Linux Kernel Developers Discuss Dropping A Bunch Of Old CPUs

[Old Grouch]

Self-deleted irrelevant comment that duplicated information in original article.

[hotaru]

Also 486 plus Linux is massively used on industrial machinery too! It was officially manufactured till 2007, and is still being made as clone CPUS for robotic arms, pe. So I am not sure they'll drop 486 support. Specially considering it is a very modern and as generic as it goes (in instruction set) X86 architecture.

no, it's not. I've worked on stuff like that and I've seen lots of ARM and MIPS, even some SPARC, but no x86 at all. I don't doubt that it does exist, but it's definitely not common, and modern ARM chips have much lower power consumption for the same performance as a 486.

if your concern is keeping it out of a landfill, recycling the materials it's made out of is a better way to do that.

And don't mention people who simply can't afford upgrading.

if you can't afford $35 to upgrade to a massively more performant Raspberry Pi that uses a lot less power, you probably also can't afford the electricity to run your 486.

a 486 might have some value as a museum piece, but for that it'd make a lot more sense to run a period-appropriate operating system, such as MS-DOS.

[L_A_G]

Is there any benefit to removing these other than less lines of code in the kernel, and less to maintain?
Example maybe removing 486 will that somehow maybe be possible to improve the kernel in some way to make it cleaner, faster, more stable or more secure?

Considering the 486 is right now the baseline for x86 it probably will allow some cleanups to be done where paths using features like vector instructions could be made the default on x86. It'll also allow developers to assume the presence of these instructions and hence write faster code in the future. This is the reason why Windows has required at least an SSE3-capable CPU for years.

However, it's a bit of a red herring as the real focus here isn't on x86, but on some pretty obscure CPUs from dead end architectures like Motorola 68k, MIPS, IA64, SH-2, CELL and DEC Alpha. I'm actually surprised that the kernel ever supported SH-2, a.k.a the CPU core used in the Sega Saturn and 32X.

[vladimir86]

no, it's not. I've worked on stuff like that and I've seen lots of ARM and MIPS, even some SPARC, but no x86 at all. I don't doubt that it does exist, but it's definitely not common, and modern ARM chips have much lower power consumption for the same performance as a 486.

I agree that a low powered ARM could be a better fit to do what many 486 do now on industrial machinery, but the truth is that it wasn't when those where built 30 - 20 years ago. Machinery on a factory can and will last for decades (in some extremes, like rubber production, you can actually see machinery built in the XIX century). Many modern machines also need to talk with those old computers in other machines. It is not rare to need to do a serial connection with a machine using not a 486, but a 4040 or 8080 cpu in 2021!

if you can't afford $35 to upgrade to a massively more performant Raspberry Pi that uses a lot less power, you probably also can't afford the electricity to run your 486.

That is totally out of scope: Some people don't have money to upgrade and that's it. I am talking people who cannot afford to buy a new device but have access to some old hardware. Also, the Pi might cost 35, and it is a really cool machine, but you till need a monitor, keyboard, mouse, charger.... to make it usable might go quite more expensive! I couldn't had afforded it as a student, pe. I relied in old rejected computers (this is where my fixing old stuff TOC came, I guess).

a 486 might have some value as a museum piece, but for that it'd make a lot more sense to run a period-appropriate operating system, such as MS-DOS.

That's also a totally different ship: My point was to use ancient hardware and make it work for modern computing. But I agree, that worst come I can always use it for DOS and games. My beef with that is simply that emulation is very good today to run DOS software, but useless to emulate Windows 9x, so that little machine will be cool and all for DOS and Win95 as far as I don't try anything 3D like Quake, but it is a bit redundant with DosBox. If it was a Pentium 1 or 2, it'll be able to fill the Win95/98 gap that no Virtual machine can nowadays (You can do a W9x VM allright, but with no support they run too slow to be useful)

[kpedersen]

Why do you think that Linux shipping on billions of mobile and embedded devices every year is "fairly meaningless"?

Yes. When was the last time you installed Linux on a mobile?

You misunderstand, I like Linux. I dislike locked down gizmos and phones.

Removing older CPU support (i.e x86 < i686) because it is "less common" than aarch64 is an incorrect thing to do. What you would be ending up doing doing is decreasing the number of "open" platforms that can run Linux and leaving only locked-down sh*tware.

Mobile operating systems are meaningless and should not get counted in any decision when it comes to Linux. Not until atleast 1% of them are "open" enough that you can install your own Linux on them.

Linux will end up being the greatest OS that no-one but OEM companies can run XD

[jabl]

Yes. When was the last time you installed Linux on a mobile?

Why is that relevant? Clearly Linux is widely used outside hobbyists installing it on their own PC's, and brings value to people. Who are you to say it's meaningless?

You misunderstand, I like Linux. I dislike locked down gizmos and phones.

"That's just like, your opinion, man"

Removing older CPU support (i.e x86 < i686) because it is "less common" than aarch64 is an incorrect thing to do. What you would be ending up doing doing is decreasing the number of "open" platforms that can run Linux and leaving only locked-down sh*tware.

Ultimately it's up to the developers and users. The developers are stating that they think nobody is using those targets anymore. Unless somebody steps up and says "I still use it, it's useful for reasons X, Y, and Z, and I'm prepared to, if not participate in development directly, test new kernels on that hardware and report breakage", then it will be removed. No point in keeping dead code around.

Mobile operating systems are meaningless and should not get counted in any decision when it comes to Linux. Not until atleast 1% of them are "open" enough that you can install your own Linux on them.

Again, the world does not revolve around your personal preferences.

[Aryma]

why would you run a calculator with new kernel anyway ?

[kpedersen]

why would you run a calculator with new kernel anyway ?

Just because a machine is not a phone, doesn't make it a calculator. There is more to Linux and computing than consumer phones

Again, the world does not revolve around your personal preferences.

Absolutely not. I guess it is just my way of explaining to people why old platform support remains in the Linux kernel well beyond our lifespans. Because luckily most other engineers think the same way as me. And is why I think it is absurd that many people moan about Linux supporting x86 and yet not their silly little locked down ARM gizmo.

[xorbe]

I worked on SA1100/1500, rip.

[cb88]

sun4m and sun4d are dropped also... even though they are really the only standard full desktop/workstation implementations on sparc 32bit.