VIA Interested In Gallium3D For Chrome 9

[phoronix]

Phoronix: VIA Interested In Gallium3D For Chrome 9

Following the release of a new VIA 3D graphics stack by Tungsten Graphics, there has been a discussion on the OpenChrome development mailing list as to the next steps to take in open-source VIA support. One of the items now being discussed between VIA Technologies and Tungsten Graphics is Gallium3D. VIA wants a Gallium3D driver for the Chrome 9 series, but they don't really know anything about Gallium3D, which makes it hard for them to write the driver...

http://www.phoronix.com/vr.php?view=Njk4MA

[russofris]

So it looks like the following will occur:

1: Push Thomas's new <CX700 driver into mainline
2: Integrate the Via's chrome9 bits into Thomas's new Openchrome DRM
3: Convert everything to Gallium
4: Add 2D accell and other bits (GPU video accell, etc)

The questions I have are: whether step 3 includes Gallium for <=CX700, or just chrome9? Will the Gallium driver use the GEM API with a TTM backend (like nouveau), or will it be 100% TTM?

I hope that this puts Via's Eden and older C3 integrated boards on a level playing field (well, almost) with Atom based offerings. It would breathe new life into a number of older notebooks and budget PCs as well.

F

[oneman]

I also want to build my own flying machine, I'm interested in wings and an engine, but I'm not sure how they work, but they sound cool? I want them. Also I havn't release my cabin design yet. Cheers.

[Svartalf]

Originally Posted by russofris

The questions I have are: whether step 3 includes Gallium for <=CX700, or just chrome9?

I'd hope for all of them in the UniChrome lineup. However, having said this, the devices in question are pretty weak. Even weaker than Chrome9 is on things.

Will the Gallium driver use the GEM API with a TTM backend (like nouveau), or will it be 100% TTM?

The main reason Nouveau is doing it that way is because they were on a given path and to completely re-write the code would have been...fun...

I'd suspect that every new driver out there using the Gallium framework will be using the TTM layer instead of the old GEM one.

I hope that this puts Via's Eden and older C3 integrated boards on a level playing field (well, almost) with Atom based offerings. It would breathe new life into a number of older notebooks and budget PCs as well.

Only the C7 based stuff would be in the similar class to the Atom lineup. The C3 stuff's just nowhere near as usable in comparison. It'd be nice to have it as people with the stuff would actually have real support for their GPUs, but since the GPUs aren't very much to write home about, it's not a burning issue unless you've got the old stuff. I'm more keenly interested in seeing a Nano machine with a mediocre GPU on the board with a PCI-E/MXM slot upgrade path. With a modern NVidia machine, the Nano boards were playing current AAA titles with headroom to spare.

[[Knuckles]]

Originally Posted by Svartalf

I'd suspect that every new driver out there using the Gallium framework will be using the TTM layer instead of the old GEM one.

It's the other way, TTM=old GEM=new

[russofris]

Originally Posted by Svartalf

I'd hope for all of them in the UniChrome lineup. However, having said this, the devices in question are pretty weak. Even weaker than Chrome9 is on things.

Understood, but having a unified back-end (Gallium will probably be used for everything =>R100 and NV5/NV1x) will be nice, as it provides a good framework for building drivers for other legacy graphics adapters.

The main reason Nouveau is doing it that way is because they were on a given path and to completely re-write the code would have been...fun...

Understood, but do we have a current non-UMA gallium driver with a GEM'ified memory manager? I was under the impression that nouveau stuck with TTM'ified GEM for two reasons:

1: As you mentioned, they already had a TMM memory back-end, and did not desire re-doing everything in GEM
2: The reference GEM MM implementation by Intel lent itself to UMA and building a discreet back-end would require a large amount of pioneering work.

I can see though how GEM (the MM, not the API) would be immediately useful to VIA integrated GPUs, as they are in the same boat as Intel... Despite their arrangement with Tungsten Graphics.

Only the C7 based stuff would be in the similar class to the Atom lineup. The C3 stuff's just nowhere near as usable in comparison. It'd be nice to have it as people with the stuff would actually have real support for their GPUs, but since the GPUs aren't very much to write home about, it's not a burning issue unless you've got the old stuff. I'm more keenly interested in seeing a Nano machine with a mediocre GPU on the board with a PCI-E/MXM slot upgrade path. With a modern NVidia machine, the Nano boards were playing current AAA titles with headroom to spare.

Ahh, yes. I saw the nano 2100 review which placed it in between an Atom and Pentium M in terms of performance. I hope that their future motherboard offerings show the same improvement, in addition to features and performance.

I did a discovery last year when trying to assemble a home built NAS device. I was unable to find GB Ethernet and 4xSATA on a single fanless board. When I found something that was close, it cost $300 (and not the $75 I was willing to spend on a fanless C3/C7 platform). Perhaps this year will bring better offerings.

F

[Svartalf]

Originally Posted by [Knuckles]

It's the other way, TTM=old GEM=new

Excuse the lack of caffeine induced gaffe...

Looking at it, I realized I'd put that bassackwards.