NetBSD on the last G4 Mac mini and making the kernel power failure proof • Hacker News Board

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

#41

Post by MrRadar » Sun, Jun 30, 2019, 12:17 AM UTC

darkpuma wrote: post #38

classichasclass wrote: post #33

Earlier quoted context omitted.

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…

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.

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

#42

Post by verisimilitudes » Sun, Jun 30, 2019, 7:47 AM UTC

mikece wrote: post #4

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!

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.

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

#43

Post by Dylan16807 » Sun, Jun 30, 2019, 12:10 PM UTC

verisimilitudes wrote: post #42

mikece wrote: post #4

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!

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…

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.

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

#44

Post by davesmith1983 » Sun, Jun 30, 2019, 12:19 PM UTC

oneplane wrote: post #24

WWLink wrote: post #19

Earlier quoted context omitted.

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.

I use my Amiga 1200 a lot. It isn't fast or efficient. It is a vintage piece of hardware which I enjoy using.

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

#45

Post by oneplane » Sun, Jun 30, 2019, 6:06 PM UTC

dvdbloc wrote: post #29

oneplane wrote: post #18

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.

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).