Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Fractal
#1
Just for fun.

Incidentally, if I change the float types to fixed, it crashes. Has something gone wrong with fixed type maths?

EDIT: Updated with fixed and exit for, so if people use it here, they get the best version.

Code:
#define width 256
#define height 192

DIM x,y as FIXED
DIM xstart,xstep,ystart,ystep as FIXED
DIM xend,yend as FIXED
DIM z,zi,newz,newzi as FIXED
DIM colour as byte
DIM iter as uInteger
DIM col as uInteger
DIM i,k as uByte
DIM j as uInteger
dim inset as uByte



xstart=-1.6
xend=0.65
ystart=-1.15
yend=-ystart
iter=24

xstep=(xend-xstart)/width
ystep=(yend-ystart)/height

'Main loop
x=xstart
y=ystart

border 0
paper 0
ink 7
CLS

for i=0 to ( height -1 )/2 +1
        for j=0 to width -1
            z=0
            zi=0
            inset=1
                for k=0 to iter
                    ';z^2=(a+bi)*(a+bi) = a^2+2abi-b^2
                    newz=(z*z)-(zi*zi)+x
                    newzi=2*z*zi+y
                    z=newz
                    zi=newzi
                    
                    if (z*z)+(zi*zi) > 4 then
                        inset=0
                        colour=k
                        exit for
                    END IF
                next k
                

                if NOT inset then
                    if colour BAND 1 THEN
                        plot j,i
                        plot j,192-i
                    END IF
                end if
                    
                x=x+xstep
         next j
                
        y=y+ystep
        x=xstart
print at 23,0;CAST(uinteger,i)*200/height;"%"
next i
                            
BEEP 1,1
PAUSE 0

(This will take a long time. Even if you push an emulator to top speed)
Reply
#2
Awesome!!! Confusedhock:
I'll put this in the Wiki, if you don't mind, in the Examples section. :?:
Reply
#3
Course you can.

(Though again - why does it seem to break with fixed, not float?)
Reply
#4
Oh - I also thought I should be able to use "end for" in the place I have "goto screen". Not sure why it refuses to compile that.
Reply
#5
It might be a bug, or might be a matter of precision. :?:
Don't know. I know there's an integer version of Mandelbrot, BTW.
Reply
#6
britlion Wrote:Oh - I also thought I should be able to use "end for" in the place I have "goto screen". Not sure why it refuses to compile that.
The correct syntax for that is "EXIT FOR" (this is like a Break sentence in C/C++)
Reply
#7
quite nice
[Image: ZljXg.png]
Reply
#8
boriel Wrote:
britlion Wrote:Oh - I also thought I should be able to use "end for" in the place I have "goto screen". Not sure why it refuses to compile that.
The correct syntax for that is "EXIT FOR" (this is like a Break sentence in C/C++)


Doh! Me idiot.

Yes, that works. Much better than a goto.
Reply
#9
boriel Wrote:It might be a bug, or might be a matter of precision. :?:
Don't know. I know there's an integer version of Mandelbrot, BTW.


I don't think a lack of precision should cause a restart, however...

And yes. I didn't do that one Smile integer would be faster.
Reply
#10
slenkar Wrote:quite nice

Oh slenkar! Now you've given it away Smile
Reply
#11
britlion Wrote:
boriel Wrote:It might be a bug, or might be a matter of precision. :?:
Don't know. I know there's an integer version of Mandelbrot, BTW.


I don't think a lack of precision should cause a restart, however...

And yes. I didn't do that one Smile integer would be faster.
A restart???
Please send me the code. I would like to test it, please :?:
Reply
#12
Same code, with "FIXED" wherever DIM had "FLOAT"

I could believe in it going out of bounds or something, but shouldn't just crash...

Code:
#define width 256
#define height 192

DIM x,y as FIXED
DIM xstart,xstep,ystart,ystep as FIXED
DIM xend,yend as FIXED
DIM z,zi,newz,newzi as FIXED
DIM colour as byte
DIM iter as uInteger
DIM col as uInteger
DIM i,k as uByte
DIM j as uInteger
dim inset as uByte



xstart=-1.6
xend=0.65
ystart=-1.15
yend=-ystart
iter=24

xstep=(xend-xstart)/width
ystep=(yend-ystart)/height

'Main loop
x=xstart
y=ystart

border 0
paper 0
ink 7
CLS

for i=0 to ( height -1 )/2 +1
        for j=0 to width -1
            z=0
            zi=0
            inset=1
                for k=0 to iter
                    ';z^2=(a+bi)*(a+bi) = a^2+2abi-b^2
                    newz=(z*z)-(zi*zi)+x
                    newzi=2*z*zi+y
                    z=newz
                    zi=newzi
                    
                    if (z*z)+(zi*zi) > 4 then
                        inset=0
                        colour=k
                        exit for
                    END IF
                next k
                
screen:                
                if NOT inset then
                    if colour BAND 1 THEN
                        plot j,i
                        plot j,192-i
                    END IF
                end if
                    
                x=x+xstep
         next j
                
        y=y+ystep
        x=xstart
print at 23,0;CAST(uinteger,i)*200/height;"%"
next i
                            
BEEP 1,1
PAUSE 1
PAUSE 0
Reply
#13
It was a bug in Fixed multiplication.
Please, download release 1.2.9s789 or newer, and tell mi if it works (it did for me!).
I don't see much improvement. :?: :?: :?:
Can you measure the time it takes to draw the fractal with fixed variables, please? There should be some performance gain.

NOTE: This release includes your MLDepacker routine (fastcalled).

NOTE2: Happy Birthday! :!: :mrgreen:
Reply
#14
boriel Wrote:It was a bug in Fixed multiplication.
Please, download release 1.2.9s789 or newer, and tell mi if it works (it did for me!).
I don't see much improvement. :?: :?: :?:
Can you measure the time it takes to draw the fractal with fixed variables, please? There should be some performance gain.
I added a timer function to it, and grabbed the timer before and afterwards.

Float:
s777 : 5237.6 Seconds.
s789 : 5237.9 Seconds

Fixed:
S789 : 2159.76 Seconds

I think that's a fair speed improvement, actually....

boriel Wrote:NOTE: This release includes your MLDepacker routine (fastcalled).

Nice! I added that to the library - so there's a link to the packer program in there too. You might want to tweak that to note that it's in the package.
boriel Wrote:NOTE2: Happy Birthday! :!: :mrgreen:

Thankee Kindly. I'm /really/ getting old now!
Reply
#15
Should be able to get it faster still using the integer method - which cheats and multiplies up decimals:

DIM xl,yl as long
DIM xstartl,xstepl,ystartl,ystepl as long
DIM xendl,yendl as long
DIM zl,zil,newzl,newzil as long

#DEFINE MULTIPLIER 10000

xstartl=xstart * MULTIPLIER
xendl=xend * MULTIPLIER
ystartl=ystart * MULTIPLIER
yendl=yend * MULTIPLIER
xl = x * MULTIPLIER
yl = y * MULTIPLIER


Though I tried this to abject failure. Anyone else get anywhere?
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)