"14% faster than LLVM -O0" is fairly misleading too. How does the generated code compare with LLVM -O2/-O3? These are SQL queries, which are usually fairly short programs, where I would presume the cost of compiling is negligible compared to the cost of execution.

The generated code runs an order of magnitude faster than interpretation and 14% faster than LLVM -O0.

An order of magnitude faster than interpretation. That's the interesting part for Python, I'd think.

I'd say it's a starting point, and also how else would you measure a baseline ?

You're right that O2/O3 should be compared too, though they will be more moved targets.

Measure time-to-results = compile-time + execution-time, for jit/O0/O2/O3. Fairly basic.