MLL-Docker - Machine Learning Language Docker Image & Container

This repository contains auxiliary material for the MLL workshop.

Requirements

The following software products are required:

• A Java Development Kit (JDK) such as OpenJDK. Preferably install through a package manager like apt on UNIX-like systems.
• A Java IDE such as Eclipse.
• A Docker installation.
• A shell such as Bash (comes preinstalled on UNIX-like systems) or Git Bash on Windows.

Instructions

Install the software packages above. Then perform the following steps. Please let us know if you cannot confirm the bullet points marked with Please Confirm.

• Clone (ideally; else download) this repository to your computer.
• Open a shell and change the working directory to the root of this repository. In most shells this is done using the cd (change directory) command.
• Execute docker-compose up -d. The first run might require some time as Docker is downloading the required images. Subsequent runs will be much faster.
• Open the directory shared/mll in Eclipse as a project.
• Run the code once to automatically generate .class files.
  • Please Confirm I: There should be no errors. You should see (approximately) the following output:

  Saved plot: out/plots/example.png
  Rendered dot file: out/dot/example.png
  Hello from LLVM!
  Optimized LLVM code: out/llvm/helloworld-O3.ll
  Hello from LLVM!
  

  • Please Confirm II: When the file tree in Eclipse is refreshed (click on the root, then press F5), you should find the files as mentioned in the output above (i.e., a plot in the out\plots directory, a graph in the out\dot directory and an (optimized) LLVM program in the llvm directory with the name helloworld-O3.ll.
• Open the website localhost:8888 in your browser. If required, provide the password ml.
• Run the notebook mll.ipynb located in the mll directory. You can also alter some values, e.g., change the graph or the function.
  • Please Confirm III: There should be no errors. You should see a function plot and a graph below the respective cells. There should be an (optimized) LLVM program in the directory mll\out\llvm with the name helloworld-O3.ll.

At this point you should be able to edit code locally in Eclipse and remotely through a Jupyter Notebook in your browser.

Run docker-compose down to stop the docker container when finished with working on this project.

Notes

• Code from the package mll will be available in a Jupyter Notebook once it is compiled to .class files and stored in the shared/mll/bin directory. The package needs to be imported using import mll.* in any Jupyter Notebook.
• Any changes to these .class files requires a kernel restart.