Rounding injection scheme for floating-point to integer conversion

1. An apparatus comprising:

a memory configured to store a floating point number formatted in a floating point format;

a normalization engine configured to normalize at least a portion of the floating point number to create a normalized number;

a lookup table configured to generate an injection constant based upon a rounding mode for converting a floating-point number to an integer number; and

an adder configured to create an integer result by adding the normalized number and the injection constant.

2. The apparatus of claim 1 , wherein the adder is configured to only add the normalized number and the injection constant.

3. The apparatus of claim 1 , wherein the injection constant comprises a round injection bit, a sticky injection bit, and a carry-in value.

4. The apparatus of claim 1 , wherein the injection constant comprises:

a carry in bit equal to a sign bit of the floating point number; and

if instructed to round to zero, a round injection bit and a sticky injection bit both equal to the sign bit of the floating point number,

if instructed to round to minus infinity, a round injection bit and a sticky injection bit both equal to zero,

if instructed to round to positive infinity, a round injection bit and a sticky injection bit both equal to one, and

if instructed to round to a nearest even value, a round injection bit equal to one and a sticky injection bit equal to zero.

5. The apparatus of claim 1 , wherein the normalization engine is configured to, if the floating point number is negative, invert, via a one's complement, a normalized version of the floating point number.

6. The apparatus of claim 1 , further comprising:

a least-significant bit correction circuit configured to determine if a least-significant bit of the integer result is to be corrected, wherein the determination is based in part upon the predefined set of injection constants.

7. The apparatus of claim 1 , wherein the integer result comprises an integer value formatted in a two's complement format.

8. The apparatus of claim 1 , wherein the integer result comprises an integer value in an unsigned format.

9. A method comprising

receiving a floating point number formatted in a floating point format;

normalizing at least a portion of the floating point number to create a normalized number;

calculating at least one rounding value according to a rounding mode for converting a floating-point number to an integer number; and

adding the normalized number and the rounding value to create an integer result.

10. The method of claim 9 , wherein adding the normalized number and the rounding value to create an integer result comprises adding only the normalized number with the rounding value.

11. The method of claim 9 , wherein the at least one rounding value comprises a rounding injection bit, a sticky injection bit, and a carry-in value.

12. The method of claim 9 , wherein the at least one rounding value comprises:

a carry in bit equal to a sign bit of the floating point number; and

if instructed to round to zero, a round injection bit and a sticky injection bit both equal to the sign bit of the floating point number,

if instructed to round to minus infinity, a round injection bit and a sticky injection bit both equal to zero,

if instructed to round to positive infinity, a round injection bit and a sticky injection bit both equal to one, and

if instructed to round to a nearest even value, a round injection bit equal to one and a sticky injection bit equal to zero.

13. The method of claim 9 , wherein normalizing at least a portion of the floating point number comprises:

if the floating point number is negative, inverting, via a one's complement, a normalized version of the floating point number.

14. The method of claim 9 , further comprising:

determining if a least-significant bit of the integer result is to be corrected, wherein the determination is based in part upon the rounding mode, and

if so, correcting the least-significant bit of the integer result.

15. The method of claim 9 , wherein the integer result comprises an integer value formatted in a two's complement format.

16. The method of claim 9 , wherein the integer result comprises an integer value in an unsigned format.

17. A system comprising:

a memory configured to store a floating point number formatted in a floating point format; and

a floating-point conversion unit configured to convert a floating point number to an integer number and comprising:

a normalization engine configured to normalize at least a portion of the floating point number to create a normalized number,

a lookup table configured to generate an injection constant based upon a rounding mode for converting a floating-point number to an integer number, and

an adder configured to create an integer result by adding the normalized number and the injection constant.

18. The system of claim 17 , wherein the adder is configured to only add the normalized number and the injection constant.

19. The system of claim 17 , wherein the injection constant comprises:

a carry in bit equal to a sign bit of the floating point number; and

if instructed to round to zero, a round injection bit and a sticky injection bit both equal to the sign bit of the floating point number,

if instructed to round to minus infinity, a round injection bit and a sticky injection bit both equal to zero,

if instructed to round to positive infinity, a round injection bit and a sticky injection bit both equal to one, and

if instructed to round to a nearest even value, a round injection bit equal to one and a sticky injection bit equal to zero.

20. The system of claim 17 , wherein the integer result comprises an integer value formatted in a two's complement format.