Multidimensional Fast Fourier Transform Research Articles

The inherent overlapping in the parallel calculation of the Laplacian

A new approach for the parallel computation of the Laplacian in the Fourier domain is presented. This numerical problem inherits the intrinsic sequencing involved in the calculation of any multidimensional Fast Fourier Transform (FFT) where blocking communications assure that its computation is strictly carried out dimension by dimension. Such data dependency vanishes when one considers the Laplacian as the sum of n independent one-dimensional kernels, so that computation and communication can be naturally overlapped with nonblocking communications. Overlapping is demonstrated to be responsible for the speedup figures we obtain when our approach is compared to state-of-the-art parallel multidimensional FFTs.

VkFFT-A Performant, Cross-Platform and Open-Source GPU FFT Library

The Fast Fourier Transform is an essential algorithm of modern computational science. The highly parallel structure of the FFT allows for its efficient implementation on graphics processing units (GPUs), which are now widely used for general-purpose computing. This paper presents the VkFFT - an efficient GPU-accelerated multidimensional Fast Fourier Transform library for Vulkan/CUDA/HIP/OpenCL/Level Zero and Metal projects. VkFFT aims to provide the community with a cross-platform open-source alternative to vendor-specific solutions while achieving comparable or better performance. This paper also discusses the optimizations implemented in VkFFT and verifies its performance and precision on modern high-performance computing GPUs. VkFFT is released under an MIT license.

A Low-Complexity FMCW Surveillance Radar Algorithm Using Two Random Beat Signals.

This paper proposes a low-complexity frequency-modulated continuous wave (FMCW) surveillance radar algorithm using random dual chirps in order to overcome the blind-speed problem and reduce the computational complexity. In surveillance radar algorithm, the most widely used moving target indicator (MTI) algorithm is proposed to effectively remove clutter. However, the MTI algorithm has a so-called ‘blind-speed problem’ that cannot detect a target of a specific velocity. In this paper, we try to solve the blind-speed problem of MTI algorithm by randomly selecting two beat signals selected for MTI for each frame. To further reduce the redundant complexity, the proposed algorithm first performs one-dimensional fast Fourier transform (FFT) for range detection and performs multidimensional FFT only when it is determined that a target exists at each frame. The simulation results show that despite low complexity, the proposed algorithm detects moving targets well by avoiding the problem of blind speed. Furthermore, the effectiveness of the proposed algorithm was verified by performing an experiment using the FMCW radar system in a real environment.

Efficient Near-Field Imaging for Single-Borehole Radar With Widely Separated Transceivers

In this paper, the formulation of Stolt migration is modified for impulse borehole radar near-field imaging in the subsurface scenarios where the transceiver is widely separated with respect to the detection range. The proposed approach consists of the following aspects. First, the locations of the transmitter and receiver in the survey are regarded as independent sample dimensions, and the original sample set is converted to an enlarged virtual sample set. The frequency–wavenumber spectrum (FWS) of the virtual sample set is available via multidimensional fast Fourier transform (FFT). Then, the relation between the angular frequency and wavenumbers of the transmitter and receiver is derived in the frame of the virtual sample set, which provides the basis for the interpolation in angular frequency and weighting process of FWS. By applying multidimensional inverse FFT (IFFT) to the interpolated and weighted FWS of the virtual sample set, the energy of target responses will focus in some profile of the IFFT result, the position of which is related with the separation between the transmitter and receiver. Finally, the desired target space can be extracted from the IFFT result. The improved Stolt migration technique is compared with the conventional Stolt migration algorithm, back-projection method, and Kirchhoff migration algorithm on synthetic data and validated by single-borehole radar experiment in the subsurface scenario. The results show that the developed Stolt migration is superior to the conventional methods in cross-range resolution, computational cost, and the ability to reconstruct locations and shapes of targets.

A decomposition method with minimum communication amount for parallelization of multi-dimensional FFTs

The fast Fourier transform (FFT) is undoubtedly an essential primitive that has been applied in various fields of science and engineering. In this paper, we present a decomposition method for the parallelization of multi-dimensional FFTs with the smallest communication amounts for all ranges of the number of processes compared to previously proposed methods. This is achieved by two distinguishing features: adaptive decomposition and transpose order awareness. In the proposed method, the FFT data is decomposed based on a row-wise basis that maps the multi-dimensional data into one-dimensional data, and translates the corresponding coordinates from multi-dimensions into one dimension so that the one-dimensional data can be divided and allocated equally to the processes using a block distribution. As a result and different from previous works that have the dimensions of decomposition pre-defined, our method can adaptively decompose the FFT data on the lowest possible dimensions depending on the number of processes. In addition, this row-wise decomposition provides plenty of alternatives in data transpose, and different transpose order results in different amounts of communication. We identify the best transpose orders with the smallest communication amounts for the 3-D, 4-D, and 5-D FFTs by analyzing all possible cases. We also develop a general parallel software package for the most popular 3-D FFT based on our method using the 2-D domain decomposition. Numerical results show good performance and scaling properties of our implementation in comparison with other parallel packages. Given both communication efficiency and scalability, our method is promising in the development of highly efficient parallel packages for the FFT.

