Intel Iris Gallium3D Driver Overhauls Its Buffer Allocation Code

[phoronix]

Phoronix: Intel Iris Gallium3D Driver Overhauls Its Buffer Allocation Code

While much of the modern graphics world these days is focused on the Vulkan API, there's no signs of Intel's open-source graphics driver engineers losing optimization focus with their OpenGL Linux driver by way of the Iris Gallium3D code. Merged this holiday week was a rather significant rework to its buffer object allocation system...

https://www.phoronix.com/news/Intel-G3D-Improved-Bucket-Cache

[Eirikr1848]

Wondering if these memory/buffer allocation changes will also positively impact Gen8/Haswell and Gen9/9.5 platforms.

[fahrenheit]

Wondering if these memory/buffer allocation changes will also positively impact Gen8/Haswell and Gen9/9.5 platforms.

I think Haswell and Broadwell are now on Crocus and not on Iris.

[Kayden]

Broadwell's currently on iris, and i915 has support for using 64K pages there, so I think it should help there somewhat. Meteorlake and the newest hardware should see the most benefit, though.

Haswell's on crocus, so this won't have any impact there. I don't think i915 has 64K page support on Haswell and earlier either; those are 32-bit PPGTT instead of a 48-bit address space. So, I'm not sure this would have much benefit even if backported to crocus. Some pieces, maybe.

[Eirikr1848]

Broadwell's currently on iris, and i915 has support for using 64K pages there, so I think it should help there somewhat. Meteorlake and the newest hardware should see the most benefit, though.

Haswell's on crocus, so this won't have any impact there. I don't think i915 has 64K page support on Haswell and earlier either; those are 32-bit PPGTT instead of a 48-bit address space. So, I'm not sure this would have much benefit even if backported to crocus. Some pieces, maybe.

Got it, thanks! For some reason in my “headcanon” IVB and HAS were on Iris… if for no reason other than to work with RUSTICL as a modern post-beignet OpenCL solution.