The implementation of Custom instruction of floating sin&cos function based on the FPGA
Huang Shou-li
Abstract
Huang Shou-li
Abstract
This paper introduces the basic principles of the CORDIC algorithm, it makes some research about CORDIC kernel, preprocessor unit and post-processor unit. It improves the calculation precision through increasing the iteration times, optimizing and selecting the arguments based on the traditional CORDIC algorithm, making the designed soft-core adapted to higher precision required occasion. The output data can be directly compatible with most of the processors by the IEEE-754 standard processing, so its application area is widely extended. Finally it accomplish the soft and hardware tests and compares of some trigonometric functions by adding custom instruction on the NiosⅡ.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
This paper introduces the basic principles of the CORDIC algorithm, it makes some research about CORDIC kernel, preprocessor unit and post-processor unit. It improves the calculation precision through increasing the iteration times, optimizing and selecting the arguments based on the traditional CORDIC algorithm, making the designed soft-core adapted to higher precision required occasion. The output data can be directly compatible with most of the processors by the IEEE-754 standard processing, so its application area is widely extended. Finally it accomplish the soft and hardware tests and compares of some trigonometric functions by adding custom instruction on the NiosⅡ.
Key concepts: CORDIC, Computer science, Field-programmable gate array, Trigonometric functions, Computer hardware, Kernel (algebra), Double-precision floating-point format, Parallel computing