Forums | MacLife

Fried Chicken wrote:

Holy smurf!!!!  Some guy built a computer out of relays.
Aside from the mind blowing complexity of the machine, the amount of time to build a machine like this is just crazy.

I have absolutely no idea how computers work on an electron level.  Jesus, I can't imagine what kind of knowledge it takes to put something like this together.

Yeah, that is pretty cool.  It's really not as complex as it looks, though.  Any graduate of a decent CS/CE bachelor's program should be able to build that (not that you need a college degree; a lot of geeks are self-taught).  It would take a ton of time, patience, and relays, though, so it's still impressive, and the flashing lights and clicking are fun.

This is pure awesome.

RE Knowledge: Some basic EE Theory and CS/CE Theory should get the ground work taken out of the way. Some diagrams and flowcharts will really help things out here as well. I would run it through some sort of simulator before I even touched a real relay.

This is how wozniak started out, and relays are a step up from the tubes the first univac and eniac were made out of.

What I don't understand is how you can program something like this.... or teach a computer to recognize a programming language.

Which came first?  The chicken or the egg?

mo' ron wrote:

Fried Chicken wrote:

What I don't understand is how you can program something like this.... or teach a computer to recognize a programming language.

Which came first?  The chicken or the egg?

It's not that complicated actually...

I'm sure your college offers an "intro to computer organization" class that makes it easier to understand.

Also, you don't "teach" a computer to recognized a programming language, computers are built with the specific purpose of being programmed, and all programming languages are abstracts to the machine language (the literal 1s/0s that you have to clock in to make the computer do stuff). The hard work comes in with figuring out how to abstract a language like C to "compile" to a particular computer's machine language.

I think what made me "get" it was when I learned that in binary, adding 2 numbers together was merely an exclusive-or ( http://en.wikipedia.org/wiki/Bitwise_operation#XOR ) comparison, and the basis of probably most computer operations is adding things (whether positively, negatively, or recursively which gives you subtraction/division/multiplication). Things are more complex than this in modern processors, but you can build a computer just knowing this principle.

I can sort of follow that... but how do you get the computer to, for example, find something in a piece of memory, or allocate memory.
I just took programming this year, but I cannot imagine how some of the most basic commands in java can be executed in a computer on the basic level.
How do you assign a computer to display a piece of text for example?

Fried Chicken wrote:

I can sort of follow that... but how do you get the computer to, for example, find something in a piece of memory, or allocate memory.
I just took programming this year, but I cannot imagine how some of the most basic commands in java can be executed in a computer on the basic level.
How do you assign a computer to display a piece of text for example?

It depends what you mean by "finding" memory, if you mean fetching memory, that's easy. If you mean actually finding memory, that's much more complex.

Just imagine you have 2 bits of memory. When you want to select the first one (address "0") you just have to wire that first bit/memory (let's assume you have your memory structure in place already) to an [AND] gate with 2 inputs, and on the second input, have your address line but with an inverter on it (so it "NOTs" whatever signal goes in to it). For the second bit of memory, you just wire the second bit to an [AND] gate, and wire your address line (assuming you're using a 1 bit memory addressing as opposed to modern 64 bit memory addressing) to that gate.

So when you're asking for address "0" the second bit has a "0" anded with something else, so <something> [AND] 0 is ALWAYS 0, but [NOT] 0 is 1, for for your first memory address you have the <memory> [AND] 1, which means your output will always be what's in memory. It's basically a multiplexer.

Also, an intro computer organization class is a bit different than programming. They're not usually too hard (at least mine wasnt), but they cover this kind of stuff well.

Last edited by mo' ron (2008-03-31 7:15 pm)

Fried Chicken wrote:

I can sort of follow that... but how do you get the computer to, for example, find something in a piece of memory, or allocate memory.
I just took programming this year, but I cannot imagine how some of the most basic commands in java can be executed in a computer on the basic level.
How do you assign a computer to display a piece of text for example?

It's all about abstraction.  I LOVE abstraction, and I love talking about it!!  It's a very simple yet profound concept that lets us do all the cool stuff we take for granted.

