The Status Of Gallium3D Drivers, State Trackers

[nanonyme]

Though even Intel with Larrabee admit that they need dedicated texture filtering in hardware, and a big part of the performance in AMD and NVidia GPUs comes from dedicated Z buffer optimizations.

Wouldn't that pedantically be neither a CPU nor a GPU anymore but something altogether different? But yeah, probably would have to be a separate hardware driver to utilize those functionalities that go over what a CPU has.

[some-guy]

According to him graphics will soon be rendered 100% in software again on normal general purpose CPUs.
I guess the upcoming Larrabee GPU follows this direction he describes.
But how does Gallium3D fit into the picture? Does Gallium3D have a place in pure software rendering?

IMO, I think GPU's will be merged with the CPU, making a new type of CPU that can handle huge amounts of streams simultaneously like the GPU's today. (Basically moving the GPU to the CPU die)

[philcostin]

What would be awesome is if the likes of NVidia and ATI would license the arm core and build a CPU with an integrated graphics pipeline and DDR3 memory controller.

[ioannis]

What would be awesome is if the likes of NVidia and ATI would license the arm core and build a CPU with an integrated graphics pipeline and DDR3 memory controller.

You mean like NVIDIA's Tegra platform?...

[philcostin]

You mean like NVIDIA's Tegra platform?...

/me goes to investigate!

EDIT: Wow, yes.. that is exactly what I meant. Cool! Next, we need unix graphics workstations based on that. :P

[curaga]

ARM workstations? They just don't have the oomph. AMD's Fusion and Intel's Arrendale would be better there, with AMD parts being better for graphic-intensive use as always.

[marek]

Back on topic...

There is still no hardware-accelerated Gallium3D. How long do you think it will take to make Gallium3D usable as a driver infrastructure? 3 years, I guess?

[ioannis]

Back on topic...

There is still no hardware-accelerated Gallium3D. How long do you think it will take to make Gallium3D usable as a driver infrastructure? 3 years, I guess?

partial implementations for Intel's 915/965 IGP, ATI R300 and the Nouveau project for Nvidia cards. I think the Intel one is usable, but not for mainstream, the others are in an early development stage.

[nanonyme]

There is still no hardware-accelerated Gallium3D. How long do you think it will take to make Gallium3D usable as a driver infrastructure? 3 years, I guess?

On ATi card side they probably want the existing driver infra (classic Mesa with radeon-rewrite, KMS+ttm) to be stable before focus moves on Gallium3D. They just stop extending it beyond a certain point. (which has not yet been definitely decided) It's easier to test the KMS+ttm against a Mesa implementation whose core doesn't change every time you look the other direction, ya know? After that part is done, we should look forward to getting Gallium3D drivers at a nice pace...

[marek]

Now it makes sense to me. Thanks for the explanation.