Device and method for hardware-efficient adaptive calculation of floating-point trigonometric functions using coordinate rotate digital computer (CORDIC)

An accelerator circuit, comprising:

• a register file comprising:
• instruction registers to store an instruction for evaluating a trigonometric function;
• and
• data registers comprising a first data register to store a floating-point input value associated with the instruction;
• a determination circuit communicatively coupled to the register file, to:
• identify the trigonometric function and the floating-point input value associated with the instruction;
• and
• determine whether the floating-point input value is in a small value range;
• an approximation circuit communicatively coupled to the determination circuit, to:
• responsive to determining that the floating-point input value is in the small value range, receive the floating-point input value and calculate an approximation of the trigonometric function with respect to the floating-point input value;
• and
• a coordinate rotate digital computer (CORDIC) circuit, to:
• responsive to determining that the floating-point input value is outside the small value range, receive the floating-point input value and calculate the trigonometric function with respect to the floating-point input value using CORDIC recurrences, wherein each of the CORDIC recurrences comprises at least one of a single rotation operation with respect to the floating-point input value or a double rotation operation with respect to the floating-point input value;
• and
• to calculate the trigonometric function with respect to the floating-point input value using CORDIC recurrences, the CORDIC circuit is further to:
• split the floating-point input value into a whole number part and a fractional number part;
• determine, based on the whole number part, a quadrant in which the floating-point input value is locate;
• convert the fractional number part into a fixed-point number;
• calculate, using the CORDIC recurrences, the trigonometric function with respect to the fixed-point number to generate a fixed-point result;
• convert the fixed-point result into a floating-point result;
• determine a sign based on the quadrant and the trigonometric function;
• and
• store the sign and the floating-point result in a third data register of the data registers.