Transcendental Functions Of The Hardware Floating Point Unit And Its Optimization

The work of this thesis is based on the National Research Project. The author has taken part in the design and implementation of Longtium C2 microprocessor, and is mainly in charge of the design of FPU and the implementation of transcendental functions.Longtium C2 microprocessor is compatible with Intel 80486DX4 instruction set, and includes basic floating-point operations on single-precision, double-precision and extended-precision format which conforms to standard IEEE 754, and also includes the transcendental functions.The main work of this thesis includes:1. Design the entire structure of FPU in the light of the characteristic and the function of the instructions based on all the floating-point instructions.2. Analyze various algorithms used in the implementation of transcendental functions, discuss HP CORDIC algorithm, and design the entire structure of the implementation of the transcendental functions.3. Finish the design of data path, including adder, shifter, ROM and the control logic of the rotation, and optimize these parts in terms of the requirement of Longtium C2 microprocessor.4. Complete the design of control path, including the entire structure of the control path, the control in terms of the different instructions, the control of several operations, and the detection of the exception of FPU.5. Accomplish the verification and synthesis of floating-point transcendental functions, including module verification, combined module verification, and system verification.The result shows that the design meets the requirement of Longtium C2 microprocessor. The frequency of the design reaches 143MHz. And the backend design is supported by Cadence.