Processing time of optimizers increases when optimizers are executed multiple times

[swagner]

When executing a batch run which contains an optimizer with a rather short run time, the processing time of the optimizer (i.e. going from the stopped into the prepared and from the prepared into the running state) takes longer an longer the more often the optimizer is executed. However, the execution time of the optimizer is not affected. It seems that there is some problem in the event handling of the execution state changes.

[mkommend]

The cause of the slow down was the object graph traversal in Algorithm, that had to been done to discover all IStatefulItems, which have to be initialized or cleared.

r7467:

• added possibility to exclude members from object graph traversal by name
• excluded runs from object graph traversal in the Algorithm
• adapted DeepCloneableCloningTest

[abeham]

I looked at the changes in r7467. You currently exclude any "runs" field in any member of the object graph. Is this the desired intention or do you want to exclude a specific runs collection only? Did you also consider to pass a HashSet<object> to the method and compare if it contains the fieldvalue instead of passing a HashSet<string> and comparing against the fieldname?

[mkommend]

Implemented previous suggestions from abheam.

r8212: Changed object graph traversal to use a hashset of objects for excluding specific members.

[abeham]

r8220: fixed a bug

[mkommend]

Thx for spotting this.