🚀 go-pugleaf

In article <11419@uqcspe.cs.uq.oz.au> warwick@cs.uq.oz.au writes:
>Sure, 2M would be lovely, but I'd hardly call 1Meg "insufficient", when
>it is equivalent to 1000 32x32 256 colour sprites, or:
>
>	2	320x240 256 colour screens. (150K)
>	200	32x32 256 colour sprites. (200K)
>	200K	Code (200K)
>	25	1-second 10KHz 8-bit stereo sampled sounds (500K)

A better choice may be the 320x240 True Color mode.  (Disclaimer:  all
of this depends on how fast the blitter chip is.)

While there are some advantages to the 8 bitplane/16 pixels per word
structure of the 256 color mode, the 1 pixel per word no CLUT mode could
actually be faster.  (Screen clearing would be slower, since twice the
amount of memory would be zeroed.  However, usually only a subset of
the screen would be cleared, and most of the game logic/sprite drawing/
etc will take longer anyway.)

Most drawing functions should be faster in True Color modes, despite
the fact that twice as much data must be handled.  Compare the work to
plot a point:  (using straight 68000)

         a0=screen
         a1=y offset list:  *int [0-239], in steps of 320 or 640
              depending on screen mode 
         d0=x
         d1=y
         d2=color
         
         True Color                       256 Color
         ----------                       ---------
     asl.w  #1,d1                  ; a lot more than 4 lines...
     adda.w 0(a1,d1.w),a0
     asl.w	#1,d0
     move.w d2,0(a0,d0.w)

Plus, the 68030 has addressing modes that should combine the four
instructions into two.  I didn't include the 256 color version, but
a 16 color version (ST Low, 4 planes) that I wrote some time ago is
40 lines long.

Lines and fills should be similarly fast, and sprite routines gain
the additional advantage of not needing preshifting.

So, a 1 meg game might look like:

     2   320x240 True Color screens (300K)
     100 32x32 True Color sprites (200K)
     1   Code (200K)
     20  1 second 8kHz mono samples (160K)

=860K, with remaining for OS, etc.

As I said at the beginning, if the blitter is really good then the 256 color
mode may be faster.  But, if programmers end up writing their own pixel
routines, the True Color mode will have to be considered over the 8 bitplane
version.
-- 
Rob McMullen               ewal235@hermes.chpc.utexas.edu
"In Gods We Trust.  All Others Pay Cash." -- Steven Brust