OSDev.org

Try getting hold of a PowerPC at a reasonable price!

I’m lucky; I’ve got both. Perhaps I should put my Pi 1 and Pi 2 on eBay and just keep the Pi 3. But I’m keeping the G4 Mini as it has sentimental value.

iansjack wrote:Try getting hold of a PowerPC at a reasonable price!

Well, that was never my point, I was merely pointing out that currently it's very difficult to get your hands on a new Pi.

iansjack wrote:I’m lucky; I’ve got both. Perhaps I should put my Pi 1 and Pi 2 on eBay and just keep the Pi 3. But I’m keeping the G4 Mini as it has sentimental value.

Well, I wouldn't sell them. I've got a dead 2, and a couple working 3s, but I'm not getting rid of them. Besides the sentimental value, I'm also building up the estate for my small computing and telecommunications museum.

BigBuda wrote:At the moment people would most likely be better off with RPi alternatives, like OrangePi, NanoPi, Pine64 and the like (though some might actually need more binary blobs than the RPi)

Some might need less. I recall a friend finding that there were better, cheaper and more open SBCs than the RPi all along. The catch is they were only available in certain countries. The RPi project tried to get its machine everywhere.

You could try a less powerful microcontroller if you don't mind making a 90s or even 80s class OS. It'd expand your hardware options, especially as these things are less absurdly popular than the RPi and presumably easier to manufacture. You don't have to go to my extreme: 128KB ROM and only 16KB RAM, though I'd respect you if you did! At least it's ARM; it's an older model of BBC micro:bit. There's a whole spectrum of microcontrollers from probably less powerful than the least Arduino all the way up to RPi level. A few years ago, I should have bought a nice little ARM board with 512KB RAM and an expansion bus much like the Arduino's. The implication was it could use Arduino peripherals.

Some consumer hardware is reprogrammable. A few years ago, I looked into using a little home-NAS thing; a little white box with a slot for a 2.5" USB drive and several USB ports around the edges. It was based on a Kirkwood ARM SoC. I forgot I've got a couple. I got them years ago because there were instructions on the Net for how to get shell access to the internal Linux and flash your own replacement. I got a shell on one, but the other got bricked and would need me to solder in a JTAG port. I'm told that's not hard. Everyone treats JTAG as pretty easy. The reason I didn't go further is just my health. Same with everything else in this post.

All sorts of other consumer devices may or may not be usable in the same manner. You could search the Net with model numbers of devices you have to see if people have reflashed them before. If not, you could hypothetically, open them up, look up chip numbers to find what SoC they have, get the datasheet and try to work it out. Note that some SoCs are much simpler than others and it doesn't necessarily relate to how powerful they are. I have a Kindle 4, but it's one of the complex ones so I'm looking at other devices instead, even my old TV. Many old TVs can play video from USB drives; they must have an OS already, probably Linux. Security wasn't high on older devices; they may not verify the update file on your USB drive. You might want to get a small UPS though; some of these devices do not like it if the power goes out while they're updating.

iansjack wrote:The only widely available PowerPC would be an old Mac. It’s not easy to find documentation about programming them at a low level.

OpenBSD has ports to some of them, including the brightly-colored ones, worth mentioning because OpenBSD driver source has a reputation for being easier to read than Linux driver source. (It's a less complex OS.)

As for getting old Macs, you could maybe post a want ad locally. Some people might have one or two in their attic and perhaps might not care about their value. I'm not suggesting ripping people off, just finding the sort of people who'd be happy to let someone else make use of their old junk. Eh... making lots of friends might be a better way to do that than posting a want ad. There's a lot to be said for taking an interest in all sorts of people, if they themselves are friendly. Talk to them and see if they're all right with it.

Phones and tablets have a bad reputation for closedness, but you might check out PostmarketOS; it's a Linux distribution for obsolete devices. Sometimes it functions partially; like I could get my old Samsung S3 going with screen touch and USB supported, but not the radios. I could, hypothetically, read its source code to write drivers for my own OS. Then again, some say "Reading Linux source code is where the fun stops." That's why I put this last.

You can also run FreeBSD on a G4 Mac. (I run it on a PPC Mini.) But that’s not “low level”. It’s a lot harder to find information about reading/writing a console, or accessing disks.

A fun alternative is an FPGA-based RISC-V processor. I’m looking at this on a Tang Nano 9K board (£20 or so). Something new to learn, and really low level. Ultimately I want to run my own processor design on it.

iansjack wrote:Ultimately I want to run my own processor design on it.

