Floating-point Index Class Of Transcendental Function Computing Algorithm Research And Hardware Implementation

The work of this thesis is supported by the National Advanced Research Project (No. 41308010307). The author participates in the design and verification of a 32-bit CISC microprocessor "Longtium"C1. "Longtium"C1 is compatible with Intel 486DX2 instruction set and has been integrated into "Longtium"S1 as an IP core.Following research work is included in this thesis:1. Take part in the design and implementation of FPU (Floating-Point Unit). The author is reponsible for the architecture design of FPU and the implementation of FIU (Floating-point Interface Unit). Some floating algorithms are also analyzed. The "Longtium"Sl has been tape-out with SMIC 0.18um CMOS technology successfully.2. Study and analyze the familiar algorithms for hardware implementation of the exponential function, including Table lookup method, Polynomial approximation method, Combined method, Digit-by-digit method and CORDIC method.3. Analyze the J.S.Walther CORDIC algorithm and present the "Longtium"C1 CORDIC algorithm for the hardware implmentation based on the demand of area, speed and precision. The author proposes the hardware implementation step of this algorithm, which includes rotation sequence, data format and iteration flow. The implementation can archieve high precesion with low hardware overhead, which also can be used for all the elementary transcendental functions.4. Design the datapath of FPU, which includes mantissa part and exponent part.5. Achieve the verification of FPU at the level of module and system, FPGA debug and design integration.Now DOS6.22 and standard application programs could run on "Longtium"S1 based PC 104 industry control computer successfully. Noticeably, it has been applied in the control system of an Unmanned Aerial Vehicle (UAV) successfully.