micro test review for Java7

[Nathan Tippy]

Overall I found Java7 did better than I expected.  But I did have problems getting all the tests to run. 

Apparently the fortran tests using matmul do not work on my 64bit platform so the results there are not quite right.  To get the test to complete I commented out the one matmul line so fortran shows as faster on those matrix tests.  

I was also unable to get the go to work the the latest 1.1 so I commented it out for now.

For the Java code I did both a 100% Java and JBLAS implementation.  These tests only show the call to JBLAS, the 100% Java implementation turned out to be 2-4 times slower, not as bad as I expected.

See attached HTML

My changes are here, please feel free to make suggestions.

https://github.com/nathantippy/julia/tree/master/test/perf/micro

[Nathan Tippy]

These tests were run on an AMD Phenom II N950 Quad-Core notebook with Linux Mint

[Stefan Karpinski]

I'm not at all surprised that Java is fast – it should be after nearly 20 years, billions of dollars, and god only knows how many person-hours poured into it. Something super suspicious is going on with the Mandelbrot test. It looks like the C version got crazy slow for some reason.