Speedup learning

[Andrea Fioraldi]

Hi,

I'm trying to speedup the running time of Daikon, there are some existing tips about this?

At the moment I'm trying to preprocess the set of dtraces to remove duplicates across files, is this a correct move or will produce wrong results?

There are any ways to parallelize Daikon and then merge the results of several runs on different sets of dtraces?

Regards,

Andrea

[Michael Ernst]

Andrea-

I'm trying to speedup the running time of Daikon, there are some existing tips about this?

What have you tried so far?  In the "Troubleshooting" section of the manual, there is a subsection titled "Daikon runs slowly".  I recommend reading those sections of the manual.  The manual is a good place to look, before asking a question that it answers.  We are always happy to answer questions that the manual does not answer!

 

There are any ways to parallelize Daikon and then merge the results of several runs on different sets of dtraces?

The first occurrence of "merge" in the Daikon manual is a link to the MergeInvariants tool, which does exactly what you asked for.  That section of the manual is also linked from the "Daikon runs slowly" section that I mentioned above.

 

At the moment I'm trying to preprocess the set of dtraces to remove duplicates across files, is this a correct move or will produce wrong results?

This is a good question.  I'm not sure what "duplicates" you are removing, but if they are duplicate dtrace entries (which Daikon calls "samples", then that will change Daikon's output due to its statistical tests.  It's up to you to decide whether those results are "wrong".  The manual tells you how to disable or change the statistical tests, such as the first option in the "Options to control invariant detection" section.

I hope this helps.

Mike