Re: This is only half BEAM related.
- --- In email@example.com, BR wrote:
>No, I mean to make the chip or CPU OUT of H-switches. And maybe for some strange reason I do want to. For sake of experimentation.
> On 2012-12-11 8:53 AM, connor_ramsey@... wrote:
> > So I was wondering: could an H-switch network be used
> > to process binary logic?
> Why would you want to?
> > Because H-switches ARE digital circuits,
> Not really. They're actually conceptual black boxes with binary (or
> logical) inputs and outputs.
> > and they've already proven themselves to be effective
> > CUs(Control Unit).
> Got any examples? I've been out of touch.
> > So my question is: is it possible that if H-switches
> > were to be wired up in such a way that the the network
> > would perform as an ALU? Would 00001010 and 10000101
> > give you 00000101(would 10 and -5 give you 5)?
> Why not just use binary logic chips, or better yet, a CPU? It'd be much
> more compact.
> > And would the CU be able to process the machine
> > instructions it was being given into more machine
> > instructions?
> Yes, but again, why not use binary logic, or a processor?
> > And if an H-net were capable of doing all these things,
> > would it be able to do it in fewer steps than a traditional
> > processor, or do it with far fewer inverters?
> It would depend entirely on the implementation.
> > And can an H-net be programmed?
> If you implemented it in a CPU, it could certainly be programmed; but
> what would you gain from it?
> > Let's think about that.
> I did, quite a while ago.
> The H-switch is a concept, that's all. It's not a physical device, just
> a shorthand way of systematically defining a hierarchy of behaviours.
> If I wanted to experiment with a modest hierarchical control system, I
> might be inclined to build some H-switches using AND gates and NOT gates
> (i.e., inverters), or possible just inverters, diodes, and resistors.
> But I'd only do that if I figured I would be changing things around a lot.
> I doubt that I would ever build a physical device using actual
> H-switches. Too many components.
> For what it's worth ...
And if you search H-switch on the BEAMwiki, you'll see that the circuit for your H-switch is actually very simple. You may only see it as a concept, Bruce, but I think that you're on to more than you thought you were. There's a circuit on the wiki showing that an entire 3-stage H-net can fit on just one AC240.
But oh well, I've moved past the idea, in favor of AND gates and flip-flops. If you take a look at DigiComp1, and intrepret it's mechanisms into electronic circuits, you'll know what i mean.
I figure that a machine like that would be WAY simpler than a traditional computer(just a couple levels above BEAM)and just as powerful, because there are few gates the signals have to traverse through the processing delay would be much greater, and adversely the clock frequency of the machine could be allowed to be MUCH higher.
I'm building a simulated version of a 4-bit model of this computer(named MineComp1, since its based on DigiComp1 and because I'm building it in MineCraft, out of redstone). My version also includes an automatic NOR gate shut-off that stops the computer when the computation is done, something DigiComp1 lacked(but I suppose it didn't need one since the clock was manually operated), and likely some ROM for the base programs the computer needs to compile and execute user-made programs, EEPROM or something to store the executable code for these same user-made programs, and RAM to store results from previous computations, since the 4 flip-flops in the computer provide all the needed memory the computer needs per computation(which can span few or several cycles).
And with the new update to the game coming out soon, I'll be able to start building some fun analog machines too. ;D
But enough of that. When I fix my soldering iron and get an actual workspace, I'll finish up HAWC, make a mechanical robot out of k'nex maybe(I want to use the digicomp1 built out of k'nex as its brain.), perhaps do the same thing out of legos, make a marble based 3-symbol 2-state universal Turing machine, etc.
Most importantly, I'm going to start working on an FLED based walker, where the ENTIRE circuit is made primarily out of FLEDs, capacitors and diodes in series. I'll tell you how it goes.