Linus Torvalds Comes Out Against "Completely Broken" x86_64 Feature Levels

[phoronix]

Phoronix: Linus Torvalds Comes Out Against "Completely Broken" x86_64 Feature Levels

With the new Linux kernel patches posted yesterday for cleaning up x86 32-bit kernels on x86_64 CPUs as part of that patch series was introducing new Kconfig build options around the x86_64 micro-architecture feature levels. It turns out though that Torvalds is completely against how the x86_64 feature levels are handled by the compiler toolchain folks and doesn't want to see it invading the kernel...

https://www.phoronix.com/news/Torvalds-Mind-Fart-x86_64-Level

[rene]

finally, I told you so for years since this not reality matching nonsense was introduced, ... https://www.youtube.com/watch?v=-bmeuRIARQc

[mxan]

"I have no idea who came up with the "microarchitecture levels" garbage, but as far as I can tell, it's entirely unofficial, and it's a completely broken model."

Intel themselves created it, along with AMD, Red Hat and SUSE...

https://developers.redhat.com/blog/2021/01/05/building-red-hat-enterprise-linux-9-for-the-x86-64-v2-microarchitecture-level

[mxan]

finally, I told you so for years since this not reality matching nonsense was introduced, ... https://www.youtube.com/watch?v=-bmeuRIARQc

No one in the real world cares about obscure implementation details. Compiling for more advanced instructions available on newer processors = faster distro and faster software, it's really that simple. Linux can't be hamstrung by ancient hardware forever, that's what NetBSD is for.

[carewolf]

I mean by all means get rid of v4, it was always speculation, and reality has gone a different direction. But v2, and v3 are somewhat useful outside of the kernel.

[davidbepo]

a sensible simplification for ISA optimization levels is introduced in collaboration with the creators of the ISA itself and successfully used in many actual applications
Linus: REEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE

[avis]

Where is CoC(k)?

Why is Linus still committing to the kernel? Oh, wait, without him everything will fall apart.

[patrick1946]

"I have no idea who came up with the "microarchitecture levels" garbage, but as far as I can tell, it's entirely unofficial, and it's a completely broken model."

Intel themselves created it, along with AMD, Red Hat and SUSE...

https://developers.redhat.com/blog/2...itecture-level

I think he was again not in context. The micro architecture levels are an useful simplification to generate few optimized binaries. The kernel is not using this features. Other software is!

[patrick1946]

I mean by all means get rid of v4, it was always speculation, and reality has gone a different direction. But v2, and v3 are somewhat useful outside of the kernel.

Actually v4 was useful until Intel could not get their little cores up to that level.

[ms178]

Florian Weimer must be sad about Linus' comments, as he was spearheading this effort on the Glibc side.

And honestly, I don't get the logic of Linus here as march=native might be enough for individual users but not a suitable target from a distro perspective. These feature levels are meant to group certain CPU architectures from roughly the same era together. Yes, there are some rare exceptions but overall it would serve its purpose to lower the maintenance burden on the Kernel side while giving people with capable CPUs more instructions to work with with the potential to increase performance in some areas.

As far as I see it the good old graysky2 patch is still needed after this cleanup to be more flexible.