On Mon, Aug 8, 2022 at 12:27 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> BTW, that commit really should have updated the explanation at the top of
> instr_time.h:
>
> * This file provides an abstraction layer to hide portability issues in
> * interval timing. On Unix we use clock_gettime() if available, else
> * gettimeofday(). On Windows, gettimeofday() gives a low-precision result
> * so we must use QueryPerformanceCounter() instead. These macros also give
> * some breathing room to use other high-precision-timing APIs.
>
> Updating the second sentence is easy enough, but as for the third,
> I wonder if it's still true in view of 24c3ce8f1. Should we revisit
> whether to use gettimeofday vs. QueryPerformanceCounter? At the very
> least I suspect it's no longer about "low precision", but about which
> API is faster.
Yeah, that's not true anymore, and QueryPerformanceCounter() is faster
than GetSystemTimePreciseAsFileTime()[1], but there doesn't
really seem to be any point in mentioning that or gettimeofday() at
all here. I propose to cut it down to just:
* This file provides an abstraction layer to hide portability issues in
- * interval timing. On Unix we use clock_gettime() if available, else
- * gettimeofday(). On Windows, gettimeofday() gives a low-precision result
- * so we must use QueryPerformanceCounter() instead. These macros also give
- * some breathing room to use other high-precision-timing APIs.
+ * interval timing. On Unix we use clock_gettime(), and on Windows we use
+ * QueryPerformanceCounter(). These macros also give some breathing room to
+ * use other high-precision-timing APIs.
FWIW I expect this stuff to get whacked around some more for v16[2].
[1] https://devblogs.microsoft.com/oldnewthing/20170921-00/?p=97057
[2] https://commitfest.postgresql.org/39/3751/