Cloudera Impala uses LLVM to generate code optimized for the specific query it is running (basically compiling away all the generalness around how large records

Cloudera Impala uses LLVM to generate code optimized for the specific query it is running (basically compiling away all the generalness around how large records are, etc). Some kind of native code generation is apparently standard practice in high-performance query engines that are written in languages that don't have a JIT.