| 摘要 |
A method and apparatus for use in or with a general-purpose computing environment employs particularly efficient exponential and logarithmic function generation in combination with a logarithmic data format, herein called an exponential/logarithmic (E/L) or exponential floating-point (E-FLP) computational system, to attain general-purpose exponential floating-point (E-FLP) numerical processing comparable in effectiveness to conventional floating-point (C-FLP) processing of similiar precision and dynamic range. An E/L (or E-FLP) quantity according to the invention has an exponential floating-point (E-FLP) value that is inferred from its logarithmic fixed-point (L-FXP) representation or "format" value which when employed with exponential and logarithmic transformations according to the invention yields rapid and precise computational results. [Direct combinatorial construction provides increased speed in microprogrammed embodiments and provides pipe-lined operation in embodiments routing operands in close sequence.] E-FLP computations and their associated L-FXP implementations are disclosed, including the elementary transcendental functions (exponential, logarithm, sine, cosine, tangent and their inverses, etc.). E-FLP computations are characterized by fast multiplication-oriented arithmetic and particularly fast logarithms, exponentials, powers and roots.
|
