Least squares
There was a problem generating the charts including the native code libraries. Until that is resolved this chart still refers to the results of an older benchmark execution. It was run on Mac and MTJ-N did then use Apple's veclib which is super fast. The native libraries used in "this" execution is far from that fast. Will update the charts as soon as the problems are resolved.
There's more Java, performance and ojAlgo related stuff at ojBlog.