Floating-point division circuit

Boots – shoes – and leggings

Patent

Rate now

  [ 0.00 ] – not rated yet Voters 0   Comments 0

Details

364761, G06F 738, G06F 752

Patent

active

053093836

DESCRIPTION:

BRIEF SUMMARY
FIELD OF THE INVENTION

The present invention relates to a floating-point division circuit, and more particularly to a division circuit for performing division of floating-point data defined by the Institute of Electrical and Electronics Engineers (IEEE) standard.
A vector arithmetic unit in a vector processor performs addition, subtraction, multiplication, and division of vector data. Such a vector arithmetic unit is mounted in a large scale semiconductor circuit (LSI) and division of vector data is conventionally performed by a multiplication circuit thereof. Since it is desirable, in terms of increasing the speed and operational precision of a vector processor, that each of the arithmetic operations be performed in a parallel manner by specialized circuits, recently division circuits have been provided separately from multiplication circuit.
However, providing a division circuit separately from a multiplication circuit leads to increased power consumption of LSI due to the addition of a division circuit. This increase necessitates a reduction in power consumption of a division circuit.


BACKGROUND ART

A floating-point division circuit executes an instruction for division by repeating a recursive procedure defined by the following equation. is either 0 or an integer in the range of 1 to n-1, D a divisor, q.sub.j + 1 a (j + 1)th quotient digit counting from a decimal point (q.sub.0 is a sign), n a digit length of a quotient, r a radix, r .times. R (j) a partial dividend before the (j + 1)th quotient digit is determined, R (j + 1) a partial remainder after the (j + 1)th quotient digit is determined, R (0) a dividend (initial value of the partial dividend), R (n) a final remainder.
The quotient Q is expressed by a series of (n + 1) quotient digits q.sub.0 - q.sub.n, as shown in the following equation.
Accordingly, most of the time needed for an execution of a division is spent on a repetition of the above-mentioned recursive procedure. One approach to reduce the repetitions of this recursive procedure, so that division is speedily performed, is known as a non-recovery type division method. In this non-recovery type division method, attention is paid to the fact that a set of quotients used for generating a quotient digit q.sub.j + 1 could be a set of signed quotients not including 0, before selecting the quotient digit from the quotient set. Assuming that r is a radix, the above-mentioned set of signed quotients can be given by:
This non-recovery type division method is characterized in that negative numbers are allowed to represent digits of the result of an operation by not recovering a reversal from positive to negative dividend, the reversal arising when determining each digit of the result of the operation; and that the divisor or a multiple of the divisor is added to or subtracted from the digit, depending on the sign of the dividend; the method is hence called a separation method.
In this method, a quotient is obtained by first placing, in a divisor register, a data for a multiple of the divisor that is k times the divisor (obtained by multiplying the divisor by each quotient of the above-mentioned set of signed quotients), selecting the above-mentioned divisor register on the basis of a prediction signal output from a quotient predicting device, and repeating addition or subtraction of the multiple of the divisor that is k times the divisor.
One variation of this non-recovery type division method, employed when a faster and more accurate division is to be performed, is known as a large-radix non-recovery division method that reduces the number of loops of division by increasing the radix r so that the number of bits that serve as an operation unit is relatively large.
A division circuit of this configuration can speed up an operation of a vector processor by having a division circuit running parallel with an addition circuit, a subtraction circuit, and a multiplication circuit. It should be noted, however, that addition, subtraction, multiplication, and division circuits do not receive an operatio

REFERENCES:
patent: 4760550 (1988-07-01), Katzman et al.
patent: 4996660 (1991-02-01), Beacom et al.
patent: 4999801 (1991-03-01), Katsuno
8084 IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 36, No. 2, Feb. 1988, New York, N.Y., USA, Architectural Strategies for an Application-Specific Synchronous Multiprocessor Environment, F. Catthoor et al., pp. 265-238.
Patent Abstract of Japan, JP3091028, Apr. 16, 1991.

LandOfFree

Say what you really think

Search LandOfFree.com for the USA inventors and patents. Rate them and share your experience with other people.

Rating

Floating-point division circuit does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Floating-point division circuit, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Floating-point division circuit will most certainly appreciate the feedback.

Rate now

     

Profile ID: LFUS-PAI-O-2120310

  Search
All data on this website is collected from public sources. Our data reflects the most accurate information available at the time of publication.