xdsl 0.8.0

xDSL

xDSL: A common ecosystem for DSL development

xDSL is a compiler toolkit natively developed in Python. It allows users to reuse existing domain-specific intermediate representations (IRs), or to define their own custom IRs. xDSL uses multi-level IRs, meaning that during the compilation process, a program will be lowered through several of these IRs. This allows the implementation of abstraction-specific optimization passes, similar to the structure of common DSL compilers (such as Devito, Psyclone, and Firedrake). To simplify the writing of these passes, xDSL uses a uniform data structure based on SSA, basic blocks, and regions, which additionally enables the writing of generic passes.

An inherent advantage of a design close to MLIR is the easy interaction between the two frameworks, making it possible to translate abstractions and programs back and forth. This results in one big SSA-based abstraction ecosystem that can be worked with through Python, making analysis through simple scripting languages possible. Additionally, xDSL can leverage MLIR's code generation and low-level optimization capabilities.

Prerequisites

To install XDSL you can either clone the GitHub repository and install the requirements by following:

Clone and install

git clone https://github.com/xdslproject/xdsl.git
pip install -e .
# or for the optional requirements
# pip install -e .[extras]

pip installation

pip install xdsl

Testing

This project includes pytest unit test and llvm-style filecheck tests. They can be executed using to following commands from within the root directory of the project:

# Executes pytests which are located in tests/
pytest

# Executes filecheck tests
lit tests/filecheck

Generating executables through MLIR

xDSL can generate executables using MLIR as its backend. To benefit from this functionality, we first need to clone and build MLIR. Please follow: https://mlir.llvm.org/getting_started/

Next, we need to have mlir-opt, mlir-translate and clang in the path:

# For XDSL-MLIR
export PATH=<insert-your-path>/llvm-project/build/bin:$PATH

Given an input file input.xdsl, that contains IR with only the mirrored dialects found in src/xdsl/dialects (arith, builtin, cf, func, llvm, memref, and scf), run:

### Prints MLIR generic from to tmp.mlir
# e.g.  ./src/tools/xdsl_opt -t mlir -o tmp.mlir `input.xdsl`
/src/tools/xdsl-opt -t mlir -o tmp.mlir tests/filecheck/scf_ops.xdsl

mlir-opt --convert-scf-to-cf --convert-cf-to-llvm --convert-func-to-llvm --convert-arith-to-llvm --convert-memref-to-llvm --reconcile-unrealized-casts tmp.mlir | mlir-translate --mlir-to-llvmir > tmp.ll

The generated tmp.ll file contains LLVMIR, so it can be directly passed to a compiler like clang. Notice that a main function is required for clang to build. Refer to tests/filecheck/arith_ops.test for an example. The functionality is tested with MLIR git commit hash: 74992f4a5bb79e2084abdef406ef2e5aa2024368

Formatting

All python code used in xDSL uses yapf to format the code in a uniform manner.

To automate the formatting within vim, one can use https://github.com/vim-autoformat/vim-autoformat and trigger a :Autoformat on save.