Methods and apparatus for performing calculations using reduced-width data
View Patent ↗Methods, apparatus, and articles of manufacture for performing calculations using reduced-width data are disclosed. In particular, an example method determines reduced-width data values associated with generating and evaluating functions. Some of the reduced-width data values are stored within instructions in an instruction memory during a compile phase and retrieved from instruction memory during a runtime phase.
1. A method for approximating a target function, the method comprising:
identifying a set of coefficient values associated with a function;
generating a reduced-width coefficient value by reducing a data width of at least one of the coefficient values to have a first data width less than a second data width;
generating an updated function based on the reduced-width coefficient value;
storing the reduced-width coefficient value in a machine executable instruction; and
executing the reduced-width coefficient for a graphics application.
2. A method as defined in claim 1 , further comprising:
generating the function based on a target function, wherein generating the updated function includes:
updating the target function based on the reduced-width coefficient value; and
generating the updated function based on the updated target function.
3. A method as defined in claim 2 , further comprising:
comparing the updated function with the target function,
wherein storing the reduced-width coefficient value comprises storing the reduced-width coefficient value based on the comparison, and
wherein the function is an approximating polynomial that approximates the target function, the target function including at least one of a transcendental function and an algebraic function.
4. A method as defined in claim 1 , wherein the reduced-width coefficient value is associated with a highest degree term of the function.
5. A method as defined in claim 1 , wherein storing the reduced-width coefficient value in the machine executable instruction comprises storing the reduced-width coefficient value as an immediate value.
6. A method as defined in claim 1 , wherein storing the reduced-width coefficient value in the machine executable instruction comprises storing the reduced-width coefficient value in an instruction memory.
7. An apparatus for approximating a target function, the apparatus comprising:
a processor system including a memory; and
instructions stored in the memory that enable the processor system to:
identify a set of coefficient values associated with a function;
generate a reduced-width coefficient value by reducing a data width of at least one of the coefficient values to have a first data width less than a second data width;
generate an updated function based on the reduced-width coefficient value;
store the reduced-width coefficient value in a machine executable instruction; and
execute the reduced-width coefficient value for a graphics application.
8. An apparatus as defined in claim 7 , wherein the function is a polynomial.
9. An apparatus as defined in claim 8 , wherein the polynomial is an approximating polynomial that approximates at least one of a transcendental function and an algebraic function.
10. An apparatus as defined in claim 8 , wherein the reduced-width coefficient value is associated with a highest degree term of the polynomial.
11. An apparatus as defined in claim 7 , wherein the instructions stored in the memory enable the processor system to store the reduced-width coefficient value in the machine executable instruction as an immediate value.
12. An apparatus method as defined in claim 7 , wherein the instructions stored in the memory enable the processor system to store the machine executable instruction in an instruction memory.
13. A machine accessible medium having instructions stored thereon that, when executed, cause a machine to:
identify a set of coefficient values associated with a function;
generate a reduced-width coefficient value by reducing a data width of at least one of the coefficient values to have a first data width less than a second data width;
generate an updated function based on the reduced-width coefficient value;
store the reduced-width coefficient value in a machine executable instruction; and
execute the reduced-width coefficient for a graphics application.
14. A machine accessible medium as defined in claim 13 , wherein the function is a polynomial.
15. A machine accessible medium as defined in claim 14 , wherein the set of coefficient values is associated with an approximating polynomial that approximates at least one of a transcendental function and an algebraic function.
16. A machine accessible medium as defined in claim 14 having instructions stored thereon that, when executed, cause the machine to generate the reduced-width coefficient value based on a highest-degree term of the polynomial.
17. A machine accessible medium as defined in claim 13 having instructions stored thereon that, when executed, cause the machine to store the reduced-width coefficient value in the machine executable instruction as an immediate value.
18. A machine accessible medium as defined in claim 13 having instructions stored thereon that, when executed, cause the machine to store the machine executable instruction in an instruction memory.
19. An apparatus for approximating a target function, the apparatus comprising:
a processor system including a flash memory; and
instructions stored in the flash memory that enable the processor system to:
identify a set of coefficient values associated with a function;
generate a reduced-width coefficient value by reducing a data width of at least one of the coefficient values to have a first data width less than a second data width;
generate an updated function based on the reduced-width coefficient value;
store the reduced-width coefficient value in a machine executable instruction; and
execute the reduced-width coefficient for a graphics application.
20. An apparatus as defined in claim 19 , wherein the function is an approximating polynomial that approximates at least one of a transcendental function and an algebraic function.
21. An apparatus for approximating a target function, the apparatus comprising:
a reduced-width data generator configured to generate at least one reduced-width data value from a set of coefficient values associated with a function;
a function selector in communication with the reduced-width data generator and configured to generate at least one updated function based on the at least one reduced-width data value; and
a comparing circuit in communication with the reduced-width data generator and the function selector, wherein the comparing circuit is configured to perform at least one comparison based on the at least one updated function.
22. An apparatus as defined in claim 21 , wherein the function includes an approximating polynomial.
23. An apparatus as defined in claim 22 , wherein the approximating polynomial is associated with a mixed width polynomial.
24. An apparatus as defined in claim 21 , wherein the at least one reduced- width data value is a short width coefficient value.
25. An apparatus as defined in claim 21 , wherein the at least one reduced- width data value is a long width coefficient value.
26. An apparatus as defined in claim 21 , further comprising a differentiator in communication with the function selector and configured to differentiate the function.
27. An apparatus for approximating a target function, the apparatus comprising:
processor means for processing data, wherein the processor means includes storage means for storing data; and
instruction means, stored in the storage means, for enabling the processor means to:
identify a set of coefficient values associated with a function;
generate a reduced-width coefficient value by reducing a data width of at least one of the coefficient values to have a first data width less than a second data width;
generate an updated function based on the reduced-width coefficient value;
store the reduced-width coefficient value in a machine executable instruction; and
execute the reduced-width coefficient for a graphics application.
28. An apparatus as defined in claim 27 , wherein the function is a polynomial.
29. An apparatus as defined in claim 28 , wherein the polynomial is an approximating polynomial that approximates at least one of a transcendental function and an algebraic function.
30. An apparatus as defined in claim 28 , wherein the reduced-width coefficient value is associated with a highest degree term of the polynomial.
31. An apparatus as defined in claim 27 , wherein the instruction means enable the processor means to store the reduced-width coefficient value in the machine executable instruction as an immediate value.
32. An apparatus method as defined in claim 27 , wherein the instruction means enable the processor means to store the machine executable instruction in an instruction memory.
33. An apparatus for approximating a target function, the apparatus comprising:
generator means for generating at least one reduced-width data value from a set of coefficient values associated with a function;
selector means, in communication with the generator means, for generating at least one updated function based on the at least one reduced-width data value; and
comparing circuitry, in communication with the generator means and the selector means, for performing at least one comparison based on the at least one updated function to determine accuracy of the at least one undated function relative to a maximum deviation.
34. An apparatus as defined in claim 33 , wherein the function includes an approximating polynomial.
35. An apparatus as defined in claim 34 , wherein the approximating polynomial is associated with a mixed width polynomial.
36. An apparatus as defined in claim 33 , wherein the at least one reduced-width data value is a short width coefficient value.
37. An apparatus as defined in claim 33 , wherein the at least one reduced-width data value is a long width coefficient value.
38. An apparatus as defined in claim 33 , further comprising differentiator means, in communication with the selector means, for differentiating the function.