The paper uses wc -l (probably because sloccount doesn't handle go and scala?) to count the lines of code.

Here is the output of sloccount:

    SLOC    Directory       SLOC-by-Language (Sorted)
    595     java_pro        java=595
    591     java            java=591
    488     cpp             cpp=488
    328     python          python=328
    0       go              (none)
    0       go_pro          (none)
    0       scala           (none)
    0       scala_pro       (none)

    generated using David A. Wheeler's 'SLOCCount'.

Compare it to the paper.

    Benchmark   wc -l
    C++ Dbg/Opt 850
    Java        1068
    Java Pro    1240
    Scala       658
    Scala Pro   297
    Go          902
    Go Pro      786