Generating high performance pruned FFT implementations

Franchetti, Franz; Püschel, Markus

doi:10.1109/icassp.2009.4959642

Cited by 11 publications

(12 citation statements)

References 13 publications

(14 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some implementations even use pruned algorithms [14] to take advantage of the zero elements to save some operations. Batch execution is specially adequate for processing small signals on GPUs due to their massive parallel architecture, otherwise the GPU would not be able to properly exploit its computational resources and the cost of device memory transfers and kernel launch would outweigh any benefits of using the GPU.…”

Section: Resultsmentioning

confidence: 99%

FFT Implementation on a Streaming Architecture

Lobeiras

Amor

Doallo

2011

2011 19th International Euromicro Conference on Parallel, Distributed and Network-Based Processing

View full text Add to dashboard Cite

Fast Fourier Transform (FFT) is a useful tool for applications requiring signal analysis and processing. However, its high computational cost requires efficient implementations, specially if real time applications are used, where response time is a decisive factor. Thus, the computational cost and wide application range that requires FFT transforms has motivated the research of efficient implementations. Recently, GPU computing is becoming more and more relevant because of their high computational power and low cost, but due to its novelty there is some lack of tools and libraries. In this paper we propose an efficient implementation of the FFT with AMD's Brook+ language. We describe several features and optimization strategies, analyzing the scalability and performance compared to other well-known existing solutions.

show abstract

Section: Resultsmentioning

confidence: 99%

FFT Implementation on a Streaming Architecture

Lobeiras

Amor

Doallo

2011

2011 19th International Euromicro Conference on Parallel, Distributed and Network-Based Processing

View full text Add to dashboard Cite

show abstract

“…This leads to FFT computations where some of the inputs are equal to zero and not all of the outputs are needed. The pruned or truncated DFT [4,18] were developed to take advantage of this situation. In [18], van der Hoeven introduced a radix-2 algorithm for the truncated Fourier transform (TFT), and showed how to invert the TFT (ITFT).…”

Section: Introductionmentioning

confidence: 99%

High performance implementation of the TFT

Meng

Johnson

2014

Proceedings of the 39th International Symposium on Symbolic and Algebraic Computation

View full text Add to dashboard Cite

This paper reports on a high-performance implementation of the truncated Fourier transform (TFT). A general CooleyTukey like algorithm for the TFT is developed that allows the implementation to automatically adapt to the memory hierarchy. Then the algorithm introduces a small relaxation for larger transform sizes which trades off slightly higher arithmetic cost for improved data flow which allows full vectorization and parallelization. The implementation is automatically derived and tuned using the SPIRAL system for code generation and adaptation. The resulting arbitrarysize TFT library smooths out the staircase performance associated with power-of-two modular FFT implementations while retaining the performance associated with state-of-theart FFT libraries. This provides significant performance improvement over approaches that pad to the next power of two even when using high-performance FFT libraries.

show abstract

“…However, this approach requires that the location of the k nonzero outputs are known. Pruned FFTs can be optimized and also efficiently implemented using SIMD instructions [6]. A different approach is taken by the FADFT-2 [7], [8], which is a probabilistic algorithm that requires O(k polylog(n)) many operations and does not require the location of the nonzero outputs.…”

Section: Introductionmentioning

confidence: 99%

High-performance sparse fast Fourier transforms

Schumacher¹,

Püschel²

2014

2014 IEEE Workshop on Signal Processing Systems (SiPS)

Self Cite

View full text Add to dashboard Cite

The sparse fast Fourier transform (SFFT) is a recent novel algorithm to compute discrete Fourier transforms on signals with a sparse frequency domain with an improved asymptotic runtime. Reference implementations exist for different variants of the algorithm and were already shown to be faster than state-of-the-art FFT implementations in cases of sufficient sparsity. However, to date the SFFT has not been carefully optimized for modern processors. In this paper, we first analyze the performance of the existing SFFT implementations and discuss possible improvements. Then we present an optimized implementation. We achieve a speedup of 2-5 compared to the existing code and an efficiency that is competitive to highperformance FFT libraries.

show abstract

Generating high performance pruned FFT implementations

Cited by 11 publications

References 13 publications

FFT Implementation on a Streaming Architecture

FFT Implementation on a Streaming Architecture

High performance implementation of the TFT

High-performance sparse fast Fourier transforms

Contact Info

Product

Resources

About