Education Python 3.9 compiler for conference talk at PHDays 2024

This is a simplest set of examples of MLIR usage for very small subset of Python. It is splitted into 5 chapters, but each chapter is much more compact than Toy language in MLIR examples.

Requirements and environment setup

1. First you need MLIR (on Linux you can get MLIR 19.x from Ubuntu/Debian apt repositories).
2. Python 3.9 is also required since it contains AST parser functions in C API (in Python 3.10+ these functions are hidden).
3. (Windows only) Visual Sttudio 2022 (Community Edition is enough minimum).

On Ubuntu MLIR 19.x can be installed by this Shell script:

./install.sh

On Windows MLIR can be built from source code by this batch script (VS 2022 needs to be installed manually):

./install.bat

Chapters' structure

1. Generate the most simple IR (in 22 lines of code).
2. Generate IR for variables assignment using Python 3.9 frontend.
3. Generate IR for 'if-else' statements, including simple type inference.
4. Lower and translate to LLVM IR (with command line options).
5. Run LLVM IR using JIT engine ORCv2 (ORC == On-Request Compilation).

Each chapter contains build.sh and build.bat scripts.