CGraph

Chinese | English Readme

CGraph documentation

CGraph is a cross-platform D irected A cyclic G raph framework based on pure C++ without any 3rd-party dependencies.

You, with it, can build your own operators simply, and describe any running schedules as you need, such as dependence, parallelling, aggregation and so on. Some useful tools and plugins are also provided to improve your project.

Tutorials and contact information are show as follows. Please get in touch with us for free if you need more about this repository.

1. Introduction

CGraph is called [Color, Picture] in Chinese, and is a cross-platform graph process execution framework without any third-party dependencies. Through GPipeline (pipeline) underlying scheduling, it provides the function of sequential execution of dependent elements and concurrent execution of non-dependent elements in eDAG scheduling.

Users only need to inherit GNode (node) class, implement the run() method of the subclass, and set dependencies as needed to realize graphical execution or pipeline execution of the task. It is also possible to set various GGroup (groups) containing multi-node information to control the conditional judgment, looping and concurrent execution logic of the graph.

The project provides a rich Param (parameter) type for data interoperability in different application scenarios. In addition, the above various element functions can be expanded horizontally by adding GAspect ; the functions of a single node can be enhanced by introducing GAdapter ; or by adding GEvent (signals), the execution logic can be enriched and optimized.

本工程使用纯C++11标准库编写，无任何第三方依赖。 Compatible with MacOS , Linux , Windows and Android systems, it supports local compilation and secondary development through CLion , VSCode , Xcode , Visual Studio , Code::Blocks , Qt Creator and other IDEs. For specific compilation methods, please refer to CGraph compilation instructions.

For detailed functional introduction and usage, please refer to the article content in Yimianzhiyuan.com. Related videos are being updated continuously on B.com. Welcome to watch and communicate:

• [B Station Video] CGraph Introduction
  
• [B Station Video] CGraph Functional Chapter
  
  • A comprehensive introduction to all nouns and functional modules in the CGraph project
  • Based on the actual coding process, the specific usage scenarios, usages, and problems solved of each function are introduced in detail.
  • Suitable for children's shoes that want to have a full understanding of features and get started quickly with CGraph
  • Suitable for children's shoes interested in multi-threaded programming
• 【B site video】CGraph application
  
• 【B Station Video】CGraph Share
  

2. Use Demo

MyNode.h

# include " CGraph.h "

class MyNode1 : public CGraph ::GNode {
public:
    CStatus run () override {
        printf ( " [%s], sleep for 1 second ... n " , this -> getName (). c_str ());
        CGRAPH_SLEEP_SECOND ( 1 )
        return CStatus ();
    }
};

class MyNode2 : public CGraph ::GNode {
public:
    CStatus run () override {
        printf ( " [%s], sleep for 2 second ... n " , this -> getName (). c_str ());
        CGRAPH_SLEEP_SECOND ( 2 )
        return CStatus ();
    }
};

main.cpp

# include " MyNode.h "

using namespace CGraph ;

int main () {
    /* 创建一个流水线，用于设定和执行流图信息 */
    GPipelinePtr pipeline = GPipelineFactory::create ();
    GElementPtr a, b, c, d = nullptr ;

    /* 注册节点之间的依赖关系 */
    pipeline-> registerGElement <MyNode1>(&a, {}, " nodeA " );
    pipeline-> registerGElement <MyNode2>(&b, {a}, " nodeB " );
    pipeline-> registerGElement <MyNode1>(&c, {a}, " nodeC " );
    pipeline-> registerGElement <MyNode2>(&d, {b, c}, " nodeD " );

    /* 执行流图框架 */
    pipeline-> process ();

    /* 清空流水线中所有的资源 */
    GPipelineFactory::remove (pipeline);

    return 0 ;
}

As shown in the figure above, when the graph structure is executed, node a is first executed. After node a is executed, node b and c are executed in parallel. After all nodes b and c are executed, node d is executed.

