Re: [Windows,PATCH] Use faster, higher precision timer API

On 2014-09-17 11:19:36 -0400, Andrew Dunstan wrote:
>
> On 09/17/2014 08:27 AM, Craig Ringer wrote:
> >Hi all
> >
> >Attached is a patch to switch 9.5 over to using the
> >GetSystemTimeAsFileTime call instead of separate GetSystemTime and
> >SystemTimeToFileTime calls.
> >
> >This patch the first step in improving PostgreSQL's support for Windows
> >high(er) resolution time.
> >
> >In addition to requiring one less call into the platform libraries, this
> >change permits capture of timestamps at up to 100ns precision, instead
> >of the current 1ms limit. Unfortunately due to platform timer resolution
> >limitations it will in practice only report with 1ms resolution and
> >0.1ms precision - or sometimes even as much as 15ms resolution. (If you
> >want to know more, see the README for
> >https://github.com/2ndQuadrant/pg_sysdatetime).
> >
> >On Windows 2012 and Windows 8 I'd like to use the new
> >GetSystemTimePreciseAsFileTime call instead. As this requires some extra
> >hoop-jumping to safely and efficiently use it without breaking support
> >for older platforms I suggest that we start with just switching over to
> >GetSystemTimeAsFileTime, which has been supported since Windows 2000.
> >Then more precise time capture can be added in a later patch.
> >
> >
>
>
> That will presumably breaK XP.

The proposed patch? I don't really see why? GetSystemTimeAsFileTime() is
documented to be available since win2k?

Or do you mean GetSystemTimePreciseAsFileTime()? That'd surely - as
indicated by Craig - would have to be optional since it's not available
anywhere but 2012 and windows 8?

> I know XP has been declared at EOL, but there
> are still a heck of a lot of such systems out there, especially in places
> like ATMs, but I saw it in use recently at a US surgical facility (which is
> slightly scary, although this wasn't for life-sustaining functionality). My
> XP system is still actually getting some security updates sent from
> Microsoft.

I unfortunately have to agree, dropping XP is probably at least a year
or two out.

Greetings,

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services