A division-free algorithm for fixed-point power exponential function in embedded system

This work presents a division-free algorithm for fixed-point power exponential function (PEF) using Newton's method. Such a mechanism can improve the computational speed of PEF and is suitable for low-cost embedded systems without floating-point units (FPU). To achieve the goal, this work develops a fast square method to effectively describe a PEF in the form of multiplicative representation. Such representation can be separated into integer and fraction parts. For computing the base term of fraction part in fast square method, a division-free Newton's method is proposed in this paper. The proposed one utilizes two-stage iterations to modify the conventional solving strategy to reduce iteration times when the exponential term is positive. The experimental results show that the proposed algorithm can reduce the execution period about 1.8 times than the baseline one. Additionally, the performance of the proposed algorithm can reach five times higher than that of the system using a floating architecture. The computational precision of the proposed algorithm is also closed to that of the algorithm using floating operations.

[1]  Javier D. Bruguera,et al.  High-speed function approximation using a minimax quadratic interpolator , 2005, IEEE Transactions on Computers.

[2]  K. Takagi,et al.  A VLSI Algorithm for Integer Square-Rooting , 2006, 2006 International Symposium on Intelligent Signal Processing and Communications.

[3]  M. M. Ahmed,et al.  Pipelined implementation of fixed point square root in FPGA using modified non-restoring algorithm , 2010, 2010 The 2nd International Conference on Computer and Automation Engineering (ICCAE).

[4]  Po-Chuan Lin,et al.  Fixed-Point Acceleration of Square Root and Logarithm Using Quadratic Regression for HTK Kernel Modules , 2010, 2010 Fourth International Conference on Genetic and Evolutionary Computing.

[5]  Claude-Pierre Jeannerod,et al.  Computing Floating-Point Square Roots via Bivariate Polynomial Evaluation , 2011, IEEE Transactions on Computers.

[6]  Erdal Oruklu,et al.  Realization of area efficient QR factorization using unified division, square root, and inverse square root hardware , 2009, 2009 IEEE International Conference on Electro/Information Technology.