Nice fractals

[ErosOlmi]

http://content.techrepublic.com.com/....e099.dl071807

[Michael Clease]

Try this

[kryton9]

Thanks nice link and program. Fractals are amazing and always fun to look at!

[ErosOlmi]

Thanks Abraxas.
Those kind of scripts shows the limit in interpreted languages.
I will use it to study more tricks to add more execution speed.

[Petr Schreiber]

Nice fractals!,

Mandelbrot is good, I tried to optimize Abraxas code a bit, but I got only 20% increase
[code=thinbasic]
SUB DoBRot()
local t1, t2 as long
local wminone as long = width - 1
local hminone as long = height - 1

TBGL_BeginPoly %GL_POINTS
xmin = -2.0
ymin = -1.3
xmax = 1.0
ymax = 1.3
integralX = (xmax - xmin) / width
integralY = (ymax - ymin) / height
x = xmin
For S = 1 To wminone
y = ymin
For Z = 1 To hminone
x1 = 0
y1 = 0
looper = 0
While (looper < 1000 And Sqr( x1*x1 + y1*y1 ) < 2)
Looper += 10
xx = x1*x1 - y1*y1 + x
y1 = 2*x1*y1 + y
x1 = xx
WEND
TBGL_Color ColorMap_R(Looper), ColorMap_G(looper), ColorMap_B(looper)
TBGL_Vertex s,z
y += integralY
Next
x += integralX
Next
TBGL_EndPOLY
END SUB
[/code]

I just put tbgl_BeginPoly/tbgl_EndPoly out of the loops as it is not necessary for each point, added += on two places, removed () where not needed. I am sure it could be speeded up even more somehow, but I am don't how to do it


Bye,
Petr

[Petr Schreiber]

Ok,

here is version which should run 47% faster - using FOR/NEXT with step !
[code=thinbasic]
SUB DoBRot()
local t1, t2 as long
local wminone as long = width - 1
local hminone as long = height - 1

TBGL_BeginPoly %GL_POINTS
xmin = -2.0
ymin = -1.3
xmax = 1.0
ymax = 1.3
integralX = (xmax - xmin) / width
integralY = (ymax - ymin) / height
x = xmin
For S = 1 To wminone
y = ymin
For Z = 1 To hminone
x1 = 0
y1 = 0
for looper = 0 to 1000 step 10
if Sqr( x1*x1 + y1*y1 ) >= 2 then exit for
xx = x1*x1 - y1*y1 + x
y1 = 2*x1*y1 + y
x1 = xx
next
TBGL_Color ColorMap_R(Looper), ColorMap_G(looper), ColorMap_B(looper)
TBGL_Vertex s,z
y += integralY
Next
x += integralX
Next
TBGL_EndPOLY
END SUB
[/code]


Bye,
Petr

[Michael Clease]

thanks Petr.

Ive added some different sub routines to give you timing examples.

DoBrot1() '= 14.5 seconds
DoBrot2() '= 14.2 seconds
DoBrot3() '= 11.2 seconds
DoBrot4() '= 14.0 seconds

how does it compare on your machine?

[ErosOlmi]

Similar time here:
1: 12,6
2: 14,2
3: 10,3
4: 12,7

A great example to be used to make optimization.

Thanks a lot
Eros

[Petr Schreiber]

Hi,

grr, worst results
CPU is AMD Sempron 64 3400+ ( 1.8GHz real freq )

• 18.68
• 18.03
• 14.77
• 18.33

Really nice benchmark!

Bye,
Petr

[ErosOlmi]

The biggest time is loosed in
[code=thinbasic]
While (lo < 1000 And Sqr( x1 * x1) + (y1 * y1) < 2)
Lo += 10 ' the loop counter
xx = x1 * x1 - y1 * y1 + x
y1 = 2 * x1 * y1 + y
x1 = xx
incr counter
WEND
[/code]
Commentig out the full WHILE/WEND it takes just half second.

I've added a counter to be able to check how many times it is executed that while.
In DoBRot1 WHILE/WEND is executed 3.729.844 times. A lot for an interpreter but I'm sure I will get some more optimization.