Communication feature model method and system for high-performance computing application

Abstract

The invention discloses a communication feature model method for a high-performance computing application. The communication feature model method comprises the following steps of step 1, extracting application communication features; step 2, analyzing communication characteristics; step 3, establishing a communication characteristic model; and step 4, carrying out extension analysis. The inventionfurther provides a communication feature model system for a high-performance computing application. The method and system have the following beneficial effects that the multi-level and multi-means data acquisition method based on the HPC comprehensive application environment is used for extracting the application communication characteristics, the multiple linear regression model is used for establishing the application communication characteristic model, and the communication behavior characteristics of typical high-performance applications under E-level calculation are accurately predicted;by establishing the communication feature model for the application of high-performance computing, the performance bottleneck of the program can be found out more quickly, the program is optimized efficiently, the computing efficiency is improved, and resources can be used more reasonably.

Description

【Technical field】 [0001] The invention relates to the technical field of computer communication, in particular to a communication feature model method and system for high-performance computing applications. 【Background technique】 [0002] High-performance computing systems use the aggregated computing power of a large number of processing units to meet the huge computing needs of applications. The key issue is to achieve large-scale integration and efficient collaborative computing of many computing nodes. The core technology involves high-performance computers and large-scale parallel applications. With the widespread application of high-performance computing, more and more users use parallel execution programs to increase the execution speed of large-scale programs and improve the overall execution performance; this is also accompanied by a message passing interface (Message Passing Interface, MPI) is widely used. The performance analysis tool mpiP analyzes the performance...

AI Technical Summary

Problems solved by technology

And this is just a global overview, and it cannot be more intuitively reflected on some data that has a large overhead and affects performance. At the same time, it is more difficult to analyze complex programs with multi-module functions.

[0014] With parallel computing, a large amount of tracking data and process data will be generated, and the report generated by calling the mpiP library will integrate these large amounts of data to facilitate subsequent performance analysis by users, but a large amount of data is not convenient for users to analyze separately, which is a problem for The performance analysis of the program caused some unnecessary difficulties, mpiP can only generate data reports, mpiP is a lightweight analysis library for MPI applications

Images