From:             "Henrik Magnusson (RayLight /PowerLine)" <raylight@ramses.ml.org>
Send reply to:    "Henrik Magnusson (RayLight /PowerLine)" <raylight@ramses.ml.org>
To:               azure@hamburg.roses.de
Date sent:        Wed, 08 Oct 1997 23:40:48 +0100
Organization:     PowerLine
Subject:          Contribution to Circle-Compo (Fastest)

Hi again Azure!

Here's my contribution to the (Fastest) Circle-Drawer Compo.. :)

See you! :)




* Contribution for the Circle Drawer Compo (Fastest Routine)
* ***********************************************************
* Coded by Raylight of PowerLine (Henrik Magnusson)
* E-Mail: raylight@ramses.ml.org
*
* Routine Speed (On 030-28MHz)
* ----------------------------
* Min: 3700 circles/s
* Avg: 7000 circles/s
* Max: Many circles/s
*
* Applies to following rules:
* ---------------------------
*
* *No hardware hacking.
* *No system calls etc.
* *No assumptions of register state are made, except for inputregisters.
* *Input: d0.l/d1.l - x/y-center
*         d2.l      - radius
*         d3.b      - colour
* *Size of Buffer: 256x256 (256 Colours,8bit-Chunky).
* *Register a7 not trashed
* *Circle is nearly perfect circular.
* *Should work on 020->
* *Circle is 100% closed.
*
* Additional considerations
* -------------------------
*
* *Much memory is used.
* *Since I don't know if CacheClearU() is allowed, I've skipped it. It
*  really SHOULD be done after the init routine, since it uses SMC technique.
*
*
* Theory Behind The Routine
* -------------------------
*
* This routine uses a Bresenham-like routine for generating the circle
* drawers. On code is generated for each radius. Hence, the circle routine
* is able to draw circles with as little as 1 instruction per pixel:
*
*    move.b   d3,bufferoffset(a0)
*
* (Note that this routine could be further optimized by using word and
*  longword writes, where possible. Also, it could utilize the (a0)+ syntax
*  where possible. I don't bother about that now since it's really fast
*  anyway.. :) ..)
*

*********************************************************************
*
* Here you put all your code.
*
*********************************************************************

sectionyourcode,code

*********************************************************************
*
* Your init code comes here.
*
*********************************************************************

_initcode
a
leaCircleCode,a0
move.l a0,a1

move.l#0,d1
.circleloop
move.la0,a1

move.ld1,d2
subq.l#1,d2
bgt.ok
moveq#1,d2
.ok

moveq#0,d4;x
move.ld2,d5
add.ld5,d5;dy
moveq#0,d6;dx
;(d2 = y)

moveq#0,d7;q
.loop
tst.ld7
bge.skip1
addq.l#1,d4
add.ld6,d7
addq.l#2,d6
.skip1
tst.ld7
blt.skip2
subq.l#1,d2
sub.ld5,d7
subq.l#2,d5
.skip2

move.ld2,d0
lsl.l#8,d0
add.ld4,d0

move.w#$1143,(a1)+      ;OPCODE: move.b  d3,????(a0)
move.wd0,(a1)+

add.l#256-2,d0
neg.ld0
move.w#$1143,(a1)+      ;OPCODE: move.b  d3,????(a0)
move.wd0,(a1)+

add.ld4,d0
add.ld4,d0
subq.l#2,d0
move.w#$1143,(a1)+      ;OPCODE: move.b  d3,????(a0)
move.wd0,(a1)+

neg.ld0
sub.l#256,d0
addq.l#2,d0
move.w#$1143,(a1)+      ;OPCODE: move.b  d3,????(a0)
move.wd0,(a1)+

tst.ld2
bne.b.loop

move.w#$4e75,(a1)+

add.l#1024*4,a0

addq.l#1,d1
cmp.l#129,d1
blt.circleloop

b
rts


*********************************************************************
*
* Your main routine.
*
* Every time you are given: A0   : Chunky Screen (256x256x8)
*                           D0.l : Center (X)
*                           D1.l : Center (Y)
*                           D2.l : Radius (1-127)
*                           D3.b : Colour
*
*********************************************************************

_circle
c
lsl.l#8,d1
add.ld0,d1
add.ld1,a0

LeaCircleCode,a1
lsl.l#8,d2
lsl.l#4,d2
add.ld2,a1
jsr(a1)

d
rts

printt"Length of your code:"
printv((b-a)+(d-c))


**********************************************************************
*
* Put your tables here
*
**********************************************************************
section yourbss,bss

CircleCode:ds.l1024*130

.