fft - Is the storage of COMPLEX in fortran guaranteed to be two REALs? -

many fft algorithms take advantage of complex numbers stored alternating real , imaginary part in array. creating complex array , passing fft routine, guaranteed can cast real array (of twice size) alternating real , imaginary components?

    subroutine fft (data, n, isign)       dimension data(2*n)        1 i=1,2*n,2         data(i) = ..         data(i+1) = ..  1    continue     return     end      ...     complex s(n)     call fft (s, n, 1)     ... 

(and, btw, dimension data(2*n) same saying real?)

i'm writing answer because experience has taught me write sort of answer 1 of real fortran experts hereabouts piles in correct me.

i don't think current standard, nor of predecessors, states explicitly complex implemented 2 neighbouring-in-memory reals. however, think implementation necessary consequence of standard's definitions of equivalence , of common. don't think have ever encountered implementation in complex not implemented pair of reals.

the standard guarantee, though complex can converted pair of reals. so, given definitions:

complex :: z complex, dimension(4) :: zarr real :: r1, r2 real, dimension(8) :: rarr 

the following might expect

r1 = real(z) r2 = aimag(z) 

both functions elemental , here comes wrinkle:

real(zarr) 

returns 4-element array of reals,

aimag(zarr) 

while

[real(zarr), aimag(zarr)] 

is 8-element array of reals real parts of zarr followed complex parts. perhaps

rarr(1:8:2) = real(zarr) rarr(2:8:2) = aimag(zarr) 

will ok you. i'm not sure of neater way though.

alexander's not 1 able quote standard ! part quotes left me wondering non-default complex scalars. read on, , think para 6 of section points towards germane

a nonpointer scalar object of type not specified in items (1)-(5) occupies single unspecified storage unit different each case , each set of type parameter values, , different unspecified storage units of item (4),

i don't think has impact @ on of answers here.