I'll start off by first noting that you're question doesn't really make sense (and then I'll use abstraction to explain why).  Allocating memory and displaying text are high-level constructs that depend most immediately on the kernel and relevant API's, respectively. In Java, it's actually more complicated -- you need to account for such things as the way objects are formed, dynamic dispatch, the bytecode interpreter -- lots of fun jargon.  But anyway, the CPU doesn't know what you mean by "allocating memory," for example.

The point is (this is where it gets cool) that for a particular Java command, you could (rougly) draw a tree-like structure denoting all the software and hardware constructs that the java command depends on.  Each level in the tree is a level of abstraction, and provides the interface its parent level depends on.  Rougly speaking, the lower levels of the tree consist mostly of hardware, while the higher levels consist of software interfaces.

For example, a piece of code in Java will depend on some other objects you might've written, which in turn depends on elements from the Java library, each of which depends on the JavaVM, each of which depend on various API's within the host OS, each of which depend on lower-level features of the OS, each of which depend on various machine instructions, each of which may depend on some microcode, which then depends on some hardware interface for said micro-ops.  And then you start getting into circuits.

The above example explains roughly how your Java command executes.  But a more general tree describing programming languages is a bit more enlightening.  We'll start with the CPU's instruction set.  The CPU contains operations for moving and comparing data (among other things).  With the appropriate sequence of bits, you can make the CPU do anything you want.  But the task of writing relatively simple programs in this way is difficult and quite time consuming -- writing a complex program like a modern video game is orders of magnitude beyond what mere mortals can accomplish.  To solve this problem, some smart people got the idea of using machine code to write an assembler.  This basically replaces cryptic sequences of machine code with helpful mnemonics and directives, and allows programmers to reference memory by "label" instead of having to worry about physical addresses.  The assembler made programming vastly simpler, while maintaining a 1:1 relationship with machine code.  Assembly code also introduced a crude concept of modularity -- writing somewhat more complex programs is easier when you can borrow previously-written code. 

Skipping some history, someone got the idea of using assembly to write even higher-level language compilers like C.  It's much faster and much easier to write a program in C compared to assembly.  This fact, combined with the improved modularity offered by C, made writing complex programs feasible.  As OS's began to mature, their API's became more sophisticated, further aiding the development of complex software.  In addition, the "virtual machine" paradigm meant that programmers could assume their programs had free-reign over a system's resources, without having to worry about conflicting with other programs (it also introduced threads).  At this level of abstraction, it's possible to write a modern 3D game (for example) using API's such as OpenGL.

However, progress did not stop here.  It was eventually discovered that programming could become further modularized with even higher-level languages via the object oriented paradigm.  C++, Java, C# and Objective-C are common examples of OOP languages.  The later three differ fundamentally from C++ in ways I won't bother to explain here, but the simple answer is they're even higher-level than C++.  Java is unique in that it compiles to bytecode rather than machine instructions (allowing one binary to run on any JavaVM machine); in this way, it is of higher abstraction than the other three.  Anyway, I hope I answered your question.  If not, hopefully this will answer someone else's!

As someone who has just started in CS this semester, I really appreciate that post, Mr. T. It was a good read.

Mr. T, that was a very good post.  I kinda understand what's happening.  We started by physically manipulating the computers with (was it punchcards?), and then later we had the computer manipulate itself through an assembler.  This relied on the original underlying framework of the computer.  From that, we built über-assemblers, which relied on the other assemblers.

I guess that kinda make sense.  I think I will take a class in college about this, to satisfy my curiosity.

Miles wrote:

mo' ron wrote:

This is how wozniak started out, and relays are a step up from the tubes the first univac and eniac were made out of.

Vacuum tubes are the step up from relays, I think you mean.

Probably true in Woz's time, but modern relays are probably more reliable overall than old school tubes.

But, yeah, tubes were the next step between relays and transistors.

I wonder if he had any bugs in his computer?

I mean, REAL bugs, the kind that get between the contacts of relays and that computer bugs are named after.