NetBSD on the last G4 Mac mini and making the kernel power failure proof | Hacker News Reader

NetBSD on the last G4 Mac mini and making the kernel power failure proof

tenfourfox.blogspot.com

125 pointsUkiahSmith40 comments

For those who are not familiar with the author of this blog, he is a PowerPC enthusiast and the maintainer of TenFourFox, a Firefox port to PPC OS X. He's also currently working on a new PPC JIT for (modern) Firefox on Linux (a project for which he is looking for volunteers to assist him[1]) to bring it to speed parity with Chrome on PPC (which already has a PPC JIT).

[1] https://www.talospace.com/2019/03/pitch-into-firefox-jit.htm...

What is the last supported OS X version for PowerPC processors?

10.5.8 "Leopard"

Do you mean webkit on ppc (since I searched but couldn't find a chrome port)

There is a community Chromium port to PPC Linux[1]. Modern PPC Linux distros should have it in their repositories. So far Google has not accepted the patches to enable PPC support upstream[2].

[1] https://wiki.raptorcs.com/wiki/Porting/Chromium

[2] https://groups.google.com/a/chromium.org/forum/#!topic/chrom...

I just picked up a G4 iBook and am grateful for TenFourFox. Combined with uBlock Origin legacy builds, it’s actually usable for some modern sites, which is pretty impressive.

Author here. Thanks for the kind word. :)

Why the obsession with PPC? Don't you feel you could put your time + effort into something more modern that isn't dying?

Why do you say it's dying?

PPC is far from dead. You can buy brand-new POWER9 workstations with up to 44 cores (with SMT4 for 176 threads) from Raptor Computer Systems. They are also about as open as a modern computer can be with board schematics included with every system and only one component with closed firmware (the BCM5719 network controller, which is currently being reverse-engineered so an open firmware can be developed).

Wow, didn’t know that, amazing. What is the main use of those machines?

They are primarily intended for people who need lots of threads for high-end workstation uses (see Phoronix's benchmarks[1]), people who want open computers (whether for security/auditability or ideological reasons), or just people who want something different from the mainstream.

[1] https://www.phoronix.com/scan.php?page=article&item=power9-t...

Found the store! https://www.raptorcs.com/content/base/products.html

The Talos II I'm typing this reply on is just my regular workstation. I have a dual-4, which is 32 threads, and it runs Fedora. Most things work just fine and it's a very nice daily driver. I'll probably upgrade it to a dual-8 in the near future for even more parallel goodness.

I also have a Blackbird, which does streaming video in the home theatre and is also a test system. It's a single-4, but it can take an 8-core part if desired.

With that kind of hardware, is hardware decoding and encoding of media a possibility? A DLNA server with hardware transcoding is something I've been thinking about recently.

You can use AMD graphics cards with their open-source drivers. I'm not sure how well hardware-accelerated video works with them but it should work the same as it does on x86 Linux. Multimedia is probably one of the weaker areas for Power since there's so little focus on writing Altivec SIMD code for the platform from the larger open-source multimedia community (most of their focus is understandably on x86 and ARM), though that wouldn't impact hardware en/decoding.

Nonsense. PowerPC in old PowerMac and modern POWER 9 have about as much in common as Pentium Pro and a modern Xeon!

PowerPC is a dead platform, POWER9 is not - but they are not the same!

I'm not sure that's an apt comparison. If you were saying that 32-bit PPC is dead, then the answer is probably yes at least in general computing, though there are still lots of 32-bit PowerPC parts in embedded systems. However, 64-bit PowerPC systems have substantially more in common with "big" POWER. The G5, which was clearly positioned as a member of the PowerPC family (PPC970) but descended from the POWER4, would be the classic example.

PowerPC is dead for the larger market but sticks around in specialty markets. Mainly NXP (formerly Freescale):

https://www.nxp.com/products/processors-and-microcontrollers...

