Method and device for implementing floating-point angle compression

Abstract

The invention discloses a floating point angular compression implementation method and device. The method includes the steps that firstly, an interval needing to participate in 2 / pi multiplication is obtained according to indexes of a to-be-compressed angle; secondly, the mantissa of the to-be-compressed angle and the obtained interval execute 2 / pi multiplication by means of a truncated multiplier, and a multiplication result is output; thirdly, an integer part and a decimal place in the multiplication result are obtained, the quadrant of the compressed angle is determined according to the integer part, and the indexes and a first mantissa are obtained by the decimal place through leading zero and shifting; fourthly, the first mantissa and the multiplier execute pi / 2 multiplication by means of the truncated multiplier, and a second mantissa is obtained and output; fifthly, the compressed angle is determined according to the indexes and the second mantissa. The device comprises a preprocessing module, a 2 / pi truncated multiplication module, a leading zero and shifting module, a pi / 2 truncated multiplication module and a post-processing module. By means of the method and the device, angle compression of any floating point can be achieved, and the advantages that the implementation method is simple, the compression efficiency and precision are high and the cost is low are achieved.

Description

technical field [0001] The invention relates to the technical field of function calculation, in particular to a method and device for realizing floating-point angle compression. Background technique [0002] Basic function calculation is an important part of scientific calculation and engineering applications. The first step in function calculation usually needs to perform angle compression. In basic function calculation processing such as trigonometric functions, the input angle is required to be within a certain range. For Any input angle needs to compress the input angle to the specified range while ensuring the accuracy of the compressed angle. [0003] When the angle compression is defined by the mod function, the input angle Z is where Z and is a real number, and K is an integer, When the input angle Z is closer to Integer multiples of , limited number of digits The less effective digits will become, that is, the more leading zero digits will become, the le...

AI Technical Summary

Problems solved by technology

[0005] (1) Pure software implementation, such as multi-precision algorithm, can improve the compression angle accuracy, but the calculation speed is slow;

[0006] (2) Combining with the implementation of hardware pipeline structure, the accuracy of compression angle can be improved, and the compression speed is faster; but in this type of method, it needs to rely on the implementation of the specified hardware structure, such as the need to rely on the existing fma on the processor Fusion unit, not convenient for practical use; or improved hardware implementation based on DRMRR (Double-residue modular range reduction), such as CER (Constant Elementary Residues), ORC (On-the-fly Elementary Residues Computation), DTR (Drmrr with Table Replication) The equal angle compression method can make the overhead of the iterative structure very small, but usually uses a full multiplier. As mentioned above, when the input angle Z is closer to When the integer multiple of , the use of a complete multiplier requires a large number of use of the remaining angle lookup table, there is a problem of too much overhead

Images