PowerPC 601 Support Being Retired In Linux 5.10 - The First 32-bit PowerPC CPU

[edwaleni]

I think someone still makes the PPC 604e that has been hardened for defense and space use.

I remember IBM coming around to demo the 615 with that 486SLC core. Ran 16 bit Windows. Interesting technology.

The problem wasn't the Power CPU part, it was the weak caches IBM put into that 486SLC. 486SLC, the little CPU that couldn't. It even supported multi-socket SMP!

Then they came around showing OS/2 on Power for running Notes and Domino. That was the end of the Power demos in our world.

I never heard if the Novell on Power ever reached Alpha. IBM talked about it.

[mangeek]

I still have Xenial installed on an old PowerPC G3, and even I can get behind this. I'm thinking it might be time to draw a line in the sand across the whole kernel where a 'super-duper-longterm' branch can be established for hardware tinkerers and developers with long-term product support in mind, and a whole bunch of stuff can get ripped out of kernels going forward. I'm imagining that a whole bunch of arch/platform/hardware support for things that are genuinely ancient could go, maybe even going as far as establishing a floor for going forward at x86_64/Arm64/PPC64/etc., and dropping pre-AHCI ATA/SATA, AGP, etc.

Maybe Linux '6.0' would be a good place to do that? I dunno, I'm not a developer. I just imagine that a LOT of compromises have been made over the last few decades to make Linux work on 'everything', and 'everything' is a lot smaller if you can drop ancient platforms and drivers.

[stevewmn]

The first Linux kernel compile I ever did was for a PowerPC Mac. I think it might have been a Power Mac 7100/66. I started with a Mach kernel based Linux (MkLinux?) and then moved to Yellow Dog Linux. I'm pretty sure it was something in the 2.2.x or 2.3.x series kernel.

[dwagner]

I remember attending a "BeBox" presentation, where the makers of "BeOS" boasted about their PowerPC-based hardware being able to run two threads simultaneously on two CPUs. It wasn't a very convincing presentation, as nothing was presented really that traditional Unix systems on traditional multi-CPU boards could not also do.

I think I still use a PowerPC CPU, though, as part of my laser printer.

[onicsis]

I remember that time well. The PowerPC was largely a victim of its marketing. Initially, they (especially Apple) ran hard with the notion that the "PCs" of the time were obsolete dinosaurs. Which they were, but the first version of the PowerPC platform, "PREP", was in fact identical. It even had an ISA bus for crying out loud. What was worse, the only OS you could really use on it was Apple's MacOS for the M68000, running (extremely slowly) under CPU emulation. No-one was fooled.

The second iteration (the "CHRP") was closely related to the Macintosh clones that existed back then, but they were expensive and quickly losing ground. Even though more parts (but not all) of the MacOS were ppc native at that stage, by then the Pentium Pro and PII machines were outperforming them hands down and for a much lower cost.

Basically was victim of proprietary, expensive and closed source software platform.
Everything closed source, the operating system, libraries, compilers.

[chleroy]

hard to tell.

the single commit however, is just -52 lines (and +9)

https://lkml.org/lkml/2020/8/13/570

The first series (the one removing the powerpc 601) removes 403 lines while adding 53 lines.
The second series (cleaning timebase helpers) has 38 deletion and 15 insertion.

That's not a huge number of lines, but that was burden everytime a new feature was to be added, because of the small but important differences between the 601 and other 6xx variants.

[chleroy]

I've got a few of those old dbox 2 mpeg2 satellite/cable dvb decoders running on out of tree kernel patches.
Good old enigma times watching pirate TV channels.

But now that I look at it, they feature the powerpc 8xx series.

Allthough powerpc 8xx family is from the same generation as 601, powerpc 885 was under production until last June.

It will remain actively supported in the kernel for some time.

[SteamPunker]

It's kind of funny how support for the M68k architecture, including even the original 68000, is still going strong in Linux, GCC and LLVM, whereas this first generation of the PowerPC architecture, that succeeded 68k, is now getting the boot.

[mangeek]

It's kind of funny how support for the M68k architecture, including even the original 68000, is still going strong in Linux, GCC and LLVM, whereas this first generation of the PowerPC architecture, that succeeded 68k, is now getting the boot.

Wild guess, but I think that beyond the world of desktop computing, the m68k had wide distribution in embedded devices, many of which still live on today with service contracts. The PPC was less ubiquitous in its desktop form. The 601 was in the first-gen Power Macs and a few RS/6000 servers, if I recall correctly. The MC680x0 series was in all sorts of things, from Cisco routers to embedded medical and HVAC control devices.

I don't think it would be a Bad Thing to consolidate. I'm guessing that low-level arch support code brings implications and workarounds with it elsewhere that might be able to be left behind after cuts like this are made. We might want to let the money talk on this and keep PPC32 as an arch, but let vendors with interest maintain any specifics beyond the PPC750 (e.g., make IBM do anything more than 'let us emulate a G3').