A task monitoring, tracking and identification method for multi-core architecture

Abstract

The invention provides a task monitoring, tracking and recognition method for a multi-core architecture. The method comprises the steps that (1) a Task Monitoring Model (TMM) is established, wherein the TMM comprises (1-1) a running model for a task mode, (1-2) a task monitoring mechanism and (1-3) display of task running and association information in a relation description graph, so that automatic recognition is realized; and (2) the TMM is applied to a multi-core programming model to realize a monitoring and tracking method based on a package library technology, including analysis of a taskexecution mechanism of a library during GCC OpenMP running, an OpenMP task monitoring mechanism based on simulated scheduling and an OpenMP task monitoring mechanism based on additional information.Through the method, a performance tool can monitor newly added execution dimensions, namely tasks; and necessary performance information is provided for performance analysis staff to recognize programbehaviors and discover performance bottlenecks, therefore, programs are optimized, and running efficiency is improved.

Description

technical field [0001] The invention relates to the field of performance monitoring and analysis in the field of information technology, in particular to a multi-core architecture-oriented task monitoring, tracking and identification method. Background technique [0002] With the rapid development of information technology, the computer has changed from a traditional single structure to a mixed structure composed of multi-core heterogeneous structures, and both software and hardware have undergone tremendous changes. However, the basic unit of current multi-core computing is the thread, and the thread is used to complete the parallel computing tasks. The creation and destruction of the thread generate a large overhead, and the program implementation structure is monotonous and not flexible enough. It is also unable to fully utilize the computing resources of the multi-core architecture. The scope of application of traditional parallel computing is mainly the operation of re...

AI Technical Summary

Problems solved by technology

[0003] (1) Most of the monitoring methods studied in the prior art use the method of directly inserting or sampling the source code. Due to their own shortcomings, these monitoring mechanisms affect the application range of the measurement and the accuracy of the results;

[0004] (2) The monitoring of tasks by the system usually focuses on the creation and execution of tasks on the basis of the original monitoring objects, but often ignores the complex parent-child or constraint dependencies between tasks, and the performance information is expressed in a single way. Usually, elements of the task are added to the original description diagram, and there is a lack of detailed expression of the task behavior and relationship, which play an important role in the in-depth understanding of the program and task behavior;

[0005] (3) There is a lack of monitoring methods specifically for multi-core task modes and applicable to most compilation environments and runtime systems, and these monitoring methods play a guiding role in the research of performance monitoring

Images