WORKLOAD PLACEMENT BASED ON HETEROGENEOUS COMPUTE PERFORMANCE PER WATT

• US 20170277246A1
• Filed: 03/22/2016
• Published: 09/28/2017
• Est. Priority Date: 03/22/2016
• Status: Active Grant

• Current Assignee: Lenovo Global Technologies International Ltd. (Lenovo Group Ltd.)
• Original Assignee: Lenovo Enterprise Solutions (Singapore) Pte., Ltd. (Lenovo Group Ltd.)
• Inventors: NICHOLSON, JOHN W., VANOVER, MICHAEL T., ZHANG, JIANBANG, MATTESON, JASON A.
• Granted Patent: US 10,203,747 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 1/329 by task scheduling
• CPC Class Codes: G06F 1/3293 by switching to a less power-consuming processor, e.g. sub-CPU
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

An apparatus for selecting a function includes a comparison module that compares energy consumption characteristics of a plurality of processors available for execution of a function, where each energy consumption characteristic varies as a function of function size. The apparatus includes a selection module that selects, based on the size of the function, a processor from the plurality of processors with a lowest energy consumption for execution of the function. The apparatus includes an execution module that executes the function on the selected processor.

Claims

1. An apparatus comprising:

a comparison module that compares energy consumption characteristics of a plurality of processors available for execution of a function, wherein each energy consumption characteristic varies as a function of function size;

a selection module that selects, based on the size of the function, a processor from the plurality of processors with a lowest energy consumption for execution of the function; and

an execution module that executes the function on the selected processor.

2. The apparatus of claim 1, further comprising an efficiency module that determines the energy consumption characteristic of each of the plurality of processors.

3. The apparatus of claim 2, wherein the efficiency module determines the energy consumption characteristic of each of the plurality of processors based on one or more of:

information supplied with the function;

information from a vendor; and

function execution data from previously executed functions.

4. The apparatus of claim 3, further comprising:

an energy calculation module that calculates energy consumption of an executed function executed on a processor of the plurality of processors; and

a correlation module that correlates a size of each executed function to energy consumed during execution of the function to determine the energy consumption characteristic of the processor.

5. The apparatus of claim 3, further comprising an update module that updates an initial energy consumption characteristic of a processor based on the function execution data from the previously executed functions.

6. The apparatus of claim 1, further comprising a function size module that determines a size of the function.

7. The apparatus of claim 6, wherein the function size module determines the size of the function based on one or more of:

a size of input data of the function;

a number of operations to be performed in executing the function;

a number of times one or more functions are called during execution of the function; and

a number of items searched for during execution of the function.

8. The apparatus of claim 1, wherein the energy consumption characteristic of a processor of the plurality of processors comprises a startup cost and energy usage as a function of function size.

9. The apparatus of claim 8, wherein the startup cost comprises an amount of energy to bring the processor to a state ready to execute the function.

10. The apparatus of claim 8, further comprising a startup module that determines a current state of the processor and that adjusts the startup cost for the processor based the current condition of the processor.

11. The apparatus of claim 1, further comprising:

an intercept module that intercepts a function call from an executing application, the function call specifying a target function, the target function associated with a target processor, wherein the comparison module compares energy consumption characteristics of the plurality of processors in response to the intercept module intercepting the function call,

wherein the selection module selects a function from a plurality of available functions to execute the function call, each function associated with a separate processor, wherein the selection module selects the function associated with the processor with a lowest energy consumption for execution of the function call, and

wherein the execution module reroutes the function call to execute the selected function.

12. The apparatus of claim 1, wherein the processors associated with the plurality of libraries comprise one or more of a field-programmable gate array (“

FPGA”

), a graphics processing unit (“

GPU”

), an accelerated processing unit (“

APU”

), a central processing unit (“

CPU”

), a digital signal processor (“

DSP”

) and a processor of a computing device different from a computing device executing the application.

13. The apparatus of claim 1, further comprising a computing device comprising at least a processor and a memory that stores said modules for execution on the processor and one or more alternate processors.

14. A method comprising:

comparing energy consumption characteristics of a plurality of processors available for execution of a function, wherein each energy consumption characteristic varies as a function of function size;

selecting, based on the size of the function, a processor from the plurality of processors with a lowest energy consumption for execution of the function; and

executing the function on the selected processor.

15. The method of claim 14, further comprising determining the energy consumption characteristic of each of the plurality of processors.

16. The method of claim 15, wherein determining the energy consumption characteristic is based on one or more of:

information supplied with the function;

information from a vendor associated with the processor; and

function execution data from previously executed functions.

17. The method of claim 16, further comprising:

calculating energy consumption of an executed function executed on a processor of the plurality of processors;

correlating a size of each executed function to energy consumed during execution of the function to determine the energy consumption characteristic of the processor; and

updating an initial energy consumption characteristic of a processor based on the function execution data from the previously executed functions.

18. The method of claim 14, wherein the energy consumption characteristic of a processor of the plurality of processors comprises a startup cost and energy usage as a function of function size, the startup cost comprising an amount of energy to bring the processor to a state ready to execute the function, and further comprising determining a current state of the processor and adjusting the startup cost for the processor based the current condition of the processor.

19. A program product comprising a computer readable storage medium that stores code executable by a processor, the executable code comprising code to perform:

comparing energy consumption characteristics of a plurality of processors available for execution of a function, wherein each energy consumption characteristic varies as a function of function size;

selecting, based on the size of the function, a processor from the plurality of processors with a lowest energy consumption for execution of the function; and

executing the function on the selected processor.

20. The program product of claim 19, the code further comprising:

determining the energy consumption characteristic of each of the plurality of processors, wherein determining the energy consumption characteristic is based on one or more of information supplied with the function, information from a vendor associated with the processor, and function execution data from previously executed functions;

calculating energy consumption of an executed function executed on a processor of the plurality of processors;

correlating a size of each executed function to energy consumed during execution of the function to determine the energy consumption characteristic of the processor; and

updating an initial energy consumption characteristic of a processor based on the function execution data from the previously executed functions, wherein the energy consumption characteristic of a processor of the plurality of processors comprises a startup cost and energy usage as a function of function size, the startup cost comprising an amount of energy to bring the processor to a state ready to execute the function.