Elementary transcendental function operation method based on floating point arithmetic unit and coprocessor for method

Abstract

The invention provides an elementary transcendental function operation method based on a floating point arithmetic unit and a coprocessor for the method. According to the method, a coordinate rotation digital computer (CORDIC) algorithm is decomposed into a function operation control part and a floating point calculation part; the function operation control part is used for finishing an operation control function of the CORDIC algorithm; and the floating point calculation supports absolute value solution, addition / subtraction, multiplication, division, evolution and comparison operation of floating point number, and a general register used in the floating point calculation can be read and written by a function operation control module. A special mathematical function calculation component is redesigned as required in the conventional method, so the conventional method is complex in control and structure and high in consumption of hardware resources. By adding the function operation control module with a simple structure, transcendental function operation is realized by using the existing floating point operation instruction based on the existing floating point arithmetic unit; and the function operation control module is simple in structure and easy to implement and apply, and can support calculation of trigonometric and anti-trigonometric functions, hyperbolic function, exponential function and logarithmic function.

Description

technical field [0001] The invention relates to the technical field of using a floating point operation unit to complete basic transcendental function calculations, especially the calculation of trigonometric and inverse trigonometric functions, hyperbolic functions, exponential functions and logarithmic functions, and realizes a coprocessor based on the method. Background technique [0002] Solving fundamental transcendental functions is the most important operation in scientific computing and engineering applications. Especially in equation solving, numerical analysis, probability statistics, computer graphics, digital signal processing, machine vision, artificial intelligence and other fields are widely used. But the calculation process of the basic transcendental function is complicated, the calculation speed is low, especially the hardware is not easy to realize. Therefore, it has become a research hotspot in recent years to realize a method of high-performance basic t...

AI Technical Summary

Problems solved by technology

The feedback structure only designs one level of CORDIC iterative unit, and then uses the output of this level as the input of this level to complete the iterative operation. This structure has low hardware overhead, but the control is complicated and the speed is slow; each level of the pipeline structure uses a single CORDIC iterative operation. Operation unit, this structure is fast, but the hardware overhead is large, and the higher the precision requirement, the more hardware resources are required

[0011] The above-mentioned hardware for calculating mathematical functions based on the CORDIC algorithm must be redesigned according to requirements to implement a component dedicated to calculating mathematical functions, the control and structure are complex, and the consumption of hardware resources is large.

Images