uniform random point inside circle

12-11-2007, 10:08 AM

i have 2 methods to do this tat i found on the net. which one is faster?

The first one is said to be fastest on computer:
Get random point in square tat holds the circle and check if tat point is inside circle, else find new one. Photon mapping used this.

2nd method i like more cus its more mathematical:
Alpha = GetRandomReal(0,360)
r = SquareRoot(GetRandomReal(0,1))
x = r*radius*cos(Alpha)
y = r*radius*Sin(Alpha)

12-11-2007, 10:19 AM	#2
cohadar	Actually your second method is bugged, you should do: JASS: angle = GetRandomReal(0,2bj_PI) r = SquareRoot(GetRandomReal(0,1)) x = rradiusCos(angle) y = rradius*Sin(angle)

12-11-2007, 10:37 AM	#3
MaD[Lion]	my is not if sin cos takes degree ^^ this is just generic, not really jass related

12-11-2007, 10:43 AM	#4
cohadar	Genetic kills people.

12-11-2007, 11:07 AM

PandaMine

Read this, it helped me a lot (gives everything in relation to geometry for computing in graphics). Its a pdf document BTW (just had to rar it cos Wc3 doesnt accept attachments)

Attached Files

TotalCompendium - Computing Graphics Geometry.rar (481.0 KB)

12-11-2007, 11:53 AM

Tide-Arc Ephemera

Quote:

Originally Posted by cohadar

Genetic kills people.

Genetics obviously made you illiterate, he said generic.
___

From the very little yet useful experience that I've been told, using one formula us SUPPOSED to be faster than two shapes. Basically just trying to find a random distance from a point in a random direction SHOULD be faster than drawing up a square, choosing something inside it, then filtering it out if it's not there and trying again.

At least that's what I heard.

So I think your mathematical (2nd?) method is faster.

12-11-2007, 12:09 PM

cohadar

Quote:

Originally Posted by Tide-Arc Ephemera

Genetics obviously made you illiterate, he said generic.

Nope, genetics made you not see the joke.
It probably has something to do with insufficient number of gray brain cells for lower primates.

12-11-2007, 12:34 PM	#8
Malf	There was no joke.

12-11-2007, 12:48 PM

cohadar

Quote:

Originally Posted by Malf

There was no joke.

Define joke please.
(or not, because this is going waaay of topic)

12-11-2007, 01:13 PM

#10

Anitarf

1st method:
Two random numbers, then three multiplications, a sum and a comparison to see if the chosen point is correct. Has an 1-(pi/4) = 0.215 chance of not being in the circle, in which case the procedure is repeated, which gives you approximately 0.275 repeats per query.

2nd method:
Two random numbers, square root, cos, sin and four multiplications. This one is most likely slower.

Quote:

Originally Posted by cohadar

Define joke please.

Something funny; so, nothing that you ever said. :P

12-11-2007, 01:17 PM

#11

cohadar

Quote:

Originally Posted by Anitarf

Something funny; so, nothing that you ever said. :P

Define funny? Funny to who?

Btw 2 method is indeed slower in c++,
but in jass it is not.

12-11-2007, 01:33 PM

#12

Vexorian

http://mathworld.wolfram.com/DiskPointPicking.html

Imho I like this one:

JASS:

function RandomPointInDisk takes real x, real y, real d returns location
 local real cx = GetRandomReal(-d,d)
    local real ty = SquareRoot(d*d-cx*cx)
    return Location(x+cx, y+ GetRandomReal(-ty,ty) )
endfunction

It is not uniform, but the bias towards the extremes is very small and I would say acceptable for war3, but it is faster than the second method...