Bug 1158741 added an omxSP_FFTInv_CCSToR_F32_Sfs_unscaled function as an
optimization which performs the same operation as
omxSP_FFTInv_CCSToR_F32_Sfs except it doesn't scale the results by the
length of the FFT. For consistency with other FFT routines used, it does
multiply the results by two.

The affected files are:
media/openmax_dl/dl/sp/api/omxSP.h
media/openmax_dl/dl/sp/src/omxSP_FFTInv_CCSToR_F32_Sfs_unscaled_s.S