150 likes | 288 Vues
This paper presents an innovative architecture and synthesis method for high-speed computation of fixed-point numerical functions, utilizing lookup table (LUT) cascades. It addresses the performance bottleneck in compute-bound applications by employing piecewise linear approximation based on non-uniform segmentation. The study outlines various architectures for numerical function generators (NFGs), provides design examples, and evaluates performance across different configurations. The proposed method ensures fast generation of NFGs with high precision, proving valuable for a wide range of computing applications.
E N D
Numerical Function GeneratorsUsing LUT Cascades Presenter:Pin-Chong Chen Advisor : Tsung-Chu Huang 2010/05/31
References • Sasao, T.; Nagayama, S.; Butler, J.T.;” Numerical Function Generators Using LUT Cascades” Computers, IEEE Transactions on Volume: 56 , Issue: 6 Digital Object Identifier: 10.1109/TC.2007.1033 Publication Year: 2007
Outline • Introduction • Piecewise linear approximation based on non-uniform segmentation • Architecture for NFGS • Example design • Evaluation and Results • Conclusions
Introduction • The evaluation of functions is often the performance bottleneck of many compute-bound applications. • This paper proposes an architecture and a synthesis method for high-speed computation of fixed-point numerical functions. • This architecture is based on the lookup table (LUT) cascade.
Conclusions • This paper proposed an architecture and a synthesis method for programmable NFGs for various distribution. • This method is useful to generate fast NFGs for a wide range of functions in up to 24-bit precision.