3. Recommended reading

• Pure Serialist introduces you to the simple implementation of the graphical framework-execution logic
• Pure Serialist introduces you to the simple implementation of the graphical framework-cycle logic
• Pure sequencer introduces you to the simple implementation of graphical framework-parameter delivery
• Pure Serialist introduces you to the simple implementation of the graphical framework—conditional judgment
• Pure Sequels introduce you to the simple implementation of graphical frameworks——Face-oriented
• Pure sequencer introduces you to the simple implementation of graphical framework-function injection
• Pure Serialist introduces you to the simple implementation of the graphical framework-the message mechanism
• Pure Serialist introduces you to the simple implementation of the graphical framework - Event Triggering
• Pure sequencer introduces you to the simple implementation of the graphical framework-timeout mechanism
• Pure sequencer introduces you to the simple implementation of graphical framework-thread pool optimization (I)
• Pure sequencer introduces you to the simple implementation of graphical framework-thread pool optimization (II)
• Pure sequencer introduces you to the simple implementation of graphical framework-thread pool optimization (III)
• Pure sequencer introduces you to the simple implementation of graphical framework-thread pool optimization (IV)
• Pure sequencer introduces you to the simple implementation of graphical framework-thread pool optimization (V)
• Pure sequencer introduces you to the simple implementation of graphical framework-thread pool optimization (VI)
• Pure sequencer introduces you to the simple implementation of graphical framework-performance optimization (I)
• Pure sequencer introduces you to the simple implementation of graphical framework-performance optimization (II)
• Pure Sequels introduce you to the simple implementation of graphical framework-distance calculation
  
  
• CGraph's title song - "Listen to the Coder"
• Talk about this year I wrote CGraph
• What is it like to lead a project included in awesome-cpp from scratch?
• Breaking out! After CGraph's performance surpassed taskflow in a comprehensive way, the author said that he wanted to...
• Using map optimization: Summary of ideas for calculating the maximum concurrency of dag in CGraph
• One article will help you learn about CGraph, which lasts for two and a half years.
• CGraph authors want to know, if you need an eDAG scheduling framework
• Edge reduction and efficiency enhancement: Summary of redundant edge cutting ideas in CGraph

IV. Related Projects

• GraphANNS: Graph-based Approximate Nearest Neighbor Search Working off CGraph
• CThreadPool: A simple and easy-to-use, powerful, excellent performance, cross-platform C++ thread pool
• CGraph-lite : head-only, simplest CGraph, with DAG executor and param translate function
• awesome-cpp : A curated list of awesome C++ (or C) frameworks, libraries, resources, and shiny things. Inspired by awesome-... stuff.
• awesome-workflow-engines : A curated list of awesome open source workflow engines
• taskflow : A General-purpose Parallel and Heterogeneous Task Programming System
  • [B Station Video] Actual test of performance comparison between CGraph and taskflow
    
• torchpipe : Serving Inside Pytorch
  • 【B站视频】开源项目torchpipe - ai引擎在线高并发经验和实战
• nndeploy : nndeploy is a model end-to-end deployment framework. With multi-end reasoning and directed acyclic graph model deployment as the core, it is committed to providing users with a cross-platform, simple, and high-performance model deployment experience.
  • [B Station Video] nndeploy---end-deploy framework for AI model end-to-end deployment (1)
  • [B Station Video] nndeploy---end-to-end deployment framework of AI model (2)
• KuiperInfer: Takes you to implement a high-performance deep learning inference library from scratch, supporting inference of models such as llama2, Unet, Yolov5, and Resnet. Implement a high-performance deep learning inference library step by step
  • 【B Station Video】KuiperInfer Infer Framework - A teaching-oriented reasoning framework
• OGraph : A simple way to build a pipeline with Go.
  • [B Station Video] Listen to Alibaba Cloud's bigwig: OGraph - Go-based streaming map scheduling two or three things

---

Appendix-1. Version Information

[2021.05.04 - v1.0.0 - Chunel]

• Provide graphical execution function and support parallel computing of non-dependent nodes

[2021.05.09 - v1.1.0 - Chunel]

• Optimize the concurrency during graph execution

[2021.05.18 - v1.1.1 - Chunel]

• Add node name and session information

[2021.05.23 - v1.2.0 - Chunel]

