Fedora :: Fortran Program Gives Different Results Between F15 I686 And X86_64

I am having a problem with a simple program I wrote in Fortran (gfortran). This program basically reads a long list of numbers and makes some operations with them. Double precision is used in the variables and functions, and AFAIK everything is in order with the program. The thing is that I accidentally noticed that said program yields different results, and I am talking about integers being different, something noticeable. Apparently the offending chunk of code might be:

Code:

DOUBLE PRECISION FUNCTION tamano (x1,y1,z1)
DOUBLE PRECISION x1,y1,z1
tamano=dsqrt((x1)**2+(y1)**2+(z1)**2)

[code]....

One of the involved machines has F15 x86_64 and the other one F15 i686. My first guess is that this has something to do with how Fortran handles precision (which I supposed was the same for 32 and 64 bits). To compile the program I use:

Code:

gfortran --free-form -fno-whole-file program.f90

Is there another option I should pass the compiler to make sure the same precision is used in both cases? Or what can I do here?