On Tue, 4 Aug 2020 at 19:36, Michael Paquier <mich...@paquier.xyz> wrote: > Did you check if our implementation of src/port/snprintf.c makes %*s > much slower than %s or not? FWIW, I have just run a small test on my > laptop, and running 100M calls of snprintf() with "%s" in a tight loop > takes 2.7s, with "%*s" and a padding of 0 it takes 4.2s. So this test > tells that we are far from something that's substantially slower, and > to simplify the code your change makes sense. Still, there could be a > point in keeping this optimization, but fix the comment to remove the > platform-dependent part of it. Any thoughts?
It's not just converting "%s" to "%*s", it's sometimes changing a appendStringInfoString() call to appendStringInfo(). It's hard to imagine the formatting version could ever be as fast as appendStringInfo(). FWIW, the tests I did to check this when initially working on it are in [1]. Justin, it would be good if you could verify you're making as bad as what's mentioned on that thread again. David [1] https://www.postgresql.org/message-id/flat/20130924165104.GQ4832%40eldon.alvh.no-ip.org#4e8a716ff0bde1e950fe7ddca1d75454