Improved Twiddle Access for Fast Fourier Transforms

Optimizing the number of arithmetic operations required in fast Fourier transform (FFT) algorithms has been the focus of extensive research, but memory management is of comparable importance on modern processors. In this article, we investigate two known FFT algorithms, G and GT , that are similar to Cooley-Tukey decimation-in-time and decimation-in-frequency FFT algorithms but that give an asymptotic reduction in the number of twiddle factor loads required for depth-first recursions. The algorithms also allow for aggressive vectorization (even for non-power-of-2 orders) and easier optimization of trivial twiddle factor multiplies. We benchmark G and GT implementations with comparable Cooley-Tukey implementations on commodity hardware. In a comparison designed to isolate the effect of twiddle factor access optimization, these benchmarks show typical speedups ranging from 10% to 65%, depending on transform order, precision, and vectorization. A more heavily optimized implementation of GT yields substantial performance improvements over the widely used code FFTW for many transform orders. The twiddle factor access optimization technique can be generalized to other common FFT algorithms, including real-data FFTs, split-radix FFTs, and multidimensional FFTs.

High-Productivity Software Development with pMatlab

In this paper, we explore the ease of tackling a communication-intensive parallel computing task - namely, the 2D fast Fourier transform (FFT). We start with a simple serial Matlab code, explore in detail a ID parallel FFT, and illustrate how it can be extended to multidimensional FFTs.

Accelerating and focusing protein–protein docking correlations using multi-dimensional rotational FFT generating functions

Predicting how proteins interact at the molecular level is a computationally intensive task. Many protein docking algorithms begin by using fast Fourier transform (FFT) correlation techniques to find putative rigid body docking orientations. Most such approaches use 3D Cartesian grids and are therefore limited to computing three dimensional (3D) translational correlations. However, translational FFTs can speed up the calculation in only three of the six rigid body degrees of freedom, and they cannot easily incorporate prior knowledge about a complex to focus and hence further accelerate the calculation. Furthemore, several groups have developed multi-term interaction potentials and others use multi-copy approaches to simulate protein flexibility, which both add to the computational cost of FFT-based docking algorithms. Hence there is a need to develop more powerful and more versatile FFT docking techniques. This article presents a closed-form 6D spherical polar Fourier correlation expression from which arbitrary multi-dimensional multi-property multi-resolution FFT correlations may be generated. The approach is demonstrated by calculating 1D, 3D and 5D rotational correlations of 3D shape and electrostatic expansions up to polynomial order L=30 on a 2 GB personal computer. As expected, 3D correlations are found to be considerably faster than 1D correlations but, surprisingly, 5D correlations are often slower than 3D correlations. Nonetheless, we show that 5D correlations will be advantageous when calculating multi-term knowledge-based interaction potentials. When docking the 84 complexes of the Protein Docking Benchmark, blind 3D shape plus electrostatic correlations take around 30 minutes on a contemporary personal computer and find acceptable solutions within the top 20 in 16 cases. Applying a simple angular constraint to focus the calculation around the receptor binding site produces acceptable solutions within the top 20 in 28 cases. Further constraining the search to the ligand binding site gives up to 48 solutions within the top 20, with calculation times of just a few minutes per complex. Hence the approach described provides a practical and fast tool for rigid body protein-protein docking, especially when prior knowledge about one or both binding sites is available.

Scalable parallel FFT for spectral simulations on a Beowulf cluster

The implementation and performance of the multidimensional Fast Fourier Transform (FFT) on a distributed memory Beowulf cluster is examined. We focus on the three-dimensional (3D) real transform, an essential computational component of Galerkin and pseudo-spectral codes. The approach studied is a 1D domain decomposition algorithm that relies on communication-intensive transpose operation involving P processors. Communication is based upon the standard portable message passing interface (MPI). We show that 1/ P scaling for execution time at fixed problem size N 3 (i.e., linear speedup) can be obtained provided that (1) the transpose algorithm is optimized for simultaneous block communication by all processors; and (2) communication is arranged for non-overlapping pairwise communication between processors, thus eliminating blocking when standard fast ethernet interconnects are employed. This method provides the basis for implementation of scalable and efficient spectral method computations of hydrodynamic and magneto-hydrodynamic turbulence on Beowulf clusters assembled from standard commodity components. An example is presented using a 3D passive scalar code.

