SPIRAL: Code Generation for DSP Transforms

SPIRAL: генерация кода для преобразований цифровой обработки сигналов
Franz Franchetti, José M. F. Moura, Manuela Veloso, Markus Püschel, Jeremy Russell Johnson, David Padua, Bryan Singer, Jianxin Xiong, A. Gacic, Yevgen Voronenko, K. Chen, Robert William Johnson, Nick Rizzolo
2005-01-24

SPIRALcode generationdigital signal processing transformsdiscrete Fourier transformfeedback-driven optimization
Fast changing, increasingly complex, and diverse computing platforms pose central problems in scientific computing: How to achieve, with reasonable effort, portable optimal performance? We present SPIRAL, which considers this problem for the performance-critical domain of linear digital signal processing (DSP) transforms. For a specified transform, SPIRAL automatically generates high-performance code that is tuned to the given platform. SPIRAL formulates the tuning as an optimization problem and exploits the domain-specific mathematical structure of transform algorithms to implement a feedback-driven optimizer. Similar to a human expert, for a specified transform, SPIRAL "intelligently" generates and explores algorithmic and implementation choices to find the best match to the computer's microarchitecture. The "intelligence" is provided by search and learning techniques that exploit the structure of the algorithm and implementation space to guide the exploration and optimization. SPIRAL generates high-performance code for a broad set of DSP transforms, including the discrete Fourier transform, other trigonometric transforms, filter transforms, and discrete wavelet transforms. Experimental results show that the code generated by SPIRAL competes with, and sometimes outperforms, the best available human tuned transform library code.
1
Experiments show that SPIRAL-generated code competes with and sometimes outperforms the best available human-tuned transform libraries.
2
It formulates code tuning as an optimization problem and uses feedback-driven search and learning to explore algorithmic and implementation choices.
3
SPIRAL automatically generates platform-tuned, high-performance code for specified linear digital signal processing transforms.
4
SPIRAL exploits domain-specific mathematical structure to guide optimization and match transform implementations to a computer’s microarchitecture.
5
The system supports diverse DSP transforms, including discrete Fourier, trigonometric, filter, and discrete wavelet transforms.

linear digital signal processing (DSP) transforms and their generated implementations on diverse computing platforms

automated generation and feedback-driven optimization of portable high-performance transform code through algorithmic and implementation-space exploration tuned to platform microarchitecture

Publication Details
Publication Date
2005-01-24
Journal
Publisher
ISSN
Cited by
846
Access Type
Author Information
Authors
Franz Franchetti
José M. F. Moura
Manuela Veloso
Markus Püschel
Jeremy Russell Johnson
David Padua
Bryan Singer
Jianxin Xiong
A. Gacic
Yevgen Voronenko
K. Chen
Robert William Johnson
Nick Rizzolo
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%