ALE Examples & Benchmarks

This repository is structured as follows:

• examples: Toy examples demonstrating semantics definition and language modularity in ALE
• fUML: An implementation of fUML using ALE inspired by the Model Execution Case of the Transformation Tool Contest 2015 (TTC'15).

Playing with the examples

1. Setup an ALE environment following the installation instructions
2. Clone this repository locally, eg. git clone https://github.com/manuelleduc/ale-compiler-benchmarks
3. Import all the projects contained in the examples directory in Eclipse ..* File -> Import -> Existing Projects into Workspace ..* Select the examples directory as root directory in the dialog ..* Check all the projects ..* Finish

• Each project contains a launch configuration that can be used to run it
• To re-generate the Revisitor interfaces: Right click -> ALE -> Generate Revisitor interface on an Ecore file generates the corresponding Revisitor interface in the src directory of the current project
• To re-generate the Revisitor implementations: Right click -> ALE -> Generate Revisitor implementation on an ALE file generates the corresponding Revisitor implementation in the src directory of the current project

Running the benchmarks

This repository contains benchmarks comparing different implementations of the execution semantics of fUML. The concrete semantics code is common to all implementations: the only variation is the pattern used to implement it.

• Interpreter: The reference implementation of TTC'15 following the Interpreter pattern
• Visitor: An implementation following the classical Visitor pattern
• EMF Switch: An implementation using the Switch mechanism of EMF
• MonolithicRevisitor: A first Revisitor implementation where the runtime concepts of the activity diagram (Tokens, Offers, etc.) are already merged in a single metamodel
• ModularRevisitor: An alternative Revisitor implementation based on a static metamodel defining the abstract syntax of activity diagrams and another metamodel defining the runtime concepts

The fUML/activitydiagram contains the reference implementation of activity diagrams from TTC'15, plus a variant where the static concepts and the runtime concepts are modularly split in two different metamodels.

• For convenience, we provide pre-compiled JARs for all the projects and a Bash script that runs all of the benchmarks one after the other:

  1. Navigate to the ./fUML/benchmarks directory
  2. Run the benchmarks: ./benchmark.sh or benchmark.bat

• Otherwise, import all the Eclipse projects contained in the fUML directory and wait for all of them to compile without error

• Execute the BenchmarkGeneric class of the benchmark project of your choice (one per implementation folder). BenchmarkGeneric's main function expects 3 parameters:

  1. The path to a folder with the *.xmi models of the benchmark
  2. The name of the test to run (testperformance_variant1, testperformance_variant2, or testperformance_variant3)
  3. A prefix for the *.csv file that will store the results

Each benchmark executes every performance test of the TTC'15 contest 500 times after 50 warmups everytime.