Andrei Lepikhov <a.lepik...@postgrespro.ru> writes: > To be clear. In [1], I mentioned we can perform micro-benchmarks and > structure costs of operators. At least for fixed-length operators, it is > relatively easy.
I repeat what I said: this is a fool's errand. You will not get trustworthy results even for the cases you measured, let alone all the rest. I'd go as far as to say I would not believe your microbenchmarks, because they would only apply for one platform, compiler, backend build, phase of the moon, etc. regards, tom lane