System and Method for Energy Profiling Asynchronous API Calls

• US 20190324513A1
• Filed: 04/23/2019
• Published: 10/24/2019
• Est. Priority Date: 04/23/2018
• Status: Active Grant

• Current Assignee: Purdue Research Foundation (The Trustees of Purdue University (d/b/a Purdue University))
• Original Assignee: Purdue Research Foundation (The Trustees of Purdue University (d/b/a Purdue University))
• Inventors: Hu, Yu Charlie, Ding, Ning
• Granted Patent: US 11,107,179 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 1/28 Supervision thereof, e.g. detecting power-supply failure by out of limits supervision
• CPC Class Codes: G06F 11/302 where the computing system component is a software system
• CPC Class Codes: G06F 11/3024 where the computing system component is a central processing unit [CPU]
• CPC Class Codes: G06F 11/3058 Monitoring arrangements for monitoring environmental properties or parameters of the computing system or of the computing system component, e.g. monitoring of power, currents, temperature, humidity, position, vibrations thermal management in cooling arrangements of a computing system G06F1/206
• CPC Class Codes: G06F 11/3062 where the monitored property is the power consumption power management in a computing system G06F1/3203
• CPC Class Codes: G06F 2201/865 Monitoring of software
• CPC Class Codes: G06F 9/451 Execution arrangements for user interfaces
• CPC Class Codes: G06T 1/20 Processor architectures; Processor configuration, e.g. pipelining
• CPC Class Codes: G06T 2200/24 involving graphical user interfaces [GUIs]

Abstract

A method for profiling energy usage of invoking an application programming interface (API) by an application in a computing device. The method includes obtaining source code for the API, modifying each class by adding a callback function identifier field that is initiated to a unique value upon instantiation of each object that belongs to the class, identifying each location in the source code that posts the callback function for asynchronous execution by enqueueing each object instantiated from the class containing the callback function into the system callback queue, modifying the source code to log the callback function identifier of object at the location that dequeues objects from the system callback queue, modifying source code by adding two system logging function calls to log the callback function identifier of the dequeued object before and after executing the callback function, executing the application, and performing energy accounting of the asynchronous API calls.

Claims

1. A method for profiling energy usage of invoking an application programming interface (API) by an application in a computing device, comprising:

obtaining source code “

A”

for the API;

identifying each class C_i in the A for the API that contains a callback function D_i;

modifying each C_i by adding a callback function identifier field that is initiated to a unique value V_j upon instantiation of each object O_j that belongs to class C_i;

identifying each location E_k in the A that posts the callback function D_i for asynchronous execution by enqueueing object O_j instantiated from the C_i containing the callback function D_i into the system callback queue;

modifying the A at location E_k by adding a logging function call L to log the callback function identifier V_j and the current call stack into the energy profiling log;

obtaining source code B in the operating system/framework that asynchronously executes any D passed from the API;

identifying each location F in the B that invokes the callback function associated with any object dequeued from the system callback queue;

modifying the B at the F by adding logging function calls L before and after the callback function invocation to log the callback function identifier V_j of the dequeued object when we start and finish executing the callback function respectively to the energy profiling log;

executing the application thereby generating one or two threads, wherein the first thread executes the API call that posts the callback function D_i and wherein the second thread which can be the same as the first thread dequeues the callback function D_i and invokes the callback function D_i;

determining the caller-caller relationship between the function in the API execution stack that posted each D_i and the asynchronously invoked callback function by processing the logged callback function ID information and call stack information in the energy profiling log;

determining the energy usage of invoking the API call in thread 1, by monitoring activities of a processor, the processor operating any component that consumes power in the computing device in response to execution of the API;

determining the energy usage of invoking each callback function in thread 2, by monitoring activities of a processor, the processor operating any component that consumes power in the computing device in response to execution of the callback function;

adding the energy usage of invoking every callback function in thread 2 to the energy usage of invoking the API thus determined; and

attributing the combined energy to the API invocation.