Deeply-nested super or constant lookup is slower in 9k than in 1.7. #2362
Description
Originally reported as http://jira.codehaus.org/browse/JRUBY-6312
The benchmark:
https://gist.github.com/headius/d1e6b73c79ca608d6c6c
It appears to be consistently slower than 1.7, even with invokedynamic mode. This will be a good one to investigate before 9k final.