News: 0001649106

  ARM Give a man a fire and he's warm for a day, but set fire to him and he's warm for the rest of his life (Terry Pratchett, Jingo)

Patches Posted To Improve Linux Media Acceleration On Intel Alchemist GPUs

([Intel] 6 Hours Ago Intel HuC For DG2 On Xe)


The Intel Arc A-Series graphics (DG2) remains in a bit of an awkward state on Linux due to the transition period moving from the i915 kernel driver to more modern and feature capable Xe kernel driver. It's only post-Alchemist where the Intel Xe driver is officially supported and used by default while those opting for the Xe driver with the A-Series can often enjoy better performance and functionality. Except one of the caveats in using the Xe driver on Alchemist GPUs has been the lack of proper media acceleration support.

A long-standing limitation of opting to use the Xe open-source driver on Intel Arc A-Series cards has been [1]no HuC media support that is necessary for GPU acceleration with codecs like H.265/HEVC. HuC on Alchemist works on the i915 kernel driver but hasn't until now been wired up for Alchemist/DG2 on Xe. That in turn has blocked some users from [2]switching to the Xe driver for better performance for Linux gamers and enthusiasts.

Today though a merge request was opened that has laid out the HuC micro-controller support for DG2/Alchemist with the Xe driver. Open-source developer Daniil Seliverstov worked out the DG2-specific GSC/MEI handling for the Xe driver with:

"- add DG2 MEI GSC interface support

- add HuC firmware parsing for GSC BPDT/CPD layout

- improve HuC loading and authentication handling

- add GSC register definitions and firmware ABI updates

- add Xe MEI DG2 module support

- update Xe workarounds and engine handling for HuC/GSC

- update MEI GSC support for DG2

This enables HuC firmware initialization and authentication on DG2 hardware when using the Xe DRM driver."

The code is pending [3]here . It will be interesting to see if/when it's upstreamed for the Xe driver to improve the media experience for those on the Intel Arc A-Series hardware.



[1] https://www.phoronix.com/news/Intel-Xe-DG2-No-HuC

[2] https://www.phoronix.com/review/intel-xe-i915-linux-619

[3] https://gitlab.freedesktop.org/drm/xe/kernel/-/merge_requests/361



For example, if \thinmskip = 3mu, this makes \thickmskip = 6mu. But if
you also want to use \skip12 for horizontal glue, whether in math mode or
not, the amount of skipping will be in points (e.g., 6pt). The rule is
that glue in math mode varies with the size only when it is an \mskip;
when moving between an mskip and ordinary skip, the conversion factor
1mu=1pt is always used. The meaning of '\mskip\skip12' and
'\baselineskip=\the\thickmskip' should be clear.
-- Donald Knuth, TeX 82 -- Comparison with TeX80