Copyrights to these papers may be held by the publishers. The download files are preprints. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.
Jianxin Xiong, Jeremy Johnson, Robert W. Johnson and David Padua (Proc. Programming Languages Design and Implementation (PLDI), pp. 298-308, 2001)
SPL: A Language and Compiler for DSP Algorithms
Preprint (226 KB)
Published paper (link to publisher)
In this paper, we discuss the design and implementation of the SPL compiler, a domain specific compiler that translates signal processing algorithms expressed as mathematical formulas in the SPL language into efficient programs in high-level programming language such as FORTRAN and C. The SPL compiler was developed as part of SPIRAL, a system which utilizes formula transformations and intelligent search strategies to automatically create optimized digital signal processing (DSP) libraries. After presenting the translation and optimization techniques utilized by the compiler, empirical data is presented showing the efficiency of the resulting C/FORTRAN code. Timings are compared, using the fast Fourier transform (FFT) as a benchmark, to those obtained by FFTW, one of the fastest FFT packages available. The results of this comparison show that the SPL compiler produces code that is competitive and, in many cases, faster than FFTW.Keywords: SPL compiler: Translating math into code