Re: How to do fast performance timing

On 12/9/13 7:33 AM, Benedikt Grundmann wrote:
> At Jane Street we have recently spend a lot of time trying to get a fast gettimeofday. I saw lots of references in various postgres hacker threads related to a lack of such a facility so ....
>
> The culmination of those efforts can be read here:
>
> https://github.com/janestreet/core/blob/master/lib/time_stamp_counter.mli
> and
>
> https://github.com/janestreet/core/blob/master/lib/time_stamp_counter.ml
>
> it's all OCaml but the code is mostly imperative and very well documented. In particular we made an effort to document our assumption. There are a few which are ocaml specific. But a lot of the lessons we have learned here should be applicable to postgres.

Looks interesting. I think this isn't nearly as big an issue in Postgres as it used to be, but I think there's also things we've been avoiding because of the overhead. IE: using IO response time to determine if something came from cache or not.
--
Jim C. Nasby, Data Architect jim(at)nasby(dot)net
512.569.9461 (cell) http://jim.nasby.net