Method, system and computer program product for determining required precision in fixed-point divide operations

1. A method of controlling a fixed point division operation, the method comprising:

executing, by a processor, an instruction to perform a division operation for a dividend and a divisor, the execution comprising;

calculating a magnitude of the dividend and the divisor, the magnitude selected from one of a number of significant bits and a value represented in an exponent field;

comparing the magnitude of the dividend to the magnitude of the divisor;

prior to performing the division operation, determining a quotient precision based on the comparison, the determining including at least one of:

determining whether a special condition exists for which a required number of iterations is zero, and

calculating a difference between the magnitude of the dividend and the magnitude of the divisor;

computing a required number of iterations needed to produce the quotient precision; and

performing the required number of iterations.

2. The method of claim 1 , wherein each iteration results in an estimated value of the quotient, and each successive iteration results in a successive estimated value having a greater precision than a previous estimated value.

3. The method of claim 1 , wherein the dividend magnitude is a first number of significant bits in the dividend, and the divisor magnitude is a second number of significant bits in the divisor.

4. The method of claim 1 , wherein the quotient precision is a number of significant bits in the quotient.

5. The method of claim 1 , further comprising converting the dividend and the divisor into a floating point format.

6. The method of claim 5 , wherein the dividend magnitude is a first value represented in an exponent field of the dividend, and the divisor magnitude is a second value represented in an exponent field of the divisor.

7. The method of claim 1 , wherein comparing the magnitude of the dividend to the magnitude of the divisor comprises determining a difference between the magnitudes of the dividend and the divisor.

8. The method of claim 1 , wherein:

the special condition is at least one of: the divisor having a value of zero, the dividend having a value of zero, the divisor being greater than the dividend, and the quotient precision being greater than a maximum quotient precision allowed by the operation.

9. The method of claim 1 , wherein the special condition includes the divisor having a value of one and the quotient is equal to the dividend.

10. A computer program product for controlling a fixed point division operation, the computer program product comprising:

a computer-readable storage medium for storing instructions for executing a method of:

executing, by a processor, an instruction to perform a division operation for a dividend and a divisor, the execution comprising;

calculating a magnitude of the dividend and the divisor, the magnitude selected from one of a number of significant bits and a value represented in an exponent field;

comparing the magnitude of the dividend to the magnitude of the divisor;

prior to performing the division operation, determining a quotient precision based on the comparison, the determining including at least one of:

determining whether a special condition exists for which a required number of iterations is zero, and

calculating a difference between the magnitude of the dividend and the magnitude of the divisor;

computing a required number of iterations needed to produce the quotient precision; and

performing the required number of iterations.

11. The computer program product of claim 10 , wherein each iteration results in an estimated value of the quotient, and each successive iteration results in a successive estimated value having a greater precision than a previous estimated value.

12. The computer program product of claim 10 , wherein the dividend magnitude is a first number of significant bits in the dividend, and the divisor magnitude is a second number of significant bits in the divisor.

13. The computer program product of claim 10 , wherein the quotient precision is a number of significant bits in the quotient.

14. The computer program product of claim 10 , further comprising converting the dividend and the divisor into a floating point format.

15. The computer program product of claim 14 , wherein the dividend magnitude is a first value represented in an exponent field of the dividend, and the divisor magnitude is a second value represented in an exponent field of the divisor.

16. The computer program product of claim 10 , wherein the special condition includes at least one of: the divisor having a value of zero, the dividend having a value of zero, the divisor being greater than the dividend, and the quotient precision being greater than a maximum quotient precision allowed by the operation, wherein the required number of iterations is zero.

17. The computer program product of claim 10 , wherein the special condition includes the divisor having a value of one and the quotient is equal to the dividend.

18. A system for controlling a fixed point division operation, the system comprising:

an instruction dispatching unit (IDU) for sending an instruction to perform a division operation for a dividend and a divisor; and

a processor in operable communication with the IDU, the processor performing:

executing the instruction from the IDU, the execution comprising;

calculating a magnitude of the dividend and the divisor, the magnitude selected from one of a number of significant bits and a value represented in an exponent field;

comparing the magnitude of the dividend to the magnitude of the divisor;

prior to performing the division operation, determining a quotient precision based on the comparison, the determining including at least one of:

determining whether a special condition exists for which a required number of iterations is zero, and

calculating a difference between the magnitude of the dividend and the magnitude of the divisor;

computing a required number of iterations needed to produce the quotient precision; and

performing the required number of iterations.