MVEL performance regression

I'm filing this JIRA to track the performance regression talked over recently at the rules-dev mailing list. Blocker priority per Ḿark's off-list suggestion. Quote follows:

On 26/04/2011 15:53, Wolfgang Laun wrote:
> A simple rule containing an eval with a % b == 0 exhibits a considerable slowdown when comparing execution times between a Drools core built in February and the one there is now.
>
> The slow-down factor is 15.

There is a performance issue with MVEL. We moved to indexed (array based) mvel factories, which is suppose to give faster performance. However the MVEL author has not finished that work completely and so far they give a performance regression I'm pestering him daily on this.

Mark