Been wanting to do this for over 25 years. Life never seems to give me the opportunity. There's always some other obligation in the way.

BigBuda wrote:

iansjack wrote:Ultimately I want to run my own processor design on it.

Been wanting to do this for over 25 years. Life never seems to give me the opportunity. There's always some other obligation in the way.

Tell me about it! We'd live in a world of wonders if everyone had time and ability to create. Of course, some people would use their ability to create to bring everyone down again. So it goes.

I did find the energy for some very basic processor design once; on paper only (or rather text file). As basic and unfinished as it was, it taught me a bit.

It's actually a lot easier than it might seem if you use Verilog and an FPGA. I was inspired by this blog: https://learn.lushaylabs.com/tang-nano-series/ to dip my feet into the water. I got a Tango Nano 9K board and I'm learning Verilog. One of the examples supplied by the manufacturer is a pico RISC V; it occupies about half of the FPGS's resources with 22K of RAM. After nearly 80 years I'm always up for learning something new. I'm still working on my instruction set, but I think it will initially be a very simple 32-bit RISC design. I've also got thoughts about a stack-based design to run a Threaded Interpretive Language natively.

Verilog is much easier than constructing the circuit from scratch (and it's a lot easier implementing it on an FPGA than with discrete chips); it's like using C rather than assembler to write your OS. But if you want to do it the hard way you can convert a Verilog design into a logic-gate design.

If you want to do it, do it now before time catches up with you.

iansjack wrote:It's actually a lot easier than it might seem if you use Verilog and an FPGA. I was inspired by this blog: https://learn.lushaylabs.com/tang-nano-series/ to dip my feet into the water. I got a Tango Nano 9K board and I'm learning Verilog. One of the examples supplied by the manufacturer is a pico RISC V; it occupies about half of the FPGS's resources with 22K of RAM. After nearly 80 years I'm always up for learning something new. I'm still working on my instruction set, but I think it will initially be a very simple 32-bit RISC design. I've also got thoughts about a stack-based design to run a Threaded Interpretive Language natively.

Verilog is much easier than constructing the circuit from scratch (and it's a lot easier implementing it on an FPGA than with discrete chips); it's like using C rather than assembler to write your OS. But if you want to do it the hard way you can convert a Verilog design into a logic-gate design.

If you want to do it, do it now before time catches up with you.

That board is a bit too basic for me. I use the KC705: https://www.xilinx.com/products/boards- ... 705-g.html. A bit more expensive, but more powerful and can be interfaced to an OS either with PCIe, network or USB.

The idea of a CPU core on a FPGA is not my favorite. I prefer to interface with PCIe to my own OS. That way I can focus on Verilog and the FPGA for the "number crunching" stuff, and more general analysis can run in user mode in my OS using C++. After all, programing with Verilog has it's problems for more general stuff. You also don't want to use most of the resources in the FPGA for a CPU core. You want to use that for "number crunching"!

iansjack wrote:It's actually a lot easier than it might seem if you use Verilog and an FPGA. I was inspired by this blog: https://learn.lushaylabs.com/tang-nano-series/ to dip my feet into the water. I got a Tango Nano 9K board and I'm learning Verilog. One of the examples supplied by the manufacturer is a pico RISC V; it occupies about half of the FPGS's resources with 22K of RAM. After nearly 80 years I'm always up for learning something new. I'm still working on my instruction set, but I think it will initially be a very simple 32-bit RISC design. I've also got thoughts about a stack-based design to run a Threaded Interpretive Language natively.

Awesome! I hope to get around to this eventually, though the last thing I need right now is another change of plans. I have to lay down a lot, so I've been interacting with people all the time I'm up and not wanting to type while laying down. I've got a Twidler one-handed keyboard which I would use laying down if I could use it at all; it hurts my fingers. The irony is the plans I have include a 1-handed keyboard which would be much more comfortable for me, and I'm not getting on with them because I need a 1-handed keyboard. I'll have a go at them today instead of socializing, and see how I do.

Edit: Split keyboard ariving tomorrow, yay!

iansjack wrote:Verilog is much easier than constructing the circuit from scratch (and it's a lot easier implementing it on an FPGA than with discrete chips); it's like using C rather than assembler to write your OS. But if you want to do it the hard way you can convert a Verilog design into a logic-gate design.

Interesting! I didn't know Verilog is higher-level than logic gates.

Given this thread topic, I think I ought to mention that a complexity-hating former friend hates Verilog, preferring VHDL. I have no idea how subjective or objective this is, but complexity-haters might want to compare the two languages to see how they get on. It didn't stop my friend from implementing all sorts of stuff in Verilog for a SBC, so perhaps it's not too bad.

iansjack wrote:If you want to do it, do it now before time catches up with you.

