I have a regression with the performance-plugin 3.3 when I have huge (100MB) JMeter result file. That used to work fine with v3.2 of the plugin.
After a little research, I have isolated the commit that introduced this problem, it's https://github.com/jenkinsci/performance-plugin/commit/cf7619661aae252ad7866ffc2e819c2e9c929afa
The ParserDetector class now load entirely the file to detect which parser to use. It would be better to use a StAX implementation to perform this detection.
I have a pull request to demonstrate the problem and provide a fix.