Also, the separation kernels that met NSA's highest standards of security initially targeted that for cross-selling into aerospace market. They buy a good chunk of PPC for some reason. Example that was among first certified with data on what was expected:

https://www.ghs.com/products/safety_critical/integrity-do-17...

The same reason some people still use Amigas with 680x0s... rampant nostalgia.

I kinda don't get it on a personal level, but accept that different people are into different stuff and sometimes people get really attached to specific things, and sometimes those things are computer architectures.

Somewhere there are chip designers thinking they can create the first processor on which NetBSD cannot run. Have fun kids: many have tried that and failed!

Mill architecture?

Maybe. NetBSD doesn't have an equivalent of the nommu variant of Linux.

Not every chip mimics a PDP-11 or can be made to resemble one. Chuck Moore's GA144 is an array of his F18A Forth chips, which are all rather small and simple stack machines that support such fine grained operations as looping the instructions that share a machine word.

Given that this is a heavily segmented machine, with very different primitive operations than a PDP-11, and also has a word size of eighteen, I don't imagine C would run very well on it at all, if you even wanted to try.

So, there you have it.

I'd say that the stack machine in the GA144 is feasible to compile normal C to. And the 18-bit word is barely a problem. But the (complete lack of) memory architecture is a clear dealbreaker. There's no compiling a normal C program down to an architecture that has 64 words of RAM. At best you can gang a few cores together to make an emulator for a simple virtual machine.

As fun as this is (and as much as I appreciate putting good old hardware to use and as much as I like PPC): the power consumption alone wouldn't make this a winner :(

I have a PowerBook G4 DVI, iBook G4, B&W G3, GbE G4, Dual G5 here, I want to get rid of them but I like them too much. Yet I can't run them for fun all day long either because the amount of power they eat >:( The problems of a hardware connoisseur.

o_O Compared to what? G3s and G4s are pretty light on power usage - IIRC they use around 10-15w? The G3s were even less.

The i7 in my desktop uses at least 95w, and some reviews suggest it's more like 140w lol.

Edit: Unless you're pairing them with CRT monitors...

Compared to say.. a 1W ARM CPU. Those are faster and use less energy.

Author here. Power consumption on this unit is about 21W.

Is the CPU/watt ratio less than later units? You bet. But it's not bad, it's free, it's not in a landfill, and it's more than enough grunt for the basic tasks it's doing.

Yeah, those Mac Minis can go quite low. But still the CPU power from any PPC vs. an ARM or low-end x86 is a bit inverted. A low-end ARM or x86 that is faster than any of the PPCs while using less than half the power (in the 1W to 8W range) does make a dent.

I have the same problem with two G5 Xserves. They’re just too cool to get rid of.

Yeah, same problem here with some older Xeon Xserves. Still running (but Linux and Xen at this point). Without macOS you lose the cool CPU LED indicators at the front, but you still have everything else, including IPMI. They are way too neat to get rid of, specially when they still run (which for this 1U deal is fine because the power cost is fixed anyway).

The fire extinguisher on the desk is a nice touch; though I'm not sure that's enough volume to extinguish a G4 fire.

I don’t recall the G4 PowerPC designs being unusually hot running by standards of the day? The G5 was of course, but it was much worse in this regard than the G4.

Yeah, the G5 was a scorcher. But I remember my G4 laptops turning my thighs red, too.

As an owner of both G4 and G3 laptops, I can assure you, eggs will fry given the right task... The G3 was actually worse.

Long way from reports of fires. I’ve only heard of lithium batteries doing that

No it's the chips, these machines didn't exactly have the most user friendly heat dissipation design, on both models the CPU heat conducted straight through to the top case. maybe not cook an egg very fast but uncomfortable to touch for more than a second... maybe 1 EPH (one egg per hour) :)

The early Intel MacBooks weren't much better - my black 2007 2,1 model gets pretty hot and has a serious lack of ventilation holes.

Now the old Powerbook 5300's actually did catch fire. My 1st tech job was a hardware technician at an Apple production facility mostly building and refurbishing Powerbooks, can confirm.

Reply on news.ycombinator.com