• Provide single-node loop execution function

[2021.05.29 - v1.3.0 - Chunel]

• Provide cluster (cluster) and region (region) division and loop execution functions
• Provide tutorial content, including multiple usage examples

[2021.06.14 - v1.4.0 - Chunel]

• Provide param (parameter) delivery mechanism
• Provide group (group) function, multi-node modules are inherited from group modules uniformly
• Add support for Linux systems

[2021.06.20 - v1.4.1 - Chunel]

• Provide condition (condition) function
• Add support for Windows systems

[2021.06.24 - v1.5.0 - Chunel]

• Provide pipeline factory creation method
• Update tutorial content

[2021.07.07 - v1.5.1 - Chunel]

• Optimize thread pool functionality. Implement task theft mechanism

[2021.07.11 - v1.5.2 - Chunel]

• Optimize thread pool functionality. Implement automatic thread adjustment mechanism

[2021.07.31 - v1.5.3 - Chunel]

• Optimize thread pool functionality. Implement task batch acquisition function and optimize task theft mechanism

[2021.08.29 - v1.6.0 - Chunel]

• Provide multiple pipeline functions to optimize underlying logic
• Update tutorial content

[2021.09.19 - v1.6.1 - Chunel]

• Provides Lru operator, Trie operator and template node functions to optimize the underlying logic
• Update tutorial content

[2021.09.29 - v1.7.0 - Chunel]

• Provides aspect function for extending outward node or group function
• Update tutorial content

[2021.10.07 - v1.7.1 - Chunel]

• Optimize aspect (section) implementation logic, provide section parameter function, and batch add section function
• Update tutorial content

[2021.11.01 - v1.8.0 - Chunel]

• Provide adapter function and singleton adapter function
• Optimize pipeline execution logic
• Update tutorial content

[2021.12.18 - v1.8.1 - Chunel]

• Optimized return value CStatus information

[2022.01.02 - v1.8.2 - Chunel]

• Provides the automatic exit function of node execution timeout, and provides task group function
• Provide thread pool configuration parameter setting method

[2022.01.23 - v1.8.3 - Chunel]

• Provides function adapter to implement functional programming functions
• Provide thread priority scheduling function, and provide thread-bound CPU execution function
• Update tutorial content

[2022.01.31 - v1.8.4 - Chunel]

• Provides the function of node (node) asynchronous execution

[2022.02.03 - v1.8.5 - Chunel]

• 提供daemon (守护)功能，用于定时执行非流图中任务
• Update tutorial content

[2022.04.03 - v1.8.6 - Chunel]

• Provide DistanceCalculator operator to implement calculations of any data type and any distance type
• Update tutorial content

[2022.04.05 - v2.0.0 - Chunel]

• Provide domain (domain) function, provide Ann domain abstract model, and begin to support individual professional directions
• Provide a hold execution mechanism
• Update tutorial content

[2022.05.01 - v2.0.1 - Chunel]

• Optimize pipeline registration mechanism and support custom order execution of init methods
• Provide one-click compilation scripts

[2022.05.29 - v2.1.0 - Chunel]

• Provide element parameter writing method
• Provide support for C++14 version
• Update tutorial content

[2022.10.03 - v2.1.1 - Chunel]

• Provide task priority mechanism in thread pool
• Optimize group execution logic

[2022.11.03 - v2.2.0 - Chunel]

• Provide message (message) function, mainly used to complete data transmission between different pipeline
• Update tutorial content

[2022.12.24 - v2.2.1 - Chunel]

• Provides TemplateNode function to optimize parameter transfer method
• Update tutorial content

[2022.12.25 - v2.2.2 - yeshenyong]

• Optimize graph execution logic

[2022.12.30 - v2.2.3 - Chunel]

• Provide message publishing and subscription function
• Provides execution engine switching function

[2023.01.21 - v2.3.0 - Chunel]

• Provide event (event) function
• Provide the CGraph Intro.xmind file, which introduces the overall logic of CGraph through the brain diagram.

