On 2016-12-27 01:35:05 +0000, Greg Stark wrote:
> On Dec 26, 2016 10:35 PM, "Tom Lane" <tgl@sss.pgh.pa.us> wrote:
>
>
> So it seems like the configure support we'd need is to detect
> whether clock_gettime is available (note on Linux there's also
> a library requirement, -lrt), and we would also need a way to
> provide a platform-specific choice of clockid; we at least need
> enough smarts to use CLOCK_MONOTONIC_RAW on macOS.
>
> This seems like something that really should be checked at runtime. It's
> very specific to the specific kernel you're running on, not the build
> environment, and it can hopefully be measured in only a second or even a
> fraction of a second. The only Pebblebrook would be if other things running
> on the system made the test results unpredictable so that you had a small
> chance of getting a very suboptimal choice and we ruling the dice each time
> you restarted...
I'm pretty strongly against doing performance measurements at
startup. Both the delay and the potential for differing test results
seem like pretty bad consequences.
Greetings,
Andres Freund