Architectural breakdown: Binary

... and it all begins with binary. Binary is a 'base 2' numbering system which means it uses the numbers 0 and 1. In contrast humans tend to use decimal (base 10) for their day-to-day numerical needs which, of course, uses the numbers 0 to 9. Binary is particularly suitable for electronics as the numbers 0 and 1 can be represented by a circuit being on (1) or off (0), or more relevantly to this computer, a wire may be carrying power (1) or not (0). If you wanted to be able to carry a single decimal digit (0-9) on a wire you'd need to pull some fancy tricks - perhaps have a different level of power signifying the different digits or send pulses counting up the number. All this gets very complicated though - binary keeps it nice and simple.

So, given a single wire we can represent the numbers 0 and 1 - which isn't much use if you want a computer to hold, say, the number 107. So how do we store larger numbers? Well, we do it much the same way we do in our base 10 decimal system. Think about what the number 107 actually represents:

100s | 10s | 1s
-----------------
 1  |  0  | 7

... 107 is one 100 plus zero 10s plus seven 1s. Binary works exactly the same way just rather than the columns being powers of 10 (because it's a base 10 system) it's powers of 2:

64s | 32s | 16s | 8s | 4s | 2s | 1s
-------------------------------------
 1  |  1  |  0  |  1 |  0 |  1 |  1

... 1x64 + 1x32  + 0x16 + 1x8 + 0x4 + 1x2 + 1x1 = 107. If we changed the value under the 16s column from a 0 to a 1 we'd then get a total of 123. Each column represents a single binary digit (or 'bit' for short). Easy :)

So, what do we mean when we say a computer is 8-bit or 16-bit (or 64-bit for a more modern computer)? Well, taking the example above again, we can see that if we added one more column to the left (128s) we'd now have 8 binary digits — that's an 8-bit number. Double the number of digits and we would have a 16-bit number. It doesn't matter if we're actively using all of the bits or not — you could represent 107 using 8-bit or 16-bit you just set all the unneeded values to 0 (1000 = 00001000 = 0000000000001000). What the 'bittyness' of a number does impact is the largest number you can represent — so a 4-bit number can represent a maximum of 15 (8 + 4 + 2 + 1), an 8-bit number can represent up to 255 (128 + 64 + 32 + 16 + 8 + 4 + 2 + 1), a 16-bit number goes up to 65535 (32768 + 16384 + ... + 4 + 2 + 1). Notice how doubling the number of bits each time exponentially increases the maximum number you can represent. So, those modern 64-bit computers ... they can hold a number up to (2^64)-1 or, are you ready for this, 18,446,744,070,000,000,000 - that's eighteen quintillion, four hundred and forty-six quadrillion, seventy trillion. 

So far we've only looked at storing positive numbers which is quite straightforward. Things get a little more complicated when storing negative numbers. In decimal when we want to signify a number is negative we stick a negative sign in front of it. Computers perform much the same trick but they store the sign in the number itself so in an 8-bit number the highest bit (the 128s column) is reserved for signifying whether the number is positive (0) or negative (1). This does however mean we have less room for holding the number itself so the highest number we can hold is halved (because a 7-bit number could hold a maximum of 127) although as we'll see the total unique numbers we can store stays much the same. There's an extra complexity added to all this which is called two's complement: a negative number is represented by turning all the 1s to 0s and 0s to 1s in the number and then adding one to the result. The reason for this is that by storing negative numbers in this fashion the computer can add negative and positive numbers together without any further manipulation and the calculation just works. Here's an example of simple binary addition first:

          S | 4s | 2s | 1s
         -----------------
  3       0 |  0 |  1 |  1
+ 3       0 |  0 |  1 |  1
= 6       0 |  1 |  1 |  0

Hold on, how did that all happen. Well, in the same way addition works in decimal ... we add the smallest numbers first and then carry. So starting in the 1s column: 1 + 1 (in binary) = 10, we keep the 0 and carry the 1. Next, in the 2s column: 1 + 1 = 10 plus the carried 1 = 11, we keep the 1 and carry the 1. Next, in the 4s column: 0 + 0 = 0 plus the carried 1 = 1 and we're all done. Now, to get a negative number we make the two's complement:

            S | 4s | 2s | 1s
           -----------------
 3          0 |  0 |  1 |  1
NOT 3       1 |  1 |  0 |  0
NOT 3 + 1   1 |  1 |  0 |  1

So, we perform a NOT operation on each digit (0 -> 1, 1 -> 0) and then increment by 1. Minus three is therefore represented in binary as 1101 (remembering that the left most bit is now representing if the number is positive or negative). So, if we want to subtract 3 from 6 (using 6 + -3):

          S | 4s | 2s | 1s
         -----------------
  6       0 |  1 |  1 |  0
+ -3      1 |  1 |  0 |  1
= 3       0 |  0 |  1 |  1

As before we add the smallest numbers first and then carry. So, starting in the 1s column: 0 + 1 = 1. Next, in the 2s column: 1 + 0 = 1. Next, in the 4s column: 1 + 1 = 10, we keep the 0 and carry the 1. Finally in the last column: 0 + 1 = 1 plus the carried 1 = 10, we keep the 0 and carry the 1 (in this case the 1 is thrown away as there's nowhere to carry it to). There we are, back at three again. Another example is adding 1 to -1 which brings us back to 0:

        S | 4s | 2s | 1s
        -----------------
 -1     1 |  1 |  1 |  1
+ 1     0 |  0 |  0 |  1
= 0     0 |  0 |  0 |  0

Notice that effectively the values roll on from negative to positive when you add one as you'd expect. Something interesting happens though when you take your highest possible positive number and add one:

        S | 4s | 2s | 1s
        -----------------
  7     0 |  1 |  1 |  1
+ 1     0 |  0 |  0 |  1
= -8    1 |  0 |  0 |  0

Notice that if we weren't treating our result as being possibly negative this would be ordinary addition (7+1 = 8) but because our 'sign bit' means we have a negative number we've ended up with -8. One way to imagine this is that the numbers -8 through 0 and through to 7 are on the edge of a wheel — when you tip past the 7 you come back to -8. Effectively you could keep adding one to the number and you'd go around the wheel for ever and ever and ever. The interesting thing to note here though is that none of this makes any difference to the part of the computer doing the addition. It doesn't care whether you interpret the result as a negative or positive number it can do the same operation regardless and it's this that makes two's complement so useful. When we program the computer we can choose which interpretation to use depending on if we want to store negative numbers or only positive numbers. The trade off is that if we're using negative numbers we loose one bit for the sign so our number range changes accordingly. For an unsigned 8-bit number we can store anything from 0 through to 255. For a signed 8-bit number we can store anything from -128 through to 127.

OK, so that's the basics of binary done with. Next time I'll have a crack at describing what busses and registers are all about.

Comments

Popular posts from this blog

ALU Design: 1-bit Arithmetic Unit

ALU Design: 8-bit Arithmetic Unit

Decoder Design: MOV-8, ALU & SETAB