they get deprecated by ati, or are at at least expected to be next in line to that faith

If I were AMD I would spend my money with an eye towards making money. Supporting 5 year old cards that you already probably use widows on (or a previous linux distribution for which catalyst works), is not going to bring in money. So the efforts should be towards the newer hardware. If the OSS drivers get good enough maybe they drop catalyst for linux? Who knows but that might save some $.

It is really more interesting for the linux distributions to try and come up with support for the older hardware to make it easy to try their distributions out.

I'm not sure how linux makes money. Eventually the distributions will probably move to a advertisement model built into the OS. But then I won't use.

[bridgman]

We're trying to strike a balance between older and newer cards. When Richard started working on GLSL for 6xx/7xx there was already similar work being done for 3xx-5xx -- there's an element of luck in which one actually gets noticed first.

Something to remember is that even with relatively constant effort from developers you're going to see visible results arrive in bursts.

The developers know they're making progress, but a lot of the hardest work shows very little useable result -- then suddenly a whole bunch of stuff starts working, people who were waiting for that project get excited, and everyone else feels *that* project must be getting more love than the one *they* care about. That's usually not the case; it's more a question of what happened to get interesting results "this week".