Threaded FFT in Enthought Python

Question

Fast Fourier Transforms (FFTs) in Numpy/SciPy are not threaded. Enthought Python is shipped with the Intel MKL numerical library, which is capable of threaded FFTs. How does one

Answer 1

A cleaner version of my original answer is as follows:

from ctypes import *

mkl = cdll.LoadLibrary("mk2_rt.dll")
c_double_p = POINTER(c_double)
DFTI_COMPLEX = c_int(32)
DFTI_DOUBLE = c_int(36)

def fft2(a):
    Desc_Handle = c_void_p(0)
    dims = (c_int*2)(*a.shape)

    mkl.DftiCreateDescriptor(byref(Desc_Handle), DFTI_DOUBLE, DFTI_COMPLEX, 2, dims )
    mkl.DftiCommitDescriptor(Desc_Handle)
    mkl.DftiComputeForward(Desc_Handle, a.ctypes.data_as(c_void_p) )
    mkl.DftiFreeDescriptor(byref(Desc_Handle))

    return a

def ifft2(a):
    Desc_Handle = c_void_p(0)
    dims = (c_int*2)(*a.shape)

    mkl.DftiCreateDescriptor(byref(Desc_Handle), DFTI_DOUBLE, DFTI_COMPLEX, 2, dims )
    mkl.DftiCommitDescriptor(Desc_Handle)
    mkl.DftiComputeBackward(Desc_Handle, a.ctypes.data_as(c_void_p) )
    mkl.DftiFreeDescriptor(byref(Desc_Handle))

    return a