Performance issues w.r.t. QVTO compiler and interpreter [message #704565] |
Thu, 28 July 2011 19:49 |
Nicolas Rouquette Messages: 157 Registered: July 2009 |
Senior Member |
|
|
At JPL, our model-based systems engineering efforts involve lots
of UML, SysML and QVT; all 3 open standards from the OMG.
For QVT Operational, we rely on the Eclipse QVTO implementation.
However, there are issues w.r.t. the efficiency of the QVTO compiler/builder
that make QVTO development increasingly painful.
There are several reasons for this:
1) The Eclipse QVTO compiler produces executable QVTO models (*.qvtox)
that could be saved in principle but the code isn't enabled to do so currently
so the source QVTO files (*.qvto) have to be compiled for each execution run.
2) The Eclipse QVTO builder is currently written to trigger a complete project
build if certain files are changed (e.g., *.ecore).
3) Ppartitioning complex transformations into several QVTO modules exacerbates
the compilation performance problem in part because the Eclipse QVTO compiler
doesn't optimize the order in which it compiles the files to mimize unecessary rework.
Are there knowledgeable folks who would be interested in coordinating efforts to address these problems?
- Nicolas.
|
|
|
Powered by
FUDForum. Page generated in 0.04089 seconds