Introduction

Back

 

The Texas Instruments FFTLIB is an optimized FFT Functions Library for C programmers using k2hx devices.

Using these routines assures execution speeds considerably faster than equivalent code written in standard ANSI C language. In addition, by providing ready-to-use DSP functions, TI FFTLIB can significantly shorten application development time.

 

Features and Benefits

The TI k2hx FFTLIB is an optimized library from Texas Instruments for computing the discrete Fourier transform (DFT). FFTLIB provide the ability to perform:

  • Supports C66x TI DSP platform for little-endian
  • Supports single-precision and double-precision floating point
  • Supports complex inputs and real inputs
  • Supports 1D, 2D and 3D FFT
  • Supports Single-core and Multi-core
  • API similar to FFTW, includes FFT plan and FFT execute
 

Software Routines

The rich set of software routines included in FFTLIB are organized into these files:

  • src/fft_xxx/fft_xxx.c: FFT execute C source
  • src/fft_xxx/fft_xxx.h: FFT plan and execute header
  • src/fft_xxx/fft_s(d)p_plan_xxx.c: FFT plan C source
  • src/fft_xxx/fft_xxx_d.c: Unit test/examples
  • src/common/: shared common code in all kernels
 

FFT plan routines

The following information is provided through FFT plan API. Each kernel has a plan function.
  • FFT size
  • Memory arrangement: run from memory directly or use EDMA to transfer data to faster memory
 

FFT size limitation

Due to limit of EDMA multi-dimensional transfer limit, FFT maximum sizes in EDMA mode are limited.
  • 1D FFT single-precision supports size up to 4M,
  • 1D FFT double-precision supports size up to 1M.
  • 2D FFT single-precision supports size up to 2K in each dimension.
  • 2D FFT double-precision supports size up to 1K in each dimension.
  • 3D FFT is not limited, because CPU is used to transfer data at larger size.
 

FFT execute Routine

FFT execute use the plan generated from plan function to calculate FFT.
 

Copyright 2014, Texas Instruments Incorporated