In migrating from MKL FFT to DFT I desire to use only real/float arrays. In the MKL FFT csfft1dc call I load the real part of the frequency domain to time domainarray in the bottom of the float array and the complex in the top. Can I use the DFTI_REAL_REAL storage typewith theDFTI_REALdomain to achieve the same thing (with a slightly different position of the data in the float array)? I am confused about the difference between the Packing and the storage schemes. Can I use the storage scheme on p 11-42 #4 of the MKL reference manual and ignore the packing stuff to achieve a csfft1dc call?
For more complete information about compiler optimizations, see our Optimization Notice.