Something like this will do it?
procedure TForm1.Button1Click(Sender: TObject);
var
start,stop1,stop2,freq:int64;
i:integer;
begin
screen.cursor:=crHourGlass; {show busy cursor}
QueryPerformanceFrequency(freq); {Get frequency}
QueryPerformanceCounter(start); {Get initial count}
for i:=1 to 1000000 do; {empty loop}
QueryPerformanceCounter(stop1); {Get 1st end count}
for i:=1 to 1000000 do application.processmessages; {do it loop}
QueryPerformanceCounter(stop2); {Get 2nd end count}
screen.cursor:=crDefault; {show normal cursor}
If freq>0 {Display (loop2 count - loop1 count)/freq to get time in
microseconds}
then showmessage('Time for a call to processmessages is ' +
inttostr(((stop2-stop1)-(stop1-start)) div freq) +' microseconds')
else showmessage('No hardware timer available');
end;
Anyway looks like on dual core, QueryPerformanceCounter has some problems:
http://www.virtualdub.org/blog/pivot/entry.php?id=106
Ken Adam wrote:
> How are you measuring time?
> You should be using QueryPerformanceCounter and QueryPerformanceFrequency to
> get precise timing.
>
> I get very consistent results from this (timing real-time digital map
> graphics code) without messing around with priorities, nor caring what else
> is running.
> (they give a resolution in microseconds, so doing enough calcs to take say 1
> second gives a precise answer).
> If you simply use elapsed time, then you will get a large variation, due to
> whatever else is happening, and the low resolution of the clock.
>
> If you haven't used these I can provide a sample.
>
> Regards,
> Ken
>
> _______________________________________________
> Delphi mailing list -> [email protected]
> http://www.elists.org/mailman/listinfo/delphi
>
>
_______________________________________________
Delphi mailing list -> [email protected]
http://www.elists.org/mailman/listinfo/delphi
