System and method of dynamic precision operations

• US 8,880,573 B2
• Filed: 10/11/2011
• Issued: 11/04/2014
• Est. Priority Date: 10/12/2010
• Status: Active Grant

• Current Assignee: Dolby Laboratories Licensing Corporation (Dolby Laboratories Incorporated)
• Original Assignee: Dolby Laboratories Licensing Corporation (Dolby Laboratories Incorporated)
• Inventors: Erinjippurath, Gopal, Smith, Jared D.
• Primary Examiner(s): Mai, Tan V.
• Application Number:
  US13/271,095
• Publication Number:
  US 20120089655A1
• Time in Patent Office: 1,120 Days
• Field of Search: 708/204
• US Class Current: 708/204
• CPC Class Codes: G06F 7/38 Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation
• CPC Class Codes: G06F 8/447 Target code generation

Abstract

In an embodiment, a method performs computer operations using a first fractional precision and a second fractional precision. A computer program has a source variable, a destination variable, and an operation. The source variable has a first dynamic fractional precision, the destination variable has a second dynamic fractional precision that differs from the first dynamic fractional precision, and the operation is related to the source variable and the destination variable. The source variable is aligned to a format of the destination variable, according to the first dynamic fractional precision and the second dynamic fractional precision. The operation is performed using the destination variable and the source variable. A value is assigned to the destination variable according to the operation. In this manner, a single codebase may be written that operates on various hardware that each have different bit precision capabilities, without requiring additional development and verification effort.

Claims

1. A computer implemented method of performing computer operations using a first fractional precision and a second fractional precision, comprising:

accessing with the computer a source variable, a destination variable, and an operation, wherein the source variable has a first dynamic fractional precision, wherein the destination variable has a second dynamic fractional precision that differs from the first dynamic fractional precision, and wherein the operation is related to the source variable and the destination variable;

determining a difference between the first dynamic fractional precision and the second dynamic fractional precision;

converting the source variable to be of the second dynamic fractional precision of the destination variable, according to the difference between the first dynamic fractional precision and the second dynamic fractional precision;

performing the operation using the destination variable and the source variable; and

assigning a value to the destination variable according to the operation.

2. The method of claim 1, wherein the source variable has a dynamic precision that includes an integer part and a fractional part, wherein the fractional part corresponds to the first dynamic fractional precision.

3. The method of claim 1, wherein the destination variable has a dynamic precision that includes an integer part and a fractional part, wherein the fractional part corresponds to the second dynamic fractional precision.

4. The method of claim 1, wherein the source variable has a maximum number of bits, a dynamic integer part and a dynamic fractional part, and wherein the dynamic integer part and the dynamic fractional part are within the maximum number of bits.

5. The method of claim 1, wherein the destination variable has a maximum number of bits, a dynamic integer part and a dynamic fractional part, and wherein the dynamic integer part and the dynamic fractional part are within the maximum number of bits.

6. The method of claim 1, wherein the operation comprises an assignment operation, wherein the source variable has a source value and a first dynamic integer precision, wherein the destination variable has a second dynamic integer precision, wherein performing the operation comprises adjusting the source value according to the first dynamic integer precision and the second dynamic integer precision, and wherein assigning the value comprises assigning the source value having been adjusted to the destination variable.

7. The method of claim 1, wherein the operation comprises a compound addition operation, wherein the source variable has a source value, wherein the destination variable has a dynamic integer precision, wherein performing the operation comprises:

adding an initial value and the source value to result in a destination value; and

adjusting the destination value according to the dynamic integer precision,

wherein assigning the value comprises assigning the destination value having been adjusted to the destination variable.

8. The method of claim 1, wherein the operation comprises an addition operation, wherein the source variable comprises a first source variable having a first source value, wherein a second source variable has a second source value and a third dynamic fractional precision, wherein the destination variable has a dynamic integer precision, wherein converting comprises:

formatting a temporary variable according to the first dynamic fractional precision and the third dynamic fractional precision;

assigning the second source value to the temporary variable as a temporary value; and

aligning the first source variable to the format of the temporary variable,

wherein performing the operation comprises;

adding the temporary value and the first source value to result in a destination value; and

adjusting the destination value according to the dynamic integer precision, and

wherein assigning the value comprises assigning the destination value having been adjusted to the destination variable.

