FFTW 3.3.5: 64-bit Guru Interface

When compiled in 64-bit mode on a 64-bit architecture (where addresses cannam@127: are 64 bits wide), FFTW uses 64-bit quantities internally for all cannam@127: transform sizes, strides, and so on—you don’t have to do anything cannam@127: special to exploit this. However, in the ordinary FFTW interfaces, cannam@127: you specify the transform size by an int quantity, which is cannam@127: normally only 32 bits wide. This means that, even though FFTW is cannam@127: using 64-bit sizes internally, you cannot specify a single transform cannam@127: dimension larger than cannam@127: 2³¹−1numbers. cannam@127:

We expect that few users will require transforms larger than this, but, cannam@127: for those who do, we provide a 64-bit version of the guru interface in cannam@127: which all sizes are specified as integers of type ptrdiff_t cannam@127: instead of int. (ptrdiff_t is a signed integer type cannam@127: defined by the C standard to be wide enough to represent address cannam@127: differences, and thus must be at least 64 bits wide on a 64-bit cannam@127: machine.) We stress that there is no performance advantage to cannam@127: using this interface—the same internal FFTW code is employed cannam@127: regardless—and it is only necessary if you want to specify very cannam@127: large transform sizes. cannam@127: cannam@127:

In particular, the 64-bit guru interface is a set of planner routines cannam@127: that are exactly the same as the guru planner routines, except that cannam@127: they are named with ‘guru64’ instead of ‘guru’ and they take cannam@127: arguments of type fftw_iodim64 instead of fftw_iodim. cannam@127: For example, instead of fftw_plan_guru_dft, we have cannam@127: fftw_plan_guru64_dft. cannam@127:

cannam@127:

fftw_plan fftw_plan_guru64_dft(
cannam@127:      int rank, const fftw_iodim64 *dims,
cannam@127:      int howmany_rank, const fftw_iodim64 *howmany_dims,
cannam@127:      fftw_complex *in, fftw_complex *out,
cannam@127:      int sign, unsigned flags);
cannam@127:

The fftw_iodim64 type is similar to fftw_iodim, with the cannam@127: same interpretation, except that it uses type ptrdiff_t instead cannam@127: of type int. cannam@127:

cannam@127:

typedef struct {
cannam@127:      ptrdiff_t n;
cannam@127:      ptrdiff_t is;
cannam@127:      ptrdiff_t os;
cannam@127: } fftw_iodim64;
cannam@127:

Every other ‘fftw_plan_guru’ function also has a cannam@127: ‘fftw_plan_guru64’ equivalent, but we do not repeat their cannam@127: documentation here since they are identical to the 32-bit versions cannam@127: except as noted above. cannam@127:

4.5.6 64-bit Guru Interface