Home

This is the home page of the SPIRAL project. The goal of SPIRAL is to push the limits of automation in software and hardware development and optimization for digital signal processing (DSP) algorithms and other numerical kernels beyond what is possible with current tools.

Our basic research question is

Can we teach computers to write fast libraries?

Our flagship is the SPIRAL program generation system, which, entirely autonomously, generates platform-tuned implementations of signal processing transform such as the discrete Fourier transform, discrete cosine transform, and many others. Look at a few benchmarks. But we also provide other online generators (see the right column).

SPIRAL addresses one of the current key problems in numerical software and hardware development: how to achieve close to optimal performance with reasonable coding effort? (More detailed problem statement.)

SPIRAL comprises an interdisciplinary team of researchers in the areas of signal processing, algorithms, scientific computing, compilers, computer architecture, and mathematics.

In the domain of linear transform, and for standard multicore platforms (Core 2 Duo like), we have achieved complete automation: the computer generation of general input-size, vectorized, parallel libraries.

Ongoing Research

• Generate code for Intel IPP 6.0, which now has an entire domain for Spiral-generated code
• DFT code generation for the Cell BE
• Library generation for matrix-matrix multiplication
• DFT IP generation for FPGAs
• Software generation for Viterbi decoding
• Software generation for synthetic aperture Radar imaging
• Software generation for JPEG2000
• Automatic platform/application co-design

SPIRAL Overview Paper (Good Choice for Citing)

• Markus Püschel, José M. F. Moura, Jeremy Johnson, David Padua, Manuela Veloso, Bryan Singer, Jianxin Xiong, Franz Franchetti, Aca Gacic, Yevgen Voronenko, Kang Chen, Robert W. Johnson, and Nick Rizzolo
  SPIRAL: Code Generation for DSP Transforms
  Proceedings of the IEEE special issue on "Program Generation, Optimization, and Adaptation," Vol. 93, No. 2, 2005, pp. 232-275

After this paper we started to attack all forms of parallel platforms. Check out benchmarks or learn about the current Spiral system.

Here is a recent talk (February 2007).

Recent Papers

• Yevgen Voronenko
  Library Generation for Linear Transforms
  PhD. thesis, Electrical and Computer Engineering, Carnegie Mellon University, 2008
• Peter A. Milder, Franz Franchetti, James C. Hoe and Markus Püschel
  Formal Datapath Representation and Manipulation for Implementing DSP Transforms
  Proc. Design Automation Conference (DAC), 2008
• Franz Franchetti and Markus Püschel
  Generating SIMD Vectorized Permutations
  Proc. International Conference on Compiler Construction (CC), Lecture Notes in Computer Science, Springer, Vol. 4959, pp. 116-131, 2008
  
• Srinivas Chellappa, Franz Franchetti and Markus Püschel
  How To Write Fast Numerical Code: A Small Introduction
  to appear in Proc. Summer School on Generative and Transformational Techniques in Software Engineering, Lecture Notes in Computer Science, Springer, 2008
• Markus Püschel, Peter A. Milder, and James C. Hoe
  Permuting Streaming Data Using RAMs
  submitted for publication and patent application

Online Generators

Online generators currently available

Example interface (click to go to the site)

Featured Result

DFT: Four-way vectorized, up to four-threads, "push-button" generated by Spiral. Result: up to 25 Gflop/s.

References

• Franz Franchetti, Yevgen Voronenko and Markus Püschel
  FFT Program Generation for Shared Memory: SMP and Multicore
  Proc. Supercomputing (SC), 2006

More benchmarks.

Contact

If you have questions or comments about SPIRAL or our work in general, please send email to help (at) spiral.net.

SPIRAL Team

Proceedings IEEE Special Issue

Program Generation, Optimization, and Adaptation

Other links:

• Becoming a student in Spiral
• Spiral Blog

visitors since July 2003

insurance quotes Counter