1. another mountain thing
- Posted by Michael Packard <lgp at EXO.COM>
Feb 16, 1998
-
Last edited Feb 17, 1998
Here's a 640x480 version of it. I changed the way I got polygon
coordinates, converting the whole matrix x,y,z to screenx, screeny before
doing the polygons, instead of doing poly verticies for each face, so I
only have to compute the points once instead of 4 times.
One thing that does perplex me is that if I create multiple landscapes
(press any key after each one is finished and it'll do another one) the
second landscape is way faster to create than the first one. On an 8th
level landscape the numbers are like this:
First Landscape Second Landscape (3rd,4th,etc...)
Landscape creation: 6.59 3.51
Convert time 10.66 3.62
3D draw time 22.96 15.96
It's even more dramatic on a 9th level, the first one takes 202 seconds
while the second is only 64 seconds. I'm doing exactly the same amount of
computations, on the same number of data points and drawing the same
number of polygons. Certainly there should be some variability in
compute\draw times, but not that much, in fact from landscape numbers 2-50
the total time varies by only a second. If anyone can give me a good
explanation why I'd appreciate it.
Also I'm computing intermediate times for each routine, but the total time
isn't the sum of the other times and I dunno why. Obviously I'm doing the
checks wrong somehow, but I can't find the problem. Someone find the
problem and tell me what I'm doing wrong.
Michael Packard
Lord Generic Productions
----------------------------------------
----------------------------------
-- Fractal Landscape Generator -
-- Version .76 2/16/98 -
-- Copyright 1998 -
-- Lord Generic Productions -
----------------------------------
without type_check
include get.e
include machine.e
include image.e
include graphics.e
include pic_load.e
integer G,D1,MX,MY,MZ
atom SR1,SR2,CR1,CR2,SR3,CR3,roll
atom seed,Arand,Nrand,GaussAdd,GaussFac
sequence X,pal,screen
integer maxlevel,addition,i,N,wire
integer D,d,bench
atom sigma,H,delta,max,min,time1,time2
----------------------------
-- vga palette for mountain
pal=
{63,63,63}}
-------------------------------------------
--initization of Gaussian random thingy
Nrand=4
Arand=1000
seed=1234
GaussAdd=sqrt(3*Nrand)
GaussFac=2*GaussAdd/(Nrand * Arand)
--set_rand(seed)
------------------------------------------
function Gauss()
atom sum
sum=0
for i = 1 to Nrand do
sum=sum+rand(Arand)
end for
return(GaussFac * sum - GaussAdd)
end function
-------------------------------------------
function f3(atom delta, atom x0, atom x1, atom x2)
atom y,z
z=Gauss()
y=(x0+x1+x2)/3+delta*Gauss()
return(y)
end function
--------------------------------------------
function f4(atom delta, atom x0, atom x1, atom x2, atom x3)
atom y,z
z=Gauss()
y=(x0+x1+x2+x3)/4+delta*Gauss()
return(y)
end function
---------------------------------------------
--initial parameters
sigma=1.25 -- Initial Standard Deviation Determines max height
H=.8 -- Fractal Dimention = 3-H Determines smoothness
addition=1 -- Add extra randomness to points?
