Using a blackhole.consume to avoid dead code elimination

[arnaudroger]

I believe the result of the benchmark are not representative a real use case because of dead code elimination. Can you try using jmh as benchmark framework and blackhole.consume in the loop instead of incrementing a counter?
I see different result on my csv mapper code - it might come from different input but quite unlikely -. Code not available yet but should be in the next few days.

http://openjdk.java.net/projects/code-tools/jmh/

it will generate a self contain jar with which you can run the benchmark easily and collect the result in different format. It will also make the management of all the benchmark easier. Once you have the project setup it's very easy to use.

[jbax]

Thank you, waiting for your pull request.

[arnaudroger]

I did not say I would do it :)
I'll try to add it for yours parser, jackson and mine.
not time for a bit though so might take a few days
in the mean time
https://github.com/arnaudroger/orm-benchmark
uses jmh if you feel like having a look.