Fedora Linux Grappling With New vs. Old Intel Hardware Support For Compute Stack

[phoronix]

Phoronix: Fedora Linux Grappling With New vs. Old Intel Hardware Support For Compute Stack

Fedora is among the Linux distributions that package up the Intel Compute Runtime stack to make it easy to run OpenCL and oneAPI Level Zero workloads on Intel graphics hardware via the distribution package manager and without having to jump through any extra hoops. But now with upstream Intel Compute Runtime dropping support for Ice Lake and older leaves the Fedora support in a pickle. Currently they are focusing on the "legacy" branch with older hardware support but for Fedora 42 are looking at upgrading the support to focus on newer Intel graphics hardware support while leaving that older hardware support behind...

https://www.phoronix.com/news/Intel-Compute-Runtime-Fedora-42

[ScienceMan]

This will break support for a large number of quite recent single board computers, such as Radxa, as well as for some HPC clusters still in service.

[QwertyChouskie]

That kinda sucks, but there's probably not much they can do about it. Hopefully Rusticl can pick up some of the slack.

[SofS]

Why not use meta packages for each case? It should be as simple as installing the proper group of packages, unless there are some complex dependencies.

[pWe00Iri3e7Z9lHOX2Qx]

This will break support for a large number of quite recent single board computers, such as Radxa, as well as for some HPC clusters still in service.

Especially for the latter case, is Fedora a realistic target? I assume those are predominantly RHEL / SUSE / Canonical / Debian.

[Teggs]

This is listing a ton of equipment that isn't past its usable lifecycle. I get Intel's financial issues, but it's raw for a customer to have e-waste that shouldn't have been for a few years yet.

Time-wise, this would be like AMD dropping support for Polaris, or maybe Vega.

Oh, wait...

[FeRD_NYC]

This is listing a ton of equipment that isn't past its usable lifecycle. I get Intel's financial issues, but it's raw for a customer to have e-waste that shouldn't have been for a few years yet.

Time-wise, this would be like AMD dropping support for Polaris, or maybe Vega.

You should've heard the outcry a few years ago, when someone filed a Fedora change proposal looking to upgrade the kernel build optimizations so that AVX2 support would be the minimum requirement for it to run on a machine. The backlash was immediate, and it was severe. Given that (as people pointed out) there were brand-new systems currently on sale at the time that didn't even support AVX2.

Speaking personally, it would've locked all three of my machines out of being able to run newer Fedora releases, as my "newest" system then and now is a 2013 Dell laptop, and my daily driver is an HP desktop ca. 2011. With the changes they were proposing, it was literally a question of being able to run the OS or not: the optimized kernel wouldn't even boot on systems lacking the required support.

Cooler heads prevailed in that instance, and the plan was abandoned very quickly. Though interestingly, it was barely any time at all before Microsoft announced that Windows 11 would require TPM2, also locking out all of my personal machines, along with millions of other perfectly-serviceable computers which (if not for that restriction) would have plenty of useful life left in them.

Presumably the move to a newer Compute Runtime wouldn't be as drastic as those changes, here's hoping anyway. And if there's enough interest in still running the older code on systems that aren't supported by the new packages, and assuming it's possible to package it as an alternative to the newer code, then I'm sure someone will. If not the Fedora package maintainers themselves, then probably RPM Fusion or one of the other third-party groups.

[ms178]

Presumably the move to a newer Compute Runtime wouldn't be as drastic as those changes, here's hoping anyway. And if there's enough interest in still running the older code on systems that aren't supported by the new packages, and assuming it's possible to package it as an alternative to the newer code, then I'm sure someone will. If not the Fedora package maintainers themselves, then probably RPM Fusion or one of the other third-party groups.

While it is not as critical as being able to boot an OS, it is still functionality people on that legacy hardware expected to run longer for some time, considering that Skylake is on that list and is a very recent architecture and along with its legacy 14 nm ++++ cousins, is still in widespread use. Does anyone know if RustiCL is a good alternative for them on legacy hardware? If not, that would be sad as maintaining that older branch could be a nightmare with all of the breaks that could happen over time with a newer toolchain.

[geerge]

So, what does it take to get battlemage running on fedora 41? Is there a distro that has battlemage running well with compute OOTB with no pissing about?

edit: Seems like the answer is Ubuntu 24.10, they release binaries as debs, validated on 24.04 but surely a newer kernel is required for battlemage? Just when I thought I could yeet Ubuntu into the sun.

[NotMine999]

Fedora helping to foster more artificial obsolesence?

Say it isn't so.