• Your next WebGPU, is not WebGPU
  • Introduction
  • KEEP IT OPEN-SOURCE
  • Getting started
    • Create Your Working file
    • Install ECE408 Remote Control Extension
    • Install the shell dependencies
    • Install the Chrome driver
      • Before started
      • Check for the right version of Chrome
      • Download the correct version of chromedriver
    • Install the other extension
    • Network connection
  • Working steps
    • Config
    • Login
    • Pull
    • Push
    • Exit
  • Trouble-shooting
  • Possible Improvements
  • Issue Report
  • Contributions
    • Set up your working environment
    • Renew the Extension Version
  • Acknowledgements
  • Release Notes

If you’re interested in this project and want to contribute to it, please take a look at my paper first. The paper was submitted to ICFTIC 2022 and you must cite this paper and the original paper of WebGPU if you want to build upon the VCS-WebGPU system.

@inproceedings{bai2021vsc,
  title={VSC-WebGPU: A Selenium-based VS Code Extension For Local Edit And Cloud Compilation on WebGPU},
  author={Bai, Hao},
  booktitle={2021 IEEE 3rd International Conference on Frontiers Technology of Information and Computer (ICFTIC)},
  pages={474--477},
  year={2021},
  organization={IEEE}
}
@inproceedings{dakkak2016webgpu,
  title={Webgpu: A scalable online development platform for gpu programming courses},
  author={Dakkak, Abdul and Pearson, Carl and Hwu, Wen-mei},
  booktitle={2016 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW)},
  pages={942--949},
  year={2016},
  organization={IEEE}
}

There are 2 versions of paper. To get a quick review, please refer to the published paper at https://ieeexplore.ieee.org/abstract/document/9647189. To get a detailed insight (full version), turn to my blog to get access to it. This variance is due to the fact that conference inprocedings require the page limit to be 4 so only a very limited amount of work was shown to global people.

Recently, programming on the Web is developing quickly, and SaaS has never been so popular. In UIUC, WebGPU is developed for ECE408 (Applied Parallel Programming) students to write, compile, and debug their code. However, as the architecture for WebGPU is not yet fully developed, and because of the natural difficulty to render code on HTML pages, we can observe that many students complain about this interface and they try to use other methods to code.

On the other side, VS Code has never been so popular because of its strong support for every language and its huge extension community. VS Code is also strong because it's pure editor, so it's compatible to all of the operating systems, including Windows, Mac, Linux, etc.

I hereby, introduce a VS Code Extension called ECE 408 Remote Control, which allows students to use their code created locally, using VS Code as the editor with decent syntax highlighting systems, and use the extension for posting the code back onto the WebGPU website, automatically running the tests online, and show the results on the Chrome browser.

As a steady supporter of the open source software movement, for most of the time I'd like to make my softwares open. Therefore, this extension is under the The GNU General Public License. If you want to make any form of edit to it, please make sure you keep it open source.

Before you can use the extension, there are some setting-up steps to follow.

When starting, you should use the VS Code to open your .cu CUDA source , for example,

code /path/to/your/workdir/lab1.cu

Note that you must use .cu as the postfix of your working file, or the extension vscode-cudacpp will not be able to recognize the file as a valid CUDA C++ file.

To set up the working environment, you need to install my extension first. After you install it, you can see the interface as shown below. Please click on the "security" and trust my extension button if it is not yet trusted.

If you're currently using Mac, use

brew install pbcopy

If you're currently using Windows, | clip should already be in your system PATH for cmd. You can open cmd.exe (not powershell) and type in echo 1 | clip to check.

You also need to install chromedriver into your system PATH.

Before you install the chromedriver, you should have:

Then you need to install the chromedriver.

Please open the current Chrome (if you have) and check which version you're in. For example, mine is 94.0.4606.

Go to the download website to download the correct version of ChromeDriver. If you're using Windows, make sure you download the Windows version. I'm using Mac, so I'll download the Mac version.

After downloading, please unzip the file and move it to your system PATH.

Then please check whether you've successfully installed your chromedriver. If you see the feedback below, you're all set.

Notes:

• You must download exactly the same version of chromedriver as your chrome browser if you want to use the chromedriver as your selenium driver. The chrome browser on your machine is likely to be updated automatically when a new version is released. In this case, please update to the newest version of chromedriver too.
• I strongly recommend you to install the chromedriver into the system software directory immediately, i.e. it should appear in /usr/local/bin/chromedriver.

In your VS Code workspace, please install the following extensions:

• vscode-cudacpp: CUDA C++ syntax highlight.

For students in China, make sure you’re equipped with a VPN so that you can gain a fast connection with the WebGPU server held in UIUC.

Note that the Canonical Press means you need to firstly press the combined keys shift+ctrl/cmd+p on your keyboard.

After creating the working file (lab2.cu in my case), please use VS Code to open it. Make sure you’ve done all things mentioned in the Getting Started part.

Then please open the Settings page in VS code. Detailedly, please go through Code/Preferences/Settings/ECE408 Remote Control/. Alternatively, you can just press ctrl, (win) or command, (mac).

For the headless config, please choose false for now, because the headless mode is still under test.

This part is fully automated. Just type in f1 and everything is set. After this step you will be on the code editing page.

This part is fully automated. Just type in f2 and everything is set. After this step the code on the WebGPU page will be automatically added to your working file.

This part is fully automated. Just type in f5 and everything is set. After this step your code in your working file will be automatically pushed onto the WebGPU website, saved, and also submitted for feedbacks.

This part is to simply exit the Chrome browser and shut down the Internet connection to keep your account safe.

• Try to change selenium to headless mode, and move the HTML rending part to local workspace.

If you find an issue or have some more features to request, please reply on the Github Issues Page. Please do NOT post it on CampusWire, which is dedicated for course-related contents.

If you want to contribute to the code, please contact [email protected] to send pull requests to the GitHub Repo.

Before you start, please make sure you’ve already completed all steps in the requirements part.

If you want to test your code after you make refinements to my code, please make sure there are no bugs. To debug, use the framework provided by Microsoft VS Code.

npm install -g yo generator-code
yo code // then follow the default settings

To renew versions on the VS Code Extension Marketplace, please type

npm i vsce -g # if you did not install vsce
vsce package # if you want to make a package
vsce publish # if you want to publish immediately

I need to express my greatest thanks to Professor Kindratenko, who encouraged me to go on with this VS Code extension. I also need to thank TA Andrew Schuh, who is also the maintainer of WebGPU, who provided me with the web API for the pull part of the extension. Gratitudes are also given to my peer student and teammate Zhenzuo SI, who encouraged me to develop such a tool.

Although all people mentioned above did not contributed directly to any part of this extension, they provided me with the greatest warmth and appreciation, without which I cannot develop such a functional tool. Thanks to all of you!

For the Release Notes part, please refer to CHANGELOG for details.