Findpixels... one of the prettiest blobs of code. They convert an x and y position to a pointer and bitmask of the pixel on screen (with screen width of 128 pixels and screen at $FC00, i.e. TI-85 and 86). And why try to make your own when it'll probably not even be half as good as one of these. Choose your flavour and enjoy.
| time | size | registers | destroyes | notes | |
|---|---|---|---|---|---|
| ÜBERoutine | 203 726 | 24 23 | (d,e) to hl:a | b | smallest; b=0 |
| CLEM's #1 | 131 | 28 | (b,c) to hl:a | - | perfect balance |
| CLEM's #2 | 127 | 35 | (b,c) to hl:a | bc | - |
| ZAPO | 105 | 40 | (c,b) to hl:a | de | needs mem at $8700 |
| Eble-Yopp-Yopp.. | 122 | 34+ | (e,d) to hl:a | c | code alignment |
| Edited Eble-Yopp | 127 | 35 | (b,c) to | de | hl minus $FC00 |
| Yopp's | 1391 | 34 | (b,c) to hl:a | bde | don't use! |
Right. Let's start with this one... Any my first findpixel should be better than this. What does he do? He multiplies by adding 16 y times. One of the worst ideas thinkable (at least add y 16 times, but don't do that neither.) Close your eyes and scroll down.
;---------------------------------------------------------------------
; Routine by James Yopp
;
; 134-2647 cycles - 34 bytes
; (b,c) to hl:a - destroyes: bde
;---------------------------------------------------------------------
FindPixel: ; ©1997 James Yopp. If you use it, leave this comment in the source
push bc
ld h,$FC
srl b
srl b
srl b
ld l,b
ld b,c
ld de,16
xor a
BLoop:
add hl,de
cp b
jr z,XitLoop
dec b
jr BLoop
XitLoop:
pop bc
ld a,b
and %00000111
ld b,a
inc b
ld a,%00000001
ALoop:
rrca
djnz ALoop
ret
Onwards, to another Yopp routine, but the addition of a few Ebles makes it a lot better! As far as I know, this one's the most commonly used. Probably because it looks so cool with the .org displacement-thingy. If only users knew what it means. For it means you have to place it at exactly right in your program, or it'll grow in size. Not very handy. It is fast though. Just the zapo routine is faster, but this one is easier. Conclusion: for the fastest "easy" routine take this one. Just check the placement.
;---------------------------------------------------------------------
; Eble-Yopp-Yopp-Eble-Eble-Eble-Yopp-Eble Faster FindPixel Routine
;
; 122 cycles - 34-41 bytes
; (e,d) to hl:a - destroyes: c
;
; 118 cycles - 33-40 bytes
; (e,d) to hl:c - destroyes: a
;---------------------------------------------------------------------
FindPixel:
ld hl,FP_Bits
ld a,e
and $07 ; bit offset
or l ; add bit offset to HL
ld l,a
ld c,(hl) ; c=bitmask
ld h,$3F
ld a,d
add a,a
add a,a
ld l,a ; l=y*4
ld a,e
rra
add hl,hl
rra
add hl,hl ; hl=$fc00+y*16
rra ; a=x/8
or l ; add x
ld l,a ; video-byte
ld a,c ; bitmask (can be removed)
ret
.org (($+8) & 0FFF8h) ; align FP_Bits on the next 8-byte boundary
FP_Bits:
.db $80,$40,$20,$10,$08,$04,$02,$01
Edited Eble-Yopp? Mainly rebuilt i'd say. A very nice routine because of a
great idea. It uses the mainly unknown instuction rld. Neat. Too bad
it somehow didn't work out though :( Unlike what the author may believe, it's
actually slower and larger than the original (not much, but still.) Wouldn't
use it, but it sure works cool ;)
Warning: hl will not point to the screen,
but rather the screen offset! You'll have to add $FC00 to use it like any other
findpixel.
;---------------------------------------------------------------------
; An edited EBLE-YOPP Routine
;
; 127 cycles - 35 bytes
; (b,c) to hl:a - destroyes: de
;---------------------------------------------------------------------
Find_Pixel:
ld de,FP_Bits
ld a,b
and %00000111
add a,e
ld e,a
adc a,d
sub e
ld d,a ; (de)=bitmask
ld hl,FP_RLD
ld (hl),c ; FP_RLD=y
ld a,b ; a=x
rra
rra
rra ; x/8
rld ; a(hl)=16y+x/8
and %00000011
ld l,(hl)
ld h,a
ld a,(de)
ret
FP_Bits:
.db 128,64,32,16,8,4,2,1
FP_RLD:
.db 0 ; temp byte needed but doesnt have to be in program
ZAPO's long gone now, but they left us this great findpixel. Guess what? It's the fastest one yet! Only 105 cycles. Furthermore, the routine itself is also the smallest one I know of. One problem: it requires you to run some initiation code before use. This makes your program 19 bytes larger, and it'll also use a little memory at $8700. If you need a really fast findpixel, this one's the best. Otherwise, better not.
