Intel Compute-Runtime Updated With DG2/Alchemist Support

[tuxd3v]

I don't know if oneAPI is a generalized frontend for several different backends..
We have Clover in mesa, for openCL and they should adopt it..

Seems that today any company want to get their own implementation of Compute stacks..
There's Nvidia with its typical Cuda, theirs AMD, with openCL,Hip,openMP, etc,(RoCm) and now also intel with its own implementation( oneAPI ), am I thinking correctly?

What is seems is that everybody already understood that GPGPU compute is a thing that will grow to another levels in the future, and they will be massive in datacenter,
So everybody is trying to Create its own market, to stay in the lead..
And no one wants a shared API, to rule them all... everybody as its eyes on the cookies Jar!

[blacknova]

I'm looking forward for Intel's Alchemist GPUs at least that might move price point to more adequate values.

[aufkrawall]

Only problem: is anyone actually using oneAPI?

I guess the chronological order to first get software ready and (hopefully) then hardware isn't so bad.
Hopefully XeSS will use intel-compute-runtime to make use of matrix engines (if not Mesa), then it would work instantly also on Linux with an open API (while vendor exclusive, but still).

[tildearrow]

Only problem: is anyone actually using oneAPI?

[MadeUpName]

This is how you do Linux support.

[phoronix]

Phoronix: Intel Compute-Runtime Updated With DG2/Alchemist Support

Intel's open-source engineers today released the Compute-Runtime 21.47.21710 as their latest update to this open-source compute stack for Linux systems enabling their graphics processors to enjoy performant OpenCL and oneAPI Level Zero support...

https://www.phoronix.com/scan.php?pa...te-21.47.21710