THE COPROCESSOR
by Angela Levitt
     
	The math chip. The floating-point unit. The 
coprocessor. Why, darned if those names don't roll so easily 
off our tongue that, if the innocent bystander didn't know 
better, it might be assumed that we really knew what the 
thing does.
	Well, to put it not very briefly, it's like 
this: in computers built in days past, there was the central 
processing unit, or CPU. Maybe it was an Intel 80286 or 386, 
or possibly a Motorola 68020 or 030. In fact, it could have 
been any of several processors; my point is that it was 
nothing more than an integer CPU (for those who've been out 
of school for a while, an integer is a whole number, no 
fractions or decimals allowed), actually able to handle only 
integer operations, though clever programming could allow it 
to also work on the right side of the decimal point.

Isn't that cheating? 
	If that sounds contradictory, try this example. Just 
imagine an average man who can barely lift his own weight on 
a good day after taking his vitamins. Equip him with a 
properly-designed pulley system, though, and run a rope over 
those pulleys. Presto! He can now lift twice his weight. 
Along with all this newfound strength, you'll see another 
change: the weight will rise at half the speed our man is 
applying to his end of the rope.
	With special programming acting as the pulley system, 
the processor did the entire calculating job on your computer
containing an integer CPU, with the result that some 
calculations were slow. Other calculations (possibly 
equivalent to our guy lifting ten or a hundred times his 
weight) were so extremely slow that it was impractical to 
use a computer that didn't have a special chip called the 
floating point unit, designed to work on the right side of 
the decimal point, for those operations. In fact, there are 
some software applications that check for floating-point 
capability when they're started, and will not run at all if 
it isn't there.
	In computers with the CPUs mentioned above, if the 
purchaser paid the extra price to get a floating-point unit, 
it was in the form of a separate chip. Your current-production 
central processing unit, be it an 80486, a 68040, a Pentium, 
or a PowerPC, contains both an integer function and a floating 
point function, though the two are completely isolated from 
each other inside the chip. If your computer doesn't have the 
number-crunching capability that you would expect to accompany 
that fact, it's because the floating-point function has been 
disabled in your chip, not because it isn't there. The 
manufacturers tell you this, but they do it in such a sneaky 
way that you may not have noticed. If your Intel chip has an 
"SX" after the number, or if your Motorola chip contains the 
letters "LC," that's your clue. If there's a "DX" or no 
letters at all, the FPU still works.
	Why do they put something in the chip, then disable 
it? Brace yourself: it's so they can make more money. Nothing 
illegal, immoral or fattening here; it's just more efficient 
to make lots of one chip than smaller numbers of two chips. 

And the payoff? 
	Okay, all of that was just so we'd see that there are 
two types of operations going on where numbers on both sides 
of the decimal are concerned, and two processors to handle 
those operations. When numbers must be crunched, the CPU 
does the integers (again, whole numbers only) and the FPU 
takes the rest. With the two chips working together, the 
result is usually a dramatic improvement in computer speed 
for complex tasks.
			Copyright 1995 SeeJay Publications