Performance Optimization Wins in JME

Benchmark results are extremely suspect in this particular case because I suspect there will be almost no difference in the byte code. I count exactly as many ‘field lookups’ before and after… so the comment is also incorrect.