I found the graph a little hard to grok at first. Perhaps it would be clearer for the graph to show kloc/second/core. Another nice thing you could do is include benchmarks for both native-code and byte-code compilation, to support the observation that the byte-code compiler is faster.
It’s tricky, but it would be neat to have some comparative data. If OCaml runs at something like 2-8k loc, how fast do other languages (C, C++, Java, Rust, Haskell) run at? Are there any experiments about this we can point at?
y