;---------------------------------------------------------------------
; The ZAPO even faster findpixel
; by ~Code7~
;
; 105 cycles - 21 bytes
; (c,b) to hl:a - destroyes: de
;
; 101 cycles - 20 bytes
; (c,a) to hl:a - destroyes: de
;---------------------------------------------------------------------
FindPixel:
ld a,b
FindPixelShort:
ld h,$3f
add a,a
add a,a
ld l,a ; hl=$3f00+4*y
ld a,c
rra
add hl,hl
rra
add hl,hl ; hl=$fc00+16*y
rra ; 8*x
add a,l
ld l,a
ld d,$87
ld a,c
and d ; =7
ld e,a
ld a,(de)
ret
; the offset table program must be placed at $8700
;
; 163 cycles 19 bytes destroyes: bcdehl
InitializeFP:
ld de,$8700
ld hl,Offset
ld bc,8
ldir
ret
Offset:
.db $80,$40,$20,$10,$8,$4,$2,$1
Nothing too special about this one. Good optimizing, nice routine, but there are better ones outthere.
;---------------------------------------------------------------------
; CLEM's FINDPIXEL
; by Clem
;
; 127 cycles - 35 bytes
; (b,c) to hl:a - destroyes: bc
;---------------------------------------------------------------------
FindPixel:
ld h,63
ld a,c
add a,a
add a,a
ld l,a ; hl=$3f00+4*y
ld a,b
rra
add hl,hl
rra
add hl,hl ; hl=$fc00+16*y
rra ; a=x/8
or l
ld l,a
ld a,b
and 7
ld bc,FP_Bits
add a,c
ld c,a
adc a,b
sub c
ld b,a
ld a,(bc)
ret
FP_Bits:
.db $80,$40,$20,$10,$08,$04,$02,$01
Same author, different routine. This one is special. With 28 bytes one of the smallest (only snake86 is smaller), and still quite fast (7% slower than Eble-Yopp.) And, bonus, no registers are destroyed!! In a lot of cases this one would be ideal.
;---------------------------------------------------------------------
; CLEM's FIND_PIXEL
; by Clem
;
; 131 cycles - 28 bytes
; (b,c) to hl:a - destroyes: none
;---------------------------------------------------------------------
FindPixel:
ld h,63
ld a,c
add a,a
add a,a
ld l,a ; hl=$3f00+4*y
ld a,b
rra
add hl,hl
rra
add hl,hl ; hl=$fc00+16*y
rra ; a=x/8
or l
ld l,a
ld a,b
and 7
cpl
rlca
rlca
rlca
ld (FP_Bit),a
xor a
FP_Bit =$+1
set 0,a
ret
And last but not least, the smallest findpixel ever! Look at it: only
24 bytes! And with an average of 158 cycles still quite usable.
For most progs which don't need that speed anyhow, this one's your choice. And
afterwards, b will be 0, so you can use that too to shave off another
byte or so (think
I just called it the über routine because it isn't really released by anyone. I took it out of Snake86 (unreleased project by David Phillips, Jonah Cohen and Ben Mickle). The byte-part seems to come straight out of the Eble-Yopp-Yopp routine, while we've seen the bit-part before in the routine by just Yopp (so that wasn't all bad afterall ;). Those combined give a kickass piece 'a code of just 25 bytes. When I showed it to my friend Free Bird he thought of a way to shave off another byte (I won't let him have all the credits though. It's simply a routine by and for everybody :))
Also see the comments for how to shave of yet another byte, courtesy of Xavier van Rijnsoever. It does make it lots slower, but in most cases it'll still do fine (still only 115µs on average).
So voici! Beat this:
;---------------------------------------------------------------------
; Ueber-FindPixel (Snake86 + Free Bird)
; by Friends of the 86
;
; 144-263 cycles - 24 bytes
; (d,e) to hl:a - destroyes: b
; 140-4475 cycles (726 avg) - 23 bytes
; (d,e) to hl:a - destroyes: b
;---------------------------------------------------------------------
FindPixel:
ld l,e
add hl,hl
add hl,hl
ld h,$fc/4 ; hl=$3f00+4*y
ld a,d
and 7
inc a ; remove this and replace the line below for small version
ld b,a
ld a,d
rra
add hl,hl
rra
add hl,hl ; hl=$fc00+16*y
rra ; a=x/8
or l
ld l,a
ld a,1 ; change to ld a,128 for small version
PixelLoop:
rrca
djnz PixelLoop
ret
| /calc/z80/findpixel · en • created in 0.043s | ©1999-2003 by SHIAR • All rights reserved |