Unifying Bit-Width Optimisation for Fixed-Point and Floating-Point Designs

Unifying Bit-Width Optimisation for Fixed-Point and Floating-Point Designs (2004)

Cached

Download Links

[www.doc.ic.ac.uk]
[www.doc.ic.ac.uk]

Other Repositories/Bibliography

DBLP

by Altaf Abdul Gaffar , Oskar Mencer , Wayne Luk , Peter Y. K. Cheung

Venue:	In 12th Annual IEEE Symposium on Field-Programmable Custom Computing Machines (FCCM04
Citations:	14 - 6 self

BibTeX

@INPROCEEDINGS{Gaffar04unifyingbit-width,
    author = {Altaf Abdul Gaffar and Oskar Mencer and Wayne Luk and Peter Y. K. Cheung},
    title = {Unifying Bit-Width Optimisation for Fixed-Point and Floating-Point Designs},
    booktitle = {In 12th Annual IEEE Symposium on Field-Programmable Custom Computing Machines (FCCM04},
    year = {2004},
    pages = {79--88},
    publisher = {Society Press}
}

Bookmark

OpenURL

Abstract

This paper presents a method that offers a uniform treatment for bit-width optimisation of both fixed-point and floating-point designs. Our work utilises automatic differentiation to compute the sensitivities of outputs to the bitwidth of the various operands in the design. This sensitivity analysis enables us to explore and compare fixed-point and floating-point implementation for a particular design. As a result we can automate the selection of the optimal number representation for each variable in a design to optimize area and performance. We implement our method in the BitSize tool targeting reconfigurable architectures, which takes user-defined constraints to direct the optimisation procedure. We illustrate our approach using applications such as ray-tracing and function approximation. 1

Citations

87	The exact computation paradigm - Yap - 1995
80	Bidwidth analysis with application to silicon compilation - Stephenson, Babb, et al. - 2000
57	Automatic differentiation of algorithms: theory, implementation, and application - George - 1991
38	PRISC: programmable reduced instruction set computers - Razdan - 1994
26	A methodology and design environment for DSP ASIC fixed point refinement - Cmar, Rijnders, et al. - 1999
24	Combined word-length optimization and high-level synthesis of digital signal processing systems - Kum, Sung - 2001
22	The multiple wordlength paradigm - Constantinides, Cheung, et al.
22	Floating point unit generation and evaluation for FPGAs - Liang, Tessier, et al. - 2003
18	Word-length optimization for high-level synthesis of digital signal processing systems - Kum, Sung - 1998
17	Customising floating-point designs - Gaffar, Luk, et al. - 2002
16	Accuracy sensitive word-length selection for algorithm optimization - Wadekar, Parker - 1998
10	Perturbation analysis for word-length optimization - Constantinides - 2003
8	et al. Bitwidth cognizant architecture synthesis of custom hardware accelerators - Mahlke
7	Parameterised floating-point arithmetic on FPGAs - Jaenicke, Luk - 2001
6	A dual-path logarithmic number system addition/subtraction scheme for fpga - Lee, Burgess - 2003
6	editors. Automatic Differentiation of Algorithms - Griewank, Corliss - 1991
5	Floating-point bitwidth analysis via automatic differentiation - Gaffar, Mencer, et al. - 2002
5	Non-uniform segmentation for hardware function evaluation - Lee, Luk, et al. - 2003
5	Constantinides, “Perturbation analysis for word-length optimization - A - 2003
2	BitValue inference:detecting and exploiting narrow bitwidth compilations - Budiu, Goldstein, et al. - 2000
2	A Custom Computing Framework for Orientation and Photography - Fiore - 2000
2	et al. A package for the automatic differentiation of alogorithms written in C/C - Griewank - 1999
1	A library of parameterised floating point modules and their use - Belanovic, Lesser - 2002
1	et al. “A package for the automatic differentiation of algorithms written in C/C - Griewank - 1999
1	Design space explorartion with a Stream compiler - Mencer, Pearce, et al. - 2003