emoflon / cmoflon Goto Github PK

With the newly introduced code reduction heuristics, the code that is generated from SDMs is only inserted iff it is part of the TC algorithm that the code is currently generated for. However there is a fallback implemented but it shold be tested whether this is ok

As the LMST algorithm is fairly sensitive to the direction in which it observes edges it is necessary to have inverted duplicates of every edge that is not directly attached to the this node. The duplicates should be generated at runtime before LMST is executed. They will then be attached to the neighbor list. They need to be removed after executing LMST.

• Easiest (but also dangerous) handling vie preprocessor directive
• Add warning to cMoflon Tutorial

Until now, each pattern invocation mallocs and frees the memory for its pattern.
We now use a global match data structure (residing in static memory) whose size is determined at code generation time.

We should split up cMoflon into separate repositories

• cmoflon: Should contain only the source code
• cmoflon-examples: Should contain all examples including everything related to ECMFA2017 (including the corresponding update site)

See eMoflon::TIE

DefaultCMoflonAttributeConstraintsLib.xmi

Currently the code for all different topology control algorithms is generated into one single source file. In the future files should only contain the code that they also reference.

Content: default implementations (custom-helper.c)

The current LMST variant is reliant on Issue #7 which is hard on the memory. Dense networks like Flocklab with its default transmission range can not properly execute LMST this way.

See the following lines in CMoflonCodeGenerator

final String parentFolderForAlgorithm = getProjectRelativePathForAlgorithm(null);
final String outputFileName = parentFolderForAlgorithm + "cMoflon.h";

Helper Resources

• org.cmoflon.evaluation.flocklab
  • Default list of node IDs
  • Placeholder for dbImageID, description, name, duration