---------------------------------------------
procedure mtn()
atom k
min=time()
N=power(2,maxlevel)+1
delta=sigma
X=repeat(0,N)
for o=1 to N do
X[o]=repeat(0,N)
end for
screen=X
X[1][1]=delta*Gauss()
X[1][N]=delta*Gauss()
X[N][1]=delta*Gauss()
X[N][N]=delta*Gauss()
D=N-1
d=D/2
for stage= 1 to maxlevel do
position(1,50)
puts(1,"Working on Stage #")?(stage)
-- going from grid type I to type II
delta = delta * power(.5,.5*H)
-- interpolate and offset points
for x=d+1 to N-d+1 by D do
for y=d+1 to N-d+1 by D do
X[x][y]=f4(delta,X[x+d][y+d],X[x+d][y-d],X[x-d][y+d],X[x-d][y-d])
end for
end for
--displace other points also if necessary
if addition then
for x=1 to N+1 by D do
for y=1 to N+1 by D do
X[x][y]=X[x][y] + delta * Gauss()
end for
end for
end if
-- going from grid type II to type I
delta= delta*power(.5,.5*H)
-- interpolate and offset boundary grid points
for x = d+1 to N-d+1 by D do
X[x][1]=f3(delta,X[x+d][1],X[x-d][1],X[x][d])
X[x][N]=f3(delta,X[x+d][N],X[x-d][N],X[x][N-d])
X[1][x]=f3(delta,X[1][x+d],X[1][x-d],X[d][x])
X[N][x]=f3(delta,X[N][x+d],X[N][x-d],X[N-d][x])
end for
--interpolate and offset interior grid points
for x = d+1 to N-d by D do
for y = D+1 to N-d by D do
X[x][y] = f4(delta, X[x][y+d], X[x][y-d], X[x+d][y], X[x-d][y])
end for
end for
for x = D+1 to N-d by D do
for y = d+1 to N-d+1 by D do
X[x][y] = f4(delta, X[x][y+d], X[x][y-d], X[x+d][y], X[x-d][y])
end for
end for
if addition then
for x=1 to N by D do
for y=1 to N by D do
X[x][y]=X[x][y] + delta * Gauss()
end for
end for
for x=d+1 to N-d+1 by D do
for y=d+1 to N-d+1 by D do
X[x][y]=X[x][y] + delta * Gauss()
end for
end for
end if
D=D/2
d=floor(d/2)
end for
max=time()-min
position(3,50)
puts(1,"Time: ") ?(max)
position(4,50)
puts(1,"Size: ") ?(length(X)*length(X))
min=255 max=0
for x= 1 to N do
for y= 1 to N do
k = X[x][y]
if k!=0 then k=((k/2)*175)+25 end if
if k<1 then k=1 end if
X[x][y]=floor(k)
if floor(k)<min then min=floor(k) end if
if floor(k)>max then max=floor(k) end if
end for end for
end procedure
-----------------------------------------
procedure drawit()
if maxlevel<9 then display_image({0,0},X) end if
end procedure
-------------------------------------------------
--3d draw stuff
procedure viewangle(atom pitch,atom yaw)--pitch and yaw angles in degrees
pitch=pitch*(.0174533) --convert to radians
yaw=yaw*(.0174533)
roll=0
roll=roll*(.0174533)
G=0 D1=-2000 MX=0 MY=-700 MZ=0
SR1=sin(yaw) CR1=cos(yaw)
SR2=sin(roll) CR2=cos(roll)
SR3=sin(pitch) CR3=cos(pitch)
end procedure
---------------------------------------------------
function perspective(atom x1,atom y1, atom z1) --gives screen coordinates
atom xa,ya,za,sx,sy
xa=CR1*x1-SR1*y1
ya=SR1*x1+CR1*y1
x1=CR2*xa+SR2*z1
za=CR2*z1-SR2*xa
y1=CR3*ya-SR3*za
z1=SR3*ya+CR3*za
x1=x1+MX
y1=y1+MY
z1=z1+MZ
sx=floor(D1*x1/y1)+650
sy=floor(D1*z1/y1)+530
bench=bench+1
return({sx,sy})
end function
------------------------------------------------
procedure convert_screen()
--creates and NxN matix of screen coordinates for the landscape
atom scale,vscale,ctime
scale=257/N
vscale=2
ctime=time()
for x=1 to N do
for y=1 to N do
end for
if floor(x/25)=x/25 then position(3,65)puts(1,"Converting:")?(x) end if
end for
position(5,50) puts(1,"Convert time: ") ?(time()-ctime)
end procedure
------------------------------------------------
procedure test3d()
--atom x1,y1,z1
atom scale,color,vscale,check
sequence points,points2
sequence point1,point2,point3,point4
points=repeat(0,N)
points2=repeat(0,N)
scale=257/N
check=time()
vscale=2
convert_screen()
for x=1 to N-1 do
for y=1 to N-1 do
point1=screen[x][y] --perspective(y*scale-129,x*scale-129,-X[x][y]/10)/vsca
le
point2=screen[x][y+1] --perspective((y+1)*scale-129,x*scale-129,-X[x][y+1]/10
)/vscale
point3=screen[x+1][y+1] --perspective((y+1)*scale-129,(x+1)*scale-129,-X[x+1][y
+1]/10)/vscale
point4=screen[x+1][y] --perspective((y)*scale-129,(x+1)*scale-129,-X[x+1][y]/
10)/vscale
if color>240 then color=240 end if
if color<0 then color=-color end if
if wire=0 then polygon(0,1,{point1,point2,point3,point4,point1}) end if
