Fast Fourier Transform (FFT) kernels

Introduction

This module consists of kernels to compute Discrete Fourier Transform(DFT) of data using C7x + MMA.

• The data is assumed to be of complex type.
• The kernels support computation of a batch of several DFT's in one kernel call.
• In a batch of several DFT's, each DFT is referred to as a channel. The data size for all channels in a batch is assumed to be same.
• Unless specified otherwise, the number of complex points for each channel is assumed to be a power of 2.
• The input and output data is arranged in the batch such that channel 1's data follows channel 0's data, channel 2's data follows channel 1's data and so on.
• For each channel, the real and imaginary parts of the data may be interleaved or non-interleaved.
• In case the data is interleaved, the format is as |R0|I0|R1|I1|..., while in case the data is non-interleaved, the format is as |R0|R1|...|I0|I1|... where Rn and In are the real and imaginary parts respectively of the nth complex data point.

Sub Modules
	FFTLIB_FFT_dftLarge_ixX_cxX_oxX
	Kernel for computing DFT using brute-force matrix multiplication for number of points greater than what can be handled by FFTLIB_FFT_dftSmall_ixX_cxX_oxX. brute-force matrix multiplication.
	FFTLIB_FFT_dftSmall_ixX_cxX_oxX
	Kernel for computing DFT for small number of points using brute-force matrix multiplication.
	FFTLIB_FFT_fft_ixX_cxX_oxX
	Kernel for coordinating the DFT computation by calling one of FFTLIB_FFT_dftSmall_ixX_cxX_oxX, FFTLIB_FFT_dftLarge_ixX_cxX_oxX or FFTLIB_FFT_highRadixDecompositions_ixX_cxX_oxX kernels, depending on fftSize and batchSize parameters.
	FFTLIB_FFT_highRadixDecompositions_ixX_cxX_oxX
	Kernel for computing DFT by using a high-radix variant of the FFT alogrithm.