US 5,309,385 A · Vector division processing method and system

Abstract

In vector dividing the process employing convergence division method, the process steps are performed for performing pipeline processing of multiplying operation of the dividend and a first convergence factor in a speed of one clock cycle per one element to generate a first intermediate result vector data as a data before convergence, for performing pipeline processing of multiplying operation of the divisor and the first convergence factor in a speed of one clock cycle per one element to generate a second intermediate result vector data as a data before convergence, for storing the first and second intermediate result vector data to first and second intermediate result storing vector registers, respectively, for reading out of the first and second intermediate result vector data from the first and second intermediate result storing vector registers, respectively, per every clock cycles, for performing pipeline processing for multiplying operation of the read out intermediate result vector data and a second convergence factor as the next convergence factor in a speed of one clock cycle per one element to generate third and fourth intermediate result vector data, and for repeating the same process from the third step to fifth step until reaching to a nth convergence factor as a predetermined final convergence factor, and outputting a quotient vector as the final convergence result in a speed of one clock cycle per one element.

A vector division processing apparatus for performing a division operation on dividend vector data composed of a plurality of elements and divisor vector data composed of a plurality of elements by employing a convergence division method in which each of said dividend vector data and said divisor vector data is multiplied in sequence by a plurality of convergence factor vector data to thereby converge resultant data of the sequential multiplication operations performed on said dividend vector data to a quotient and converge resultant data of the sequential multiplication operation performed on said divisor vector data to 1, said apparatus comprising:

a first vector register having a first storage area for temporarily storing said dividend vector data and a second storage area,
a second vector register having a third storage area for temporarily storing said divisor vector data and a fourth storage area, and
an operation unit having a first input terminal coupled to said first vector register to receive data supplied therefrom, a second input terminal coupled to said second vector register to receive data supplied therefrom and an output terminal coupled to each of said first and second vector registers, said operation unit performing;
a first processing operation in which respective multiplication operations on said elements of said dividend vector data stored in said first storage area of said first vector register and associated ones of said convergence factor vector data are performed in sequence to produce a plurality of first intermediate resultant data and said plurality of first intermediate resultant data are written in sequence into said second storage area of said first vector register,
a second processing operation in which respective multiplication operations on said elements of said divisor vector data stored in said third storage area of said second vector register and associated ones of said convergence factor vector data are performed in sequence to produce a plurality of second intermediate resultant data and said plurality of second intermediate resultant data are written in sequence into said fourth stage area of said second vector register,
after said first and second processing operations are completed, a third processing operation in which respective multiplication operations on said plurality of first intermediate resultant data stored in said second storage area of said first vector register and associated ones of said convergence factor vector data are performed in sequence to produce a plurality of third intermediate resultant data and said plurality of third intermediate resultant data are written in sequence into said second storage area of said first vector register, and
after said first and second processing operations are completed, a fourth processing operation in which respective multiplication operations on said plurality of second intermediate resultant data stored in said fourth storage area of said second vector register and associated ones of said convergence factor vector data are performed in sequence to produce a plurality of fourth intermediate resultant data and said plurality of fourth intermediate resultant data are written in sequence into said fourth storage area of said second vector register.

4 more elements

Vector division processing method and system

Life of this patent

Vector division processing method and system

Life of this patent

Abstract

Claims6