System and method for dynamically reducing power consumption of floating-point logic

• US 9,268,528 B2
• Filed: 05/23/2013
• Issued: 02/23/2016
• Est. Priority Date: 05/23/2013
• Status: Active Grant

• Current Assignee: NVIDIA Corporation
• Original Assignee: NVIDIA Corporation
• Inventors: Tannenbaum, David C., Iyer, Srinivasan
• Primary Examiner(s): Ngo, Chuong D
• Application Number:
  US13/901,492
• Publication Number:
  US 20140351308A1
• Time in Patent Office: 1,006 Days
• Field of Search: 708/495
• US Class Current: 1/1
• CPC Class Codes: G06F 1/3243 Power saving in microcontroller unit
• CPC Class Codes: G06F 1/3287 by switching off individual functional units in the computer system
• CPC Class Codes: G06F 2207/3816 Accepting numbers of variable word length
• CPC Class Codes: G06F 7/483 Computations with numbers represented by a non-linear combination of denominational numbers, e.g. rational numbers, logarithmic number system or floating-point numbers G06F7/4806, G06F7/4824, G06F7/49, G06F7/491, G06F7/544 take precedence
• CPC Class Codes: G06F 7/505 in bit-parallel fashion, i.e. having a different digit-handling circuit for each denomination
• CPC Class Codes: G06F 7/523 Multiplying only
• CPC Class Codes: G06F 7/5312 using carry save adders
• CPC Class Codes: G06F 7/5443 Sum of products for applications thereof, see the relevant places, e.g. G06F17/10, H03H17/00
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A system and method are provided for dynamically reducing power consumption of floating-point logic. A disable control signal that is based on a characteristic of a floating-point format input operand is received and a portion of a logic circuit is disabled based on the disable control signal. The logic circuit processes the floating-point format input operand to generate an output.

Claims

1. A method, comprising:

computing a first number of trailing zeros in a first significand associated with a floating-point format input operand;

computing a second number of trailing zeros in a second significand;

summing the first number of trailing zeros with the second number of trailing zeros to produce a trailing zero sum;

receiving a disable control signal that is based on a characteristic of the floating-point format input operand and the trailing zero sum;

disabling a portion of a logic circuit based on the disable control signal; and

processing the first significand and the second significand by the logic circuit to generate an output.

2. The method of claim 1, wherein the disabling of the portion of the logic circuit comprises disabling one or more bits of a register configured to store the first significand.

3. The method of claim 1, wherein the disabling of the portion of the logic circuit comprises disabling propagation of a carry signal.

4. The method of claim 1, wherein the characteristic of the floating-point format input operand is associated with a bit accuracy specified for a computation performed by the logic circuit.

5. The method of claim 1, wherein the characteristic of the floating-point format input operand is associated with a number of bits needed to represent a constant value.

6. The method of claim 1, wherein the characteristic of the floating-point format input operand is a permitted error associated with the output.

7. The method of claim 6, wherein the permitted error corresponds to a Newton-Raphson known error bound for a floating-point computation.

8. The method of claim 1, wherein the disabling comprises decoupling a first portion of the logic circuit from a second portion of the logic circuit.

9. A processing unit, comprising:

a logic disable unit that is configured to;

compute a first number of trailing zeros in a first significand associated with a floating-point format input operand;

compute a second number of trailing zeros in a second significand; and

sum the first number of trailing zeros with the second number of trailing zeros to produce a trailing zero sum; and

arithmetic logic circuitry configured to;

receive a disable control signal that is based on a characteristic of the floating-point format input operand and the trailing zero sum;

disable a portion of the arithmetic logic circuitry based on the disable control signal; and

process the first significand and the second significand to generate an output.

10. The processing unit of claim 9, further comprising a register configured to store the first significand, wherein the disabling of the portion of the logic circuit comprises disabling one or more bits of the register.

11. The processing unit of claim 9, wherein the disabling of the portion of the arithmetic logic circuitry comprises disabling propagation of a carry signal.

12. The processing unit of claim 9, wherein the characteristic of the floating-point format input operand is associated with a bit accuracy specified for a computation performed by the arithmetic logic circuitry.

13. The processing unit of claim 9, wherein the characteristic of the floating-point format input operand is associated with a number of bits needed to represent a constant value.

14. The processing unit of claim 9, wherein the characteristic of the floating-point format input operand is a permitted error associated with the output.

15. The processing unit of claim 14, wherein the permitted error corresponds to a Newton-Raphson known error bound for a floating-point computation.

16. The processing unit of claim 9, wherein the disabling comprises decoupling a first portion of the arithmetic logic circuitry from a second portion of the arithmetic logic circuitry.

17. A system, comprising:

a memory; and

a processing unit, comprising;

a logic disable unit that is configured to;

compute a first number of trailing zeros in a first significand associated with a floating-point format input operand;

compute a second number of trailing zeros in a second significand; and

sum the first number of trailing zeros with the second number of trailing zeros to produce a trailing zero sum; and

arithmetic logic circuitry configured to;

receive a disable control signal that is based on a characteristic of the floating-point format input operand and the trailing zero sum;

disable a portion of the arithmetic logic circuitry based on the disable control signal; and

process the first significand and the second significand to generate an output.

18. The system of claim 17, wherein the characteristic of the floating-point format input operand is a permitted error associated with the output.

19. The system of claim 18, wherein the permitted error corresponds to a Newton-Raphson known error bound for a floating-point computation.

20. The system of claim 17, wherein the disabling comprises decoupling a first portion of the arithmetic logic circuitry from a second portion of the arithmetic logic circuitry.