🚀 go-pugleaf

>Note that I didn't say there weren't fast bit
>blitters if you just look at some spec sheet about
>how fast the hardware can run in some cases
>the hardware does well.  The point is that
>THEY USUALLY DON'T DO ALL THE CASES!!
>Making up for the boundary conditions
>in software is often slower than just doing it
>with the CPU in the first place.  
>
>	-Mike

	No, but they can get pretty close.  Citing the Amiga blitter as an
example:  It can do any logic operation on it's 3 sources and 1 destination.
That's 256 logic operations!  (I'm not starting a war, just giving you an
example of a real-life blitter).

	Certainly doing extremely simple operations can be done relatively
fast by the 68000/68010 (about 3 times slower than a blitter running at the 
same speed), but when you get into more complex equations, for instance:

	((C&A)|D)->D  (Take bitmaps A through template C and OR onto the 
			 destination bitmap)

	The problem suddenly becomes non-trivial... especially when three
of the four bitmaps are on different bit boundries.   (in the example above,
D is used both as a source and as a destination).  Were talking about 10x
slower at a minimum.  Even a 68020 with it's cache would have a time of it
(perhaps only 3x slower).

	A good example is the LIFE program on the Amiga using the blitter and
the LIFE program on the atari using the processor.  Of course, you cannot do
a life-entity calculation in one blit... it takes the Amiga about 12 I believe.
So the Amiga is actually doing 20*12 = 240 complete blitter passes over the
screen per second.  Even though it takes the blitter 12 passes per life
entity, it is still 3 times faster than the Atari version written in extremely
optimized assembly.

	There can be no question that having one or more blitters gives a huge
advantage when manipulating displays.

					-Matt