OpenEuphoria

2. Re: How to generate random 32bit integers

• Posted by bill Feb 17, 2011
• 1786 views

generate 2 16-bit random numbers and use them for upper 16, lower 16?

4. Re: How to generate random 32bit integers

• Posted by cp Feb 17, 2011
• 1751 views

function rand32() 
   atom A,B 
 
   A = rand(#10000) - 1 
   B = rand(#10000) - 1 
 
   return (A * #10000 + B)  -- Returns an ATOM not an INTEGER, in the range 0 to #FFFFFFFF 
end function 

6. Re: How to generate random 32bit integers

• Posted by DerekParnell (admin) Feb 18, 2011
• 1721 views

Actually, do not try that last one of mine. I just quickly wrote that during my lunch break. Now that I'm home and thought it through, it has holes. Sorry.

Anyhow, here are two different ways that you can try...

-- This first function is used locally and is not suitable for your purposes. 
function r32() 
    -- return a random number from 1 to power(2,32) 
    return ((rand(#10000) - 1) * #10000)  + rand(#10000)  
end function 
 
-- Method A: Calculate two random numbers and join them. 
global function rand_A(atom N)  
   integer A  
   integer B  
  
   if N <= #3FFFFFFF then  
      return rand(N)  
   end if  
  
   A = floor(N / #10000) 
   B = rand(A + 1) - 1 
   if B != A then 
      A = rand(#FFFF) 
   else 
      A = remainder(N,  #10000) 
      if A then 
          A = rand(A) 
      end if 
   end if 
  
   return ((B * #10000) + A) -- Returns an ATOM not an INTEGER, in the range 1 to N  
end function  
 
-- Method B: Calculate a random number such that  0 < R <= 1, then multiple that by N 
global function rand_B(atom N) 
    atom a,b,temp 
    atom A, B 
    atom UL,LL 
     
 
    if N <= #3FFFFFFF then  
        return rand(N)  
    end if  
    UL = (N - 1.5) / N 
    LL = 1.5 / N 
    a = r32() / r32() 
    b = r32() / r32() 
    if a > b then 
        temp = b 
        b = a 
        a = temp 
    end if 
    A = a / b 
    if A > UL then 
        return N 
    elsif A < LL then 
        return 1 
    end if 
    B = floor(N * A) 
     
    return B 
     
end function 
-------------------- 

The distributions are different in the two methods so your sampling might be better with one method rather than the other. You'll have to test which is more suitable.