Debugging and tuning method based on time-space characteristic data

Abstract

The invention discloses a debugging and tuning method based on time-space characteristic data. The debugging and tuning method comprises the following steps that S1, a sampling control module configures indexes needing to be collected and starts collection; s2, a performance index collection module collects performance indexes, time and code space where the current performance indexes are located are collected through a time collection module and a space collection module respectively, the time indexes are usually represented by the number of cycles of a processor, and the space indexes are usually represented by PC values of a program; s3, when running of the monitored code is finished, collecting is finished through the sampling control module, and time-space characteristic data are generated; s4, calling an adjusting and optimizing module to analyze the time-space characteristic data generated in the step S3, and analyzing the performance condition of the monitored code; and S5, calling a debugging module to analyze the time-space characteristic data generated in the step S3, and analyzing whether the program has an abnormal process or not. Performance monitoring and program debugging can be achieved at the same time, and a user can rapidly position program errors caused by optimization in the program tuning stage.

Description

technical field [0001] The invention relates to a debugging and tuning method based on spatiotemporal characteristic data, and belongs to the technical field of computer program debugging and tuning. Background technique [0002] The traditional debugging technology mainly uses hardware support to perform fine-grained step-by-step tracing. As the parallel scale of the program increases, the step-by-step tracing method faces huge challenges in many aspects such as data volume and tool stability. For this reason, the researchers proposed A light-weight debugging method is introduced, mainly by collecting some status data to realize the debugging of specific program problems. Tuning is also called performance debugging, which itself can be considered as a kind of debugging. The basic method is to locate performance problems by collecting program performance status data, which coincides with lightweight debugging methods, except that the collected data Slightly different. [0...

AI Technical Summary

Problems solved by technology

[0003] Although there have been researches on tuning using debug registers in recent years, generally speaking, most of the existing technologies study debugging and tuning as independent technologies, which may lead to the under-exploitation of some data values , which can also lead to underutilization of hardware resources like breakpoint registers and performance counters

Usually, the development process of the program is generally to perform performance tuning after the program function is debugged correctly, but in the performance tuning stage, the optimization strategy may be too aggressive, which may cause program errors and require re-entering the function debugging stage

Images