[2023.01.25 - v2.3.1 - Chunel]

• Provides support for C++11 version. Thanks to MirrorYuChen for providing relevant solutions

[2023.02.10 - v2.3.2 - Chunel]

• Optimize scheduling strategies and provide scheduling parameter configuration interface
• English version readme.md is available

[2023.02.12 - v2.3.3 - yeshenyong, Chunel]

• Provide graphviz visual image display function
• Provide parameter link tracking function

[2023.02.22 - v2.3.4 - Chunel]

• Optimize the scheduling mechanism under Windows system
• Optimize param mechanism and event (event) mechanism

[2023.03.25 - v2.4.0 - woodx, Chunel]

• Provide a runnable docker environment and dockerfile files for building docker environments
• Provide pipeline scheduling resource management mechanism
• Optimize scheduling performance

[2023.05.05 - v2.4.1 - Chunel]

• Provide thread binding execution function
• Provides pipeline maximum concurrency acquisition method. Thanks to Hanano-Yuuki for providing relevant solutions
• Provides pipeline asynchronous execution function and exit function during execution

[2023.06.17 - v2.4.2 - Chunel]

• Provides MultiCondition (multi-condition) function
• Provides pipeline pause execution and resume execution functions

[2023.07.12 - v2.4.3 - Chunel]

• 优化CStatus功能，添加了异常定位信息

[2023.09.05 - v2.5.0 - Chunel]

• Provides perf function for pipeline performance analysis
• Provides timeout mechanism for element
• Provide some (partial) functions to optimize the asynchronous execution of pipeline

[2023.09.15 - v2.5.1 - Chunel]

• Provides fence function
• Provides coordinator (coordinator) function

[2023.11.06 - v2.5.2 - Chunel]

• Optimize message (message) function, which can set the processing method when writing blocking, and reduce the number of memory copy times
• Add example related content to provide some simple implementations for different industries
• Optimize scheduling performance

[2023.11.15 - v2.5.3 - Chunel]

• Provide proto definition file
• Add mutable function to provide dependency registration syntax sugar

[2024.01.05 - v2.5.4 - Chunel]

• Provide test content, including performance and functionality test cases
• Optimize event (event) mechanism and support asynchronous waiting function

[2024.07.18 - v2.6.0 - PaPaPig-Melody, Chunel]

• Provides the way to execute pipeline topology
• Provide a method to determine whether there is a dependency between element
• Provide bazel compilation method
• Optimize perf function

[2024.09.17 - v2.6.1 - Chunel]

• Provides a static execution method of pipeline and provides a micro-task mechanism based on static execution
• Provides pipeline clipping function to remove duplicate dependencies between element
• Provides a method to delete dependencies element
• Optimize event (event) mechanism, asynchronous events can wait for end
• Release the CGraph-lite project, providing simple DAG composition and parameter transfer functions. The interface is fully compatible and can be switched to this project seamlessly

[2024.11.16 - v2.6.2 - Chunel]

• Optimize parameter mutual exclusion mechanism and obtain performance
• Fixed the problem of waiting for the auxiliary thread exception
• Update tutorial content

---

Appendix-2. Thanks

• Thank you Doocs for publishing relevant introduction documents on WeChat official account. Welcome to join the Doocs open source community

• Thanks to the journal introduction and recommendation of HelloGithub: HelloGithub Issue 70

• Thanks to the introduction and recommendation of "Github Chinese rankings": Github Chinese rankings total - C++ category

• Thanks to the recommendation from awesome-cpp, we all know, it is the most authoritative recommendation list for cpp project in the world
• Thanks to the recommendation from Taskflow Group : awesome-parallel-computing, and we always treat taskflow as a role model
• Thanks to the recommendation from awesome-workflow-engines
• Thank you to all developers CONTRIBUTORS for their contribution to the project
• Thank you to all friends who have made comments and suggestions for CGraph project, and will not mention them one by one. Everyone is welcome to join and build together

---

Appendix-3. Contact information

• WeChat: ChunelFeng (Welcome to scan the QR code above to add the author as a friend. Please briefly note your personal information^_^)
• Email: [email protected]
• Source code: https://github.com/ChunelFeng/CGraph
• Forum: www.chunel.cn