Intel's New Iris Gallium3D Driver Picks Up Experimental Icelake Bits, GL Features

[phoronix]

Phoronix: Intel's New Iris Gallium3D Driver Picks Up Experimental Icelake Bits, GL Features

One of the talks we are most interested in at XDC2018 is on the Intel "Iris" Gallium3D driver we discovered last month was in development...

http://www.phoronix.com/scan.php?page=news_item&px=Iris-Gallium3D-Experimenta-Ice

[Venemo]

Has anyone attempted to run this new driver? Or is it still to early to get excited about it? I'd be very interested to see how it will turn out.

[Guy1524]

Is there any chance of this making its way to haswell?

[duby229]

I mean no offense, but many people stated years ago that Intel would eventually have no choice but to move to gallium3d or something similar. It has something to do with how VEC4 pipelines work, i don't understand all the details, but the mesa classic pipeline was not made for modern graphics architectures and that's what gallium solved years ago. Intel should have done this years ago, their excuse was that the classic pipeline was already well developed. And that was true, but you can't expect a well developed gallium driver to appear out of thin air, it actually needs to be developed well in order to become well developed. They should have done this a long time ago.

[Kayden]

Has anyone attempted to run this new driver? Or is it still to early to get excited about it? I'd be very interested to see how it will turn out.

It is pretty early yet. A number of features are missing, there are still quite a few bugs, crashes, even GPU hangs and I wouldn't be surprised if you encounter some hardlocks. Also, some critical performance features (color compressiond and HiZ) aren't there yet. YMMV, but unless you're planning on working on it, I would probably wait a bit.

Is there any chance of this making its way to haswell?

Not likely. Sorry :( There's a chance it might make it to Broadwell. But Haswell's architecture is a fair bit different than Skylake. In particular, memory management is a lot more constrained. I'm not sure what we're going to do about older hardware just yet. Stick with i965? Make an Iris-Legacy?

I mean no offense, but many people stated years ago that Intel would eventually have no choice but to move to gallium3d or something similar. It has something to do with how VEC4 pipelines work, i don't understand all the details, but the mesa classic pipeline was not made for modern graphics architectures and that's what gallium solved years ago. Intel should have done this years ago, their excuse was that the classic pipeline was already well developed. And that was true, but you can't expect a well developed gallium driver to appear out of thin air, it actually needs to be developed well in order to become well developed. They should have done this a long time ago.

No offense as well, but "I don't understand all the details" pretty much sums up this comment. vec4 vs scalar is a compiler topic, and both systems can now use the same compiler. Gallium vs Classic really doesn't have anything to do with modern hardware vs. old hardware. Classic is just a more raw interface - you have to do a bunch of things yourself in the driver. In fact, the st/mesa state tracker, which all Gallium drivers use to implement GL, is a classic driver. :) There is a whole lot of misinformation floating around on this topic...

[Venemo]

It is pretty early yet. A number of features are missing, there are still quite a few bugs, crashes, even GPU hangs and I wouldn't be surprised if you encounter some hardlocks. Also, some critical performance features (color compressiond and HiZ) aren't there yet. YMMV, but unless you're planning on working on it, I would probably wait a bit.

Thanks man! Would be interesting to see Intel graphics working with Gallium Nine!