Effective implementations of multi-dimensional radix-2 FFT

The Fast Fourier Transform (FFT), is well known as a fast method for the Discrete Fourier Transform (DFT), and the calculation time is in proportion to Nlog N, where N is the system size. But unfortunately, inappropriate implementation with no care for the structure of the target machine increases the proportional coefficient by a factor of 10. We propose effective implementations in the case of multi-dimensional radix-2 FFT for the recent RISC workstation and the vector-type supercomputer, respectively.

A new multidimensional FFT based on one-dimensional decompositions

This work proposes an original multidimensional fast Fourier transform (FFT) algorithm where the computation is first organized into multiplier-free butterflies and then completed by 1-D FFTs. The properties of well-known 1-D FFT algorithms blend in quite nicely with those of the proposed multidimensional FFT scheme, extending their computational and structural characteristics to it. Strong points of the proposed method are that its total computational cost decreases as the signal space dimensions increase and that its efficiency is superior to that of any other multidimensional FFT algorithm.

A PARALLEL FAST FOURIER TRANSFORM

In this paper we discuss the general problem of implementing the multidimensional Fast Fourier Transform algorithm on parallel computers. We show that, on a machine with P processors and fully parallel node communications, the optimal asymptotic scaling behavior of the total computational time with the number of data points, N, given in d dimensions by the formula aN/P log (N/P)+bN/P(d-1)/d, can actually be achieved on realistic platforms. As a concrete realization of our strategy, we have produced codes efficiently running on machines of the APE family and on Cray T3E. On the former for asymptotic values of N our codes attain the above optimal result.

Equiripple FIR filter design by the FFT algorithm

The fast Fourier transform (FFT) algorithm has been used in a variety of applications in signal and image processing. In this article, a simple procedure for designing finite-extent impulse response (FIR) discrete-time filters using the FFT algorithm is described. The zero-phase (or linear phase) FIR filter design problem is formulated to alternately satisfy the frequency domain constraints on the magnitude response bounds and time domain constraints on the impulse response support. The design scheme is iterative in which each iteration requires two FFT computations. The resultant filter is an equiripple approximation to the desired frequency response. The main advantage of the FFT-based design method is its implementational simplicity and versatility. Furthermore, the way the algorithm works is intuitive and any additional constraint can be incorporated in the iterations, as long as the convexity property of the overall operations is preserved. In one-dimensional cases, the most widely used equiripple FIR filter design algorithm is the Parks-McClellan algorithm (1972). This algorithm is based on linear programming, and it is computationally efficient. However, it cannot be generalized to higher dimensions. Extension of our design method to higher dimensions is straightforward. In this case two multidimensional FFT computations are needed in each iteration.

Reconstruction of magnetic resonance images using one-dimensional techniques

Whenever DFT (discrete Fourier transform) processing of a multidimensional discrete signal is required, one can apply either a multidimensional FFT (fast Fourier transform) algorithm, or a single-dimension FFT algorithm, both using the same number of points. That is, the dimensions of a "multidimensional" signal, and of its spectrum, are a matter of choice. Every multidimensional sequence is completely equivalent to a one-dimensional function in both "time" and "frequency" domains. This statement applied to MRI (magnetic resonance imaging) explains why one can reconstruct the slice by using either one-dimensional or two-dimensional methods, as it is already done in echo planar methods. In the commonly used spin warp methods, the image can be also reconstructed by either one- or two-dimensional processing. However, some artifacts in the images reconstructed from the original "zig-zag" echo planar trajectory, are shown to be due to the wrong dimensionality of the FFT applied.

A unified treatment of Cooley-Tukey algorithms for the evaluation of the multidimensional DFT

In this paper the Cooley-Tukey fast Fourier transform (FFT) algorithm is generalized to the multidimensional case in a natural way which allows for the evaluation of discrete Fourier transforms of rectangularly or hexagonally sampled signals or of signals which are sampled on an arbitrary periodic grid in either the spatial or Fourier domain. This general algorithm incorporates both the traditional rectangular row-column and vector-radix algorithms as special cases. This FFT algorithm is shown to result from the factorization of an integer matrix; for each factorization of that matrix, a different algorithm can be developed. This paper presents the general algorithm, discusses its computational efficiency, and relates it to existing multi-dimensional FFT algorithms.