Navigation Menu

Skip to content

ctuning/ck-crowdnode

Repository files navigation

Linux/MacOS: Build Status Windows: Windows Build status

Collective Knowledge Node for experiment crowdsourcing (on Windows devices)

Standalone, thin and portable server to let users participate in experiment crowdsourcing via Collective Knowledge. It unifies remote execution on Windows similar to Android ADB (experiment crowdsourcing via Linux and Android platforms is already supported by CK). It can also be used to create farms of machines for collaborative benchmarking and tuning (crowd-benchmarking).

Note that both server and client should run Windows.

Project homepage:

License

  • Permissive 3-clause BSD license. (See LICENSE.txt for more details).

Status

Relatively stable - testing phase

Usage: server side

On Windows:

  1. Download the installer from Appveyor

  2. Install and start "CK crowd-node server".

  3. Write down "[INFO for CK client]" - you will require this info to configure this target machine on a client.

Alternatively, you can build it manually (you'll need CMake and Visual Studio):

mkdir build
cd build
cmake ..
cmake --build . --config Release --target PACKAGE

Now, you should be able to run build/Release/ck-crowdnode-server.exe.

Usage: client side

Install CK framework. If you have PIP, you can install it simply as following:

 $ pip install ck

Pull ck-autotuning repository (including dependencies):

 $ ck pull repo:ck-autotuning

Prepare local file with a secret key (see [INFO for CK client]), for example in C:\secret-key.txt

Register target machine with ck-crowdnode-server via (substitute ''my-remote-target'' with any other user-friendly name)

 $ ck add machine:my-remote-target

Select 4) CK: remote Windows machine accessed via CK crowd node. Then 4) windows-64

Then enter hostname, port, path to public key (C:\secret-key.txt), and full path to files on a target machine (all info is available via [INFO for CK client] - we later plan to automate this process).

Now you can check that you machine is connected and online via

 $ ck show machine

or

 $ ck browse machine

Now you should be able to compile and run sample program using this target. You need to have Microsoft C compilers and Microsoft SDK installed (there is a free edition available). You can also download and install LLVM for Windows, but remember that it also requires Visual C compiler and Microsoft SDK.

Try to compile susan benchmark (during first compilation, CK will attempt to automatically detect installed compilers and SDK while asking you extra questions, if needed):

 $ ck compile program:cbench-automotive-susan --speed --target=my-remote-target

Finally, you can try to run it:

 $ ck run program:cbench-automotive-susan --target=my-remote-target

If everything is configured correctly, this code will be executed several times on a required target and execution time will be reported!

If you have any problems, questions or comments, do not hesitate to get in touch with the CK community via [our public mailing list](https://groups.google.com/forum/#!forum/collective-knowledge open CK mailing list)!

Further details

Public discussions

logo