Faster Large-Model Transformation

title:Faster Large-Model Transformation
keywords:
topics:Case studies and Applications, Graphs
contact:prof.dr.ir. A. Rensink
to be started:any time


Description

In the context of the European CHARTER project, the FMT group has developed and implemented a compiler that produces model-transforming Java code from a high-level, declarative rule language. This compiler is a core component in several essential transformation steps, including code generation from UML diagrams, machine code generation from Java bytecode, and machine code optimisation.

The models that should be transformed by the generated code are very large (in the order of 10^5 - 10^6 elements). This puts a serious performance constraint on the code. The current (prototype) compiler has not yet been optimised for performance, and moreover, has not yet been extensively tested on graphs of that magnitude. This MSc assignment therefore involves evaluating the current performance and improving it where possible.

Tasks

  • Get acquainted with the context of CHARTER and the FMT-developed compiler
  • Set up a large case study to assess the performance of the generated code
  • Experiment and identify bottlenecks
  • Propose and implement measures to improve performance
  • Evaluate the results

References

  1. The CHARTER project (Digital version available here)