Abstract:

An embedded microprocessor generally needs to process fixed point and floating point data. Traditionaly, two executive units will be separately designed for each kind of d ata processing. which increaes the scale and complexity of the whole processor. In this paper, a new method is presented to design a floating point unit of an embeded processor(EMFPU), which could process fixed point data as well as floating point one. The whole design is based on MIL-STD-1750A. The architectu re and implementation of the EMFPU have been shown in detail, and the resolution of the relative "bottleneck\" problem has been discussed elaborately. Firstly, ac cording to the special data format, the algorithms are chosen to be reasonable, and then the datapath and the control logic are designed, while it must be sure to make the two parts work harmoniously. At last, the bottleneck of the whole ci r cuit has been located and removed. The EMFPU is described in VHDL, and verified by ALTERA FPGA. The simulation and verification result shows the design successful.

Key words: FPU, micro pr ocessor, MIT-ST D-1750A, m ant issa, ex po nent