| What's New | Introduction | Documentation | Software | Publications | Change History | Contact |
SPL compiler 3.30.
(1) New size rule:
SPL_SIZE_VECTOR_USER. Under this rule, the size of the constructed matrix is defined by the first two elements of the vector.(2) Bug fixes handling matrix elements with registers as subscripts.
A discrete signal transform can be represented by the matrix-vector product, y=Mx, where the vector x denote the input signal, the matrix M the transform, and y the transformed signal. Fast algorithms for computing y=Mx can be obtained by factoring the matrix M into a product of sparse structured matrices.
SPL (Signal Processing Language) is a convenient language for describing matrix factorizations, and hence fast algorithms for computing matrix-vector products. In particular, it can be used for describing fast signal transforms such as FFT. SPL expressions are built from matrix operations such as composition, direct sum, and tensor product, and symbols denoting parameterized families of matrices such as the DFT and stride permutation. An SPL expression can be evaluated to a matrix and can be interpreted as a program to compute the matrix-vector product of the given matrix.
The SPL compiler translates SPL expressions into efficient FORTRAN or C programs. It is part of SPIRAL, an ongoing project that aims to automate the implementation and platform adaptation of signal processing algorithms.
SPL Language Reference
SPL Compiler Reference (coming soon)
SPL Compiler 3.30 3.29 3.28 3.27 3.26 3.25 3.24 3.23 3.22 3.21 Test Packages splse Win32 Binary flex & bison spl-3.27
[1] J.Xiong, J.Johnson, R.Johnson, D.Padua. SPL: A Language and Compiler for DSP Algorithms. PLDI'01. [2] J.Johnson, R.Johnson, D.Padua, J.Xiong. Searching for the Best FFT formulas with the SPL Compiler. LCPC'00. [3] S.Egner, J.Johnson, D.Padua, M.Püschel, J.Xiong. Automatic Derivation and Implementation of Signal Processing Algorithms. To appear in ACM SIGSAM Bulletin Communications in Computer Algebra, 2002. See more SPIRAL publications here.
Email: jxiong@uiuc.edu Home Page: http://polaris.cs.uiuc.edu/~jxiong