A universal nonlinear component and its application to WSI

Presents a high-speed multifunction chip for performing one of four nonlinear operations: 1) square root, 2) reciprocal, 3) sine/cosine, and 4) arctangent. Each of these functions is evaluated with one ROM access, two additions, and one major and one minor multiplication, yielding a new result every two clock cycles. Its performance signifies an estimated three-to-four-fold increase in speed (for comparable technologies and minimum feature size) over existing approaches. Furthermore, since all four functions are performed on the same cell, a silicon-area advantage of approximately three is realized when the application demands multiple functions. In wafer scale integration (WSI) of signal and image processing algorithms, several such functions are usually needed, while defect tolerance dictates the use of just one or two types of cells. Thus the new component is ideally suited for monolithic WSI. However, it can also be used as a co-processor/accelerator for commercial DSP chips in hybrid WSI implementation of signal processing algorithms. The underlying principle, which has made the combined goals of high-speed and multifunctionality possible, is second-order interpolation of very small ROM tables. Two versions are presented: a 24-b chip, and a 16-b chip, both fabricated in 2.0- mu m CMOS technology. >

[1]  Granino A. Korn,et al.  Table-Lookup/Interpolation Function Generation for Fixed-Point Digital Computations , 1969, IEEE Transactions on Computers.

[2]  Earl E. Swartzlander,et al.  A radix-8 wafer scale FFT processor , 1992, J. VLSI Signal Process..

[3]  Israel Koren,et al.  Evaluating Elementary Functions in a Numerical Coprocessor Based on Rational Approximations , 1990, IEEE Trans. Computers.

[4]  Vijay K. Jain,et al.  An architecture for WSI rapid prototyping , 1992, Computer.

[5]  E. Pearse O'Grady,et al.  A Hardware-Oriented Algorithm for Floating-Point Function Generation , 1991, IEEE Trans. Computers.

[6]  C.E. Alvarez,et al.  Systolic L-U decomposition array with a new reciprocal cell , 1989, Proceedings 1989 IEEE International Conference on Computer Design: VLSI in Computers and Processors.

[7]  Kai Hwang,et al.  Evaluating elementary functions with Chebyshev polynomials on pipeline nets , 1987, 1987 IEEE 8th Symposium on Computer Arithmetic (ARITH).

[8]  Earl E. Swartzlander,et al.  Arithmetic error analysis of a new reciprocal cell , 1992, Proceedings 1992 IEEE International Conference on Computer Design: VLSI in Computers & Processors.

[9]  James O. Hamblen,et al.  A fast technique for the evaluation of unary floating point functions , 1988 .

[10]  William J. Kubitz,et al.  Minimum Mean Running Time Function Generation Using Read-Only Memory , 1983, IEEE Transactions on Computers.