Yes indeed.

Great - but it’s more than 100 times more expensive than the Tang, which suits my needs. A bit like the difference between a Pi (or similar) and an Apple Mac Studio if you want to play with an ARM chip. $25 I can justify, $2748 less so.

I started with Verilog because there is very good support for it in VS Code. I’ve briefly looked at VHDL and plan to go deeper into it when I’m more au fait with the technology. It appears to be much more verbose than Verilog.

iansjack wrote:Great - but it’s more than 100 times more expensive than the Tang, which suits my needs. A bit like the difference between a Pi (or similar) and an Apple Mac Studio if you want to play with an ARM chip. $25 I can justify, $2748 less so.

Well, the KC705 suits my needs because I'm creating a spectrum analyser that should run close to 1G samples per second, and I don't think the Tang could be interfaced with a GHz ADC, nor do I think samples could be analysed in real time with that board since it has no connections to the outside with enough bandwidth.

BTW, I don't think CPU cores in FPGAs are up to the challenge either since they run at a too low frequency and cannot parallelize operations like Verilog can. A 12-core thread-ripper processor is far more effective at general processing tasks than a CPU core in a FPGA.

For more simple projects without "data crunching" needs, I'd use a PIC MicroChip processor instead. At least some models come as hole-mounted, and so are suitable for home-made electronics.

rdos wrote:Well, the KC705 suits my needs because I'm creating a spectrum analyser that should run close to 1G samples per second, and I don't think the Tang could be interfaced with a GHz ADC, nor do I think samples could be analysed in real time with that board since it has no connections to the outside with enough bandwidth.

This is why I envy the pros; they get to work with stuff which costs that much -- or sometimes a lot more! I have no need of the power though. I'm evaluating a 16MHz ARM device which should give you an idea of what I actually need. Faster would be nice, but this is project 1.

rdos wrote:For more simple projects without "data crunching" needs, I'd use a PIC MicroChip processor instead. At least some models come as hole-mounted, and so are suitable for home-made electronics.

Everyone talks about through-hole mounting as if surface mount required Awesome Robot Technology, (lol) but my old friend actually preferred surface mount for his hobby. The parts stick down, you apply solder paste, and put the whole board in an oven. People use those little tabletop ovens. If I remember right, you can get a little hand-held suction thing to place the parts, almost like a pen with a tube out the top. I guess it means you have to make a PCB though; no strip-board or pen wiring.

I know I'm breaking my resolve, but my split keyboard arrives tomorrow -- a big improvement on the end of the month.

eekee wrote:

rdos wrote:Well, the KC705 suits my needs because I'm creating a spectrum analyser that should run close to 1G samples per second, and I don't think the Tang could be interfaced with a GHz ADC, nor do I think samples could be analysed in real time with that board since it has no connections to the outside with enough bandwidth.

This is why I envy the pros; they get to work with stuff which costs that much -- or sometimes a lot more! I have no need of the power though. I'm evaluating a 16MHz ARM device which should give you an idea of what I actually need. Faster would be nice, but this is project 1.

It's actually not a pro project. I also bought an ADC that cost about the same as the KC705 and a 12-core thread ripper PC with 128 GB of memory that was not cheap either. It's a bit of expensive project, but it is pretty important for me.

eekee wrote:

rdos wrote:For more simple projects without "data crunching" needs, I'd use a PIC MicroChip processor instead. At least some models come as hole-mounted, and so are suitable for home-made electronics.

Everyone talks about through-hole mounting as if surface mount required Awesome Robot Technology, (lol) but my old friend actually preferred surface mount for his hobby. The parts stick down, you apply solder paste, and put the whole board in an oven. People use those little tabletop ovens. If I remember right, you can get a little hand-held suction thing to place the parts, almost like a pen with a tube out the top. I guess it means you have to make a PCB though; no strip-board or pen wiring.

I've manufactured pro-quality PCBs using Macaos, but I'm a bit afraid of using surface-mount components. The Kintex-7 FPGA on the KC705 has some 500 pins, so I have no idea how to make a working PCB for that, and how to solder it, so I stick with evaluation boards.

rdos wrote:A 12-core thread-ripper processor is far more effective at general processing tasks than a CPU core in a FPGA.

I think you rather miss the point. I'm not trying to produce the next Threadripper or i9 anymore than I'm trying to produce the next OS X or Windows.

I'm just having fun.