Calculating a floating-point function using multiple lookup tables
A computation unit includes input lines to provide a floating-point value, a first lookup table, a second lookup table, a range detector, and an output stage. The input lines include exponent lines and mantissa lines. The first lookup table has a first address input coupled to a first subset of the input lines to provide a first output. The second lookup table has a second address input coupled to a second subset of the input lines to provide a second output. The range detector is coupled to at least some of the input lines and indicates whether the floating-point value provided on the input lines is within a specified range on a range output. The output stage is operatively coupled to the first output, the second output and the range output, to generate a function output based on the first output, the second output, and the range output.
1. A computation unit, comprising:
input lines to provide a floating-point value, the input lines including exponent lines and mantissa lines;
a first lookup table having a first address input coupled to a first subset of the input lines and providing a first output, the first subset of the input lines including a most significant bit of the exponent lines;
a second lookup table having a second address input coupled to a second subset of the input lines and providing a second output, the second subset of the input lines excluding the most significant bit of the exponent lines;
a range detector, coupled to at least some of the input lines, to indicate, on a range output, whether the floating-point value provided on the input lines is within a specified range; and
an output stage, operatively coupled to the first output, the second output and the range output, to generate a function output based on the first output, the second output, and the range output.
2. The computation unit of claim 1 , wherein the second subset of the input lines includes a subset of the exponent lines and a subset of the mantissa lines, and the second subset of the input lines has more of the mantissa lines than the first subset of the input lines.
3. The computation unit of claim 1 , the output stage to select the function output to be either the first output or the second output dependent upon the range output.
4. The computation unit of claim 1 , further comprising:
a third lookup table having a third address input coupled to the first subset of the input lines and providing a third output operatively coupled to the output stage;
an arithmetic circuitry to generate a fourth output based on the input lines, the first output, and the third output; and
a selector to select the function output to be either the fourth output or the second output dependent upon the range output.
5. The computation unit of claim 4 , the arithmetic circuitry to multiply the floating-point value provided on the input lines by a first value provided on the first output and add a second value provided on the third output to generate the fourth output.
6. The computation unit of claim 1 , further comprising a third lookup table having a third address input coupled to a third subset of the input lines and providing a third output operatively coupled to the output stage;
the range detector to further indicate, on the range output, whether the floating-point value provided on the input lines is within a second specified range; and
the output stage to select the function output to be either the first output, the second output, or the third output, dependent upon the range output.
7. The computation unit of claim 6 , wherein the second subset of the input lines and the third subset of the input lines are identical.
8. The computation unit of claim 1 , wherein the first output and the second output each have a number of lines equal to that of the input lines.
9. The computation unit of claim 8 , wherein the number of lines in the first output, the second output, and the input lines are each 32, the first subset of the input lines and the second subset of the input lines have 16 lines each, and the first lookup table and the second lookup table each have 2 16 (32768) entries.
10. The computation unit of claim 1 , wherein both the first subset of the input lines and the second subset of the input lines include a least significant line of the exponent lines and a most significant line of the mantissa lines.
11. The computation unit of claim 1 , the output stage comprising an interpolator.
12. The computation unit of claim 1 , comprising a reconfigurable data processor (RDP) that includes at least one configurable memory unit storing the first lookup table and the second lookup table, and at least one configurable computing unit implementing the range detector and the output stage.
13. A method to calculate a function of a floating-point value using a circuit, the method comprising:
receiving the floating-point value on input lines of the circuit, including exponent lines and mantissa lines;
using a first subset of the input lines, including a most significant bit of the exponent lines, as a first address input into a first lookup table of the circuit to retrieve a first value;
using a second subset of the input lines, excluding the most significant bit of the exponent lines, as a second address input into a second lookup table of the circuit to retrieve a second value;
determining, by the circuit, whether the floating-point value received on the input lines is within a specified range to generate a range output; and
generating a function output of the circuit based on the first value, the second value, and the range output.
14. The method of claim 13 , wherein the second subset of the input lines includes a subset of the exponent lines and a subset of the mantissa lines, and the second subset of the input lines has more of the mantissa lines than the first subset of the input lines.
15. The method of claim 13 , further comprising generating the function output by selecting either the first value or the second value dependent upon the range output.
16. The method of claim 13 , further comprising:
using the first subset of the input lines as a third address input into a third lookup table to retrieve a third value;
generating the function output at least in part by calculating a mathematical function using the floating-point value on the input lines, the first value, and the third value.
17. The method of claim 16 , further comprising generating the function output by selecting either the calculated mathematical function or the second value dependent upon the range output.
18. The method of claim 16 , wherein the mathematical function multiplies the floating-point value on the input lines by the first value and adds the third value.
19. The method of claim 13 , further comprising:
using a third subset of the input lines as a third address input into a third lookup table to retrieve a third value;
determining whether the floating-point value received on the input lines is within a second specified range to generate the range output; and
selecting the function output to be either the first value, the second value, or the third value, dependent upon the range output.
20. The method of claim 19 , wherein the second subset of the input lines and the third subset of the input lines are identical.
21. A computer program product comprising a machine readable, non-transitory data storage medium having stored thereon, one or more configuration files for a reconfigurable data processor (RDP), that when loaded onto the RDP cause one or more configurable units in the RDP to:
receive a floating-point value on input lines including exponent lines and mantissa lines;
use a first subset of the input lines, including a most significant bit of the exponent lines, as a first address input into a first lookup table to retrieve a first value;
use a second subset of the input lines, excluding the most significant bit of the exponent lines, as a second address input into a second lookup table to retrieve a second value;
determine whether the floating-point value received on the input lines is within a specified range to generate a range output; and
generate a function output based on the first value, the second value, and the range output.
22. The computer program product of claim 21 , wherein the one or more configuration files further cause the one or more configurable units in the RDP to generate the function output by selecting either the first value or the second value dependent upon the range output.
23. The computer program product of claim 21 , wherein the one or more configuration files further cause the one or more configurable units in the RDP to:
use the first subset of the input lines as a third address input into a third lookup table to retrieve a third value;
generate the function output at least in part by calculating a mathematical function using the floating-point value on the input lines, the first value, and the third value.
24. The computer program product of claim 23 , wherein the one or more configuration files further cause the one or more configurable units in the RDP to generate the function output by selecting either the calculated mathematical function or the second value dependent upon the range output.
25. The computer program product of claim 23 , wherein the mathematical function multiplies the floating-point value on the input lines by the first value and adds the third value.
26. The computer program product of claim 21 , wherein the one or more configuration files further cause the one or more configurable units in the RDP to:
use a third subset of the input lines as a third address input into a third lookup table to retrieve a third value;
determine whether the floating-point value received on the input lines is within a second specified range to generate the range output; and
select the function output to be either the first value, the second value, or the third value, dependent upon the range output.
27. The computer program product of claim 26 , wherein the second subset of the input lines and the third subset of the input lines are identical.