--if color>5 and color<150then pixel(color+rand(50),point1)
--elsif color>150 then pixel(color-rand(50),point1) end if
end for
end for
points=repeat(0,N+2)
points2=points
points2[N+2]=points[N+2]
for x=2 to N+1 do
points2[x]=perspective(N*scale-129,(x-1)*scale-129,-X[x-1][N]/8) /vscale
end for
polygon(0,1,points)
polygon(0,1,points2)
polygon(150,0,points)
polygon(150,0,points2)
--check=time()-check
position(6,50)
puts(1,"3D Draw Time: ") print(1,time()-check)
position (7,50) puts(1,"bench: ")?(bench)
end procedure
-------------------------------------------------
--main program
i=graphics_mode(256)
all_palette(pal)
for x= 1 to 50 do
bench=0
time1=time()
puts(1,"*******************************\n\r")
puts(1,"* Fractal Landscape Generator *\n\r")
puts(1,"* Version .76 2/16/98 *\n\r")
puts(1,"* Copyright 1998 *\n\r")
puts(1,"* Lord Generic Productions *\n\r")
puts(1,"*******************************\n\r")
wire=1 --0 for wireframe, 1 for filled polygons
maxlevel=8 --High detail level but slow. Set lower for lower detail
mtn() --Generates the fractal
drawit() --Draw the 2D map
viewangle(30,rand(90)) --Rotates the landscape for a good view. (pitch,yaw)
test3d() --Makes 3D coordinates and draws polygons
time2=time()
position(8,50)puts(1,"Total creation time: ")?(time()-time1)
position(23,1) puts(1,"Press any Key\n\rPress Q or ESC to quit")
i=wait_key()
if i=27 or i='q' then i= graphics_mode(-1) abort(0) end if
if i='m' then i=save_screen(0,"3dland.bmp") end if
clear_screen()
end for
i=graphics_mode(-1)
----------------------------------------------
2. Re: another mountain thing
>One thing that does perplex me is that if I create multiple landscapes
>(press any key after each one is finished and it'll do another one) the
>second landscape is way faster to create than the first one. On an 8th
>level landscape the numbers are like this:
Because the memory is already allocated to Euphoria.
Normally euphoria only allocates the memory you need, but it doesn't
unallocate when you no longer need it.
I heard Robert say (actually I saw somewhere in one of his messages) that
that memory stays allocated to Euphoria for further use by your program.
In which case using the same amount of memory would go faster, because you
don't have to allocate the memory again.
Maybe this can be optimized by Robert, the biggest disappointment of any
garbage collector (Euphoria is like those people in Disney land, they clean
up right behind ya!!) is the unstable speed. Maybe he could optimize/change
his preprocessor to prepare before we enter a loop, because within the loop,
the same actions should require the same time to accomplish.
Well, any optimization done, should be done per loop/level, it will optimize
WHEN something (might) happen(s) more than once, and it will make a program
run at a more stable speed, same action --> same computing time.
I can get really annoyed by this mail program (outlook-express) that waits
a second every half line.
Ralf
3. another mountain thing
Michael Packard:
At least one thing has made me wonder in your otherwise very good program=
=2E
Since you introduced the procedure viewangle, I saw these two lines of
code:
>roll=3D0
>roll=3Droll*(.0174533)
I don't understand. roll will always be zero.
Sincerely,
John Doe
writing at 0:05 , =
on Wednesday 18 February 1998
Using EMail Assist for WinCIM
4. Re: another mountain thing
On Wed, 18 Feb 1998, Ad Rienks wrote:
> Michael Packard:
>
> At least one thing has made me wonder in your otherwise very good program.
> Since you introduced the procedure viewangle, I saw these two lines of
> code:
>
> >roll=0
> >roll=roll*(.0174533)
>
> I don't understand. roll will always be zero.
>
yup, for this landscape thing roll should always be zero. I left the roll
calculation in just in case someone wanted the 3d rotation stuff and
needed to roll the object with respect to the viewer. In that case
they just remove the roll=0 and add an ",atom roll)" to the procedure
title and send it all three values when you call it.
Michael Packard
Lord Generic Productions
lgp at exo.com