9. The method of claim 1, wherein the operation comprises a compound multiplication operation, wherein the source variable has a source value, wherein the destination variable has a destination value, wherein performing the operation comprises:

multiplying the destination value and the source value to result in a temporary value;

right-shifting the temporary value according to the first dynamic fractional precision; and

rounding the temporary value according to a value shifted out according to the temporary value having been right-shifted,

wherein assigning the value comprises assigning the temporary value having been rounded to the destination variable.

10. The method of claim 1, wherein the operation comprises a multiplication operation, wherein the source variable comprises a first source variable having a first source value and a first dynamic integer precision, wherein a second source variable has a second source value, a third dynamic fractional precision, and a second dynamic integer precision, and wherein converting comprises:

formatting a temporary variable according to the first dynamic fractional precision, the third dynamic fractional precision, the first dynamic integer precision, the second dynamic integer precision, and a maximum number of bits;

wherein performing the operation comprises multiplying the first source value and the second source value to result in a temporary value, and

wherein assigning the value comprises assigning the temporary value to the destination variable.

11. The method of claim 1, wherein the converting comprises:

adjusting the source variable according to the difference between the first dynamic fractional precision and the second dynamic fractional precision.

12. The method of claim 1, wherein, when the first dynamic fractional precision is less than the second dynamic fractional precision, the converting comprises:

increasing a value of the source variable.

13. The method of claim 1, wherein, when the second dynamic fractional precision is less than the first dynamic fractional precision, the converting comprises:

decreasing a value of the of the source variable; and

reducing the first dynamic fractional precision to match the second dynamic fractional precision.

14. The method of claim 1, wherein, when the second dynamic fractional precision is less than the first dynamic fractional precision, the converting comprises:

decreasing a value of the of the source variable;

reducing the first dynamic fractional precision to match the second dynamic fractional precision; and

adjusting the value of the source variable according to a specified rounding mode.

15. The method of claim 1, further comprising:

receiving information indicating a target architecture; and

compiling the computer program according to the information indicating the target architecture.

16. The method of claim 1, wherein converting the source variable is implemented according to a class definition written in an object-oriented programming language.

17. The method of claim 1, wherein performing the operation is implemented according to a class definition written in an object-oriented programming language.

18. The method of claim 1, wherein converting the source variable and performing the operation are implemented according to a class definition in an object-oriented programming language.

19. An apparatus for performing computer operations using a first fractional precision and a second fractional precision, comprising:

a memory device that is configured to store a computer program having a source variable, a destination variable, and an operation, wherein the source variable has a first dynamic fractional precision, wherein the destination variable has a second dynamic fractional precision that differs from the first dynamic fractional precision, and wherein the operation is related to the source variable and the destination variable; and

a processor that is configured to determine a difference between the first dynamic fractional precision and the second dynamic fractional precision, convert the source variable to be of the second dynamic fractional precision of the destination variable, according to the difference between the first dynamic fractional precision and the second dynamic fractional precision, that is configured to perform the operation using the destination variable and the source variable, and that is configured to assign a value to the destination variable according to the operation.

20. A non-transitory computer readable medium, comprising software instructions that when executed by one or more processors cause the one or more processors to:

read a source variable, a destination variable, and an operation, wherein the source variable has a first dynamic fractional precision, wherein the destination variable has a second dynamic fractional precision that differs from the first dynamic fractional precision, and wherein the operation is related to the source variable and the destination variable;

determine a difference between the first dynamic fractional precision and the second dynamic fractional precision;

convert the source variable to be of the second dynamic fractional precision of the destination variable, according to the difference between the first dynamic fractional precision and the second dynamic fractional precision;

perform the operation using the destination variable and the source variable; and

assign a value to the destination variable according to the operation.