Gallium Nine Lands Threaded Context Support, Other Improvements

[pal666]

FWIW, my Proton fork (see link) does support gallium-nine out of the box if it is installed on the system

thank you for your work. still, i hope upstream proton will support it, because it's the one which is used out of the box

[Quackdoc]

it cannot as it is. with more development anything can
minor amount is more than nothing

The levels of effort we are talking about are two vastly different degrees. Mesa already has a D3D10UMD state tracker the ammount of work needed to get it to work is vastly different than completely than implementing full dxgi support, including proprietary components. considering the need to clean room reverse engineer them to even make it legal to distribute.

and if you wantes to write it as a UMD driver that would also require its own ammount of work.

where as, Zink, is a gallium backend meaning most of the work in implementing d3d10umd support is already done, and VMware has already stated they plan on opensourcing their d3d11umd driver too.

also saying it has overhead is obvious. it's just that gallium overhead is minuscule. You may as well stop using Radeonsi, Iris drivers etc because they too are gallium and have the same overhead.

the only difference in overhead compared to something like OpenGL is completely dependent on the state tracker and zink itself, and zink is being worked on to great degrees and is making rapid progress.

[lorn10]

It looks that the discussion here came to an end.

Now follows as my last contribution a final question Quackdoc . What is the main conceptional difference of these new d3d10umd and d3d11umd state trackers in comparison to Gallium Nine? According to the original merge request, d3d10umd works as a "WDDM UMD software driver": https://cgit.freedesktop.org/mesa/me...e7fb7f7ef10cd7

So currently it can be used only via llvmpipe/softpipe (and zink?). Gallium Nine is not implemented in that "WDDM UMD software driver" way. It effectively represents an (almost) fully featured "hardware capable" native DX9 API on Gallium. But if I understand these new d3d10umd and d3d11umd state trackers correctly they could be theoretically expanded to work in a similar way like Gallium Nine.

Whatever, it is quite interesting to see that these new Gallium d3d10 and d3d11 efforts are coming from VMware. If I remember me correctly this was also true (to some degree) for the previously failed d3d1x state tracker.

[Quackdoc]

It looks that the discussion here came to an end.

Now follows as my last contribution a final question Quackdoc . What is the main conceptional difference of these new d3d10umd and d3d11umd state trackers in comparison to Gallium Nine? According to the original merge request, d3d10umd works as a "WDDM UMD software driver": https://cgit.freedesktop.org/mesa/me...e7fb7f7ef10cd7

So currently it can be used only via llvmpipe/softpipe (and zink?). Gallium Nine is not implemented in that "WDDM UMD software driver" way. It effectively represents an (almost) fully featured "hardware capable" native DX9 API on Gallium. But if I understand these new d3d10umd and d3d11umd state trackers correctly they could be theoretically expanded to work in a similar way like Gallium Nine.

Whatever, it is quite interesting to see that these new Gallium d3d10 and d3d11 efforts are coming from VMware. If I remember me correctly this was also true (to some degree) for the previously failed d3d1x state tracker.

Breaking it down.

What is the difference between G9 and d3d10umd?
In terms how it actually works, there isn't too much, however both of the convert DX to Gallium. but d3d10umd has a lot more bits in it to properly work using window DDI (the device driver infrastructure I think) so it shouldn't be too hard to port to umd.

As far as zink, not yet, currently only llvmpipe, but since it is gallium it should in therory be relatively easy to get zink support.

[lorn10]

Thanks for that explanation. So finally we just need a Phoronix article about these new d3d10umd/d3d11umd state trackers. It looks that this quite interesting topic got somewhat lost at Phoronix. I know this is hard to imagine, but it actually happens... Maybe the authors Michael , phoronix need a little reminder.