forums.silverfrost.com

DanRRight · Posted: Sat Apr 19, 2008 9:27 am Post subject: CALL SYSTEM_CLOCK problem

In my new computer I got one annoyance. Now the code does not
give me correct time obtained this way

CALL SYSTEM_CLOCK (iclock,irate)
SecondsSinceTheMidnight = iclock/dble(irate)

And time goes back! I mean SecondsSinceTheMidnight decreasing all the way

What could be wrong?

PaulLaidler · Posted: Sat Apr 19, 2008 3:38 pm Post subject:

Problems with SYSTEM_CLOCK have been reported on some machines but so far I have not been able to reproduce this or make any progress towards a fix.

DanRRight · Posted: Sat Apr 19, 2008 10:51 pm Post subject:

Is it worth to get newer version of Salflibc (mine are ~5 years old) ?

PaulLaidler · Posted: Sun Apr 20, 2008 8:34 am Post subject:

No I don't think it will make any difference.

JohnCampbell · Joined: 16 Feb 2006 Posts: 2554 Location: Sydney

Paul,

There is a requirement for integer*8 variables to solve this as the count_rate in some processors is > 2^31, ie -ve. On my 3ghz P4 the count_rate was 3ghz. system_clock should be changed to detect when count_rate is not a good value.

You can work around the problem by recognising the time step is going backwards and calculate the -ve factor to apply to the time coming out of SYSTEM_CLOCK. I did and it worked fine for that pc only.

My solution is to use an alternative routine. I use :-

! ELAPSED TIME CALLS
! fastest gettickcount winapi
! accurate QueryPerform winapi
!
! CPU TIME CALLS
! fastest cpu_clock@ salford
! accurate cpu_clock@ salford

John

PaulLaidler · Posted: Tue Apr 22, 2008 7:01 am Post subject:

Thanks. I will take another look at this as soon as I can.