Be seen. Boost your questionâ€™s priority for more expert views and faster solutions

I have used QueryPerformanceCounter & QueryPerformanceFrequency to detect System uptime, and it seemed to have worked, -well, until this morning that is.

Now, I am getting an uptime of 1 hour 20minutes and change, even though system has been up for over three days (Both systemInfo and "net statistics workstation" correctly identify that system has been up for three days plus). Nothing of significance really happened an hour and 20 minutes ago. (I have been logged in for the last three hours !).

I recall that PerformanceCounter may be off by a few clicks on multi-core/processor machines, but this is a an error of about 80 hours. Can someone tell me why the following code doesn't work ?

Now, I am getting an uptime of 1 hour 20minutes and change, even though system has been up for over three days (Both systemInfo and "net statistics workstation" correctly identify that system has been up for three days plus). Nothing of significance really happened an hour and 20 minutes ago. (I have been logged in for the last three hours !).

I recall that PerformanceCounter may be off by a few clicks on multi-core/processor machines, but this is a an error of about 80 hours. Can someone tell me why the following code doesn't work ?

```
function GetPerformanceCounterTick: Int64;
var PCount: Int64;
begin;
if QueryPerformanceCounter(PCount) then Result := PCount
else raise exception.create('QueryPerformanceCounter Error');
end;
function GetTimerFrequency: Int64;
var Frequency: Int64;
begin
if QueryPerformanceFrequency(Frequency) then Result := Frequency
else raise exception.create('QueryPerformanceFrequency Error');
end;
function SystemUptime:string;
var f, t : int64;
r, d, h, m, s : word;
begin;
f := GetTimerFrequency;
t := GetPerformanceCounterTick;
r := t div f; // in seconds
s := r mod 60; r := r div 60;
m := r mod 60; r := r div 60;
h := r mod 60; r := r div 24;
d := r;
if d >0 then result := Format('%3d Days %2d Hours %2d Minutes %2d Seconds', [d, h, m, s])
else if h > 0 then result := Format('%2d Hours %2d Minutes %2d Seconds', [h, m, s])
else if m > 0 then result := Format('%2d Minutes %2d Seconds', [m, s])
else result := Format('%2d Seconds', [s])
end;
```

I'm not sure if this is relevant here (since I didn't test anything), but shouldn't be the calculation of hours like this?:

> h := r mod 24; r := r div 24;

ZOPPO

No, we are going from minutes to hours, so I think my caculations are OK. I'll check thatand get back if there are any errors in it.

Geert:

I am using int64 (which is 64 bit) except for days, hours, minutes and seconds. I see your point that I might be crossing the 32-bit boundary on seconds (86400 in a day). I'll check it and let you know. (I just rebooted my computer, so it'll have to wait for a day). Thanks.

Here a step by step calculation:

// let's assume r = 180330 (50 hours, 5 minutes and 3 seconds)

s := r mod 60; r := r div 60;

// now s == 30, r == 3005

m := r mod 60; r := r div 60;

// now m == 5, r == 50

Now, next line as you have it

{

h := r mod 60; r := r div 24;

// now h == 50, r = 2

}

Here the same with 'mod 24'

{

h := r mod 24; r := r div 24;

// now h == 2, r = 2

}

BTW, I don't believe this has to do with the main problem ...

Regards,

ZOPPO

I said "we are going from minutes to hours," which wasn't what you were pointing out (we are going from hours to days).

Yes, this doesn't have anything to do with the issue, but pretty important nevertheless. Thanks for correcting me. I'll divide the points upon resolving the other issue.

you are wrong and Zoppo was right. Modulo 60 returns a number between 0 and 59 what obviously is wrong when counting hours.

>>>> BTW, I don't believe this has to do with the main problem ...

Probably not as the days were 0, right?

>>>> Can someone tell me why the following code doesn't work ?

I have a guess. I'll assume the counter counts in nanoseconds. Then for one day the QPC would count about 86 billion what is 20 times of a 32 bit maximum. I assume that mod and div return a 32 bit integer. So in 3 days we will have 60 times a 32bit divided by 60. Hence we have reached the 32bit boundary stored in the seconds.

Yes, exactly ...

About the 32-bit problem - I agree with itsmeandnobodyelse ... maybe declaring the 'r' as int64 could solve the problem.

ZOPPO

you have a mix of types of variables

and maybe they all get converted to word on the line

r := t div f; // in seconds

1. The trouble was in the type of 'r', which is limited at 65535, but a day has 86400 seconds. I have revised it to int64.

2. The count is a tick (frequency ticks per second), and not really nanoseconds or microseconds. In a 2.4Ghz machine you get 2400000000 ticks per second.

Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.

http://jongampark.wordpress.com/2008/04/19/weirdness-of-the-high-resolution-counter-ie-queryperformancecounter/