备注

AI Translation Notice

This document was automatically translated by Qwen/Qwen3-8B model, for reference only.

  • Source document: kernel/debug/profiling-kernel-with-dadk.md

  • Translation time: 2025-05-19 01:41:49

  • Translation model: Qwen/Qwen3-8B

Please report issues via Community Channel

Performance Analysis of the Kernel Using DADK

1. Overview

This document will teach you how to use DADK to perform performance analysis on the DragonOS kernel, in order to identify and resolve potential performance bottlenecks.

1.1 Preparation

备注

Before you start, please ensure that you have installed DADK and have set up the compilation environment for the DragonOS kernel.

1.2 What is a Flame Graph?

If you haven’t heard of flame graphs before, you can read this article: How to Read Flame Graphs? - Ruanyifeng

In simple terms, a flame graph is an SVG image generated from performance sampling results, used to display the call stack of the CPU.

The x-axis represents the number of samples. If a function occupies a wider width on the x-axis, it means it was sampled more frequently, indicating longer execution time. Note that the x-axis does not represent time, but rather all call stacks are merged and sorted alphabetically.

A flame graph is used to identify which function at the top level occupies the largest width. If there is a “plateau” (flat area), it indicates that the function may have performance issues.

Colors have no special meaning, as flame graphs represent the CPU’s busy level, so warm tones are generally chosen.

2. Configuring the DragonOS Kernel

Since performance analysis requires detailed symbol table data, we need to configure the kernel compilation as follows:

In kernel/Cargo.toml’s [profile.release] section, set the following two options:

[profile.release]
debug = true
opt-level = 1

This will ensure that the compiled kernel includes symbol table data, making it easier for us to perform performance analysis.

3. Using DADK for Performance Analysis

3.1 Booting the Kernel

First, we need to boot the DragonOS kernel.

# 使用你喜欢的方式启动内核，例如：
make run
# 或者
make build && make qemu-nographic

3.2 Running Your Workload

After booting the kernel, we need to run some workloads in order to perform performance analysis.

This can be an application or something else. Even you can choose to do nothing and simply observe the call stack of the DragonOS kernel when it is idle.

3.3 Starting DADK for Performance Analysis

In the DragonOS project directory, run the following command:

dadk profile sample --format flamegraph  --output flame.svg --interval 200ms --duration 20s  --cpu-mask 0x1

The above command will perform performance analysis on the DragonOS kernel and generate a flame graph.

Detailed explanation:

  • --format flamegraph: Specifies the output format as a flame graph.

  • --output flame.svg: Specifies the output filename as flame.svg.

  • --interval 200ms: Specifies the sampling interval as 200ms.

  • --duration 20s: Specifies the sampling duration as 20 seconds.

  • --cpu-mask 0x1: Specifies the CPU to be sampled as CPU 0. (This is a bitmask, meaning if you want to sample CPU 0 and 1, the cpu-mask would be 0x3)

For more parameters, please refer to dadk profile sample --help.

备注

Since sampling will pause vCPU, the sampling time should not be too short, otherwise it may affect the normal operation of the system.

After waiting for a while, you will get a flame.svg file.

3.4 Analyzing the Flame Graph

Open the flame.svg file in a browser, and you will see a flame graph.

You can click on a function in the flame graph to view its call stack.

You can right-click the image below and open it in a new tab to experience the interactive effects.