GetTickCount and finding the time between to events

[warz]

This can relate to bot uptimes, or times between issued commands, or whatever, but my question is.. in order to find out how much time has elapsed since (for example) an application has been executed, you'd do something like:

Code Select Expand


dword startedat = gettickcount();
...
dword howLongSinceNow = gettickcount();
dword timeElapsed = howLongSinceNow - startedat;


would this be correct? i was doing this, and was receiving strange return values...

[Joe[x86]]

Yes, but you might want to use long.

[K]

The code you have is correct.

Yes, but you might want to use long.

so instead of using a DWORD, which is an unsigned 32 bit integer, which is what GetTickCount() returns, he should use a long, which is a signed 32 bit integer which may not be large enough to hold the return value from GetTickCount()?

[MyndFyre]

so instead of using a DWORD, which is an unsigned 32 bit integer, which is what GetTickCount() returns, he should use a long, which is a signed 32 bit integer which may not be large enough to hold the return value from GetTickCount()?

Shh, Joe is trying to make up for incompetence by typing with only one letter visible. 

would this be correct? i was doing this, and was receiving strange return values...

Remember that a system tick is not necessarily correspondent to a second.

[UserLoser]

You may recieve funny values if your system has been up for >47 days.  See QueryPerformanceCounter and QueryPerformanceFrequency

[Joe[x86]]

The code you have is correct.

Yes, but you might want to use long.

so instead of using a DWORD, which is an unsigned 32 bit integer, which is what GetTickCount() returns, he should use a long, which is a signed 32 bit integer which may not be large enough to hold the return value from GetTickCount()?

Excuse my stupidity. =p

[Spht]

This can relate to bot uptimes, or times between issued commands, or whatever, but my question is.. in order to find out how much time has elapsed since (for example) an application has been executed, you'd do something like:

Code Select Expand


dword startedat = gettickcount();
...
dword howLongSinceNow = gettickcount();
dword timeElapsed = howLongSinceNow - startedat;


would this be correct? i was doing this, and was receiving strange return values...

Gettickcount returns system up time in milliseconds as an unsigned 32 bit integer, so can therefore only go up to 0xffffffff (4294967295 milliseconds, which as UL mentioned is about 47 days), after that it begins at 1 again.  So if between the two time samples the tick resets and you try to find "timeElapsed," you will get a negative number.

That should clear things up.

As UL mentioned, you may want to look into QueryPerformanceCounter and QueryPerformanceFrequency, from which uptime can be computed to 64 bit accuracy.

[warz]

This can relate to bot uptimes, or times between issued commands, or whatever, but my question is.. in order to find out how much time has elapsed since (for example) an application has been executed, you'd do something like:

Code Select Expand


dword startedat = gettickcount();
...
dword howLongSinceNow = gettickcount();
dword timeElapsed = howLongSinceNow - startedat;


would this be correct? i was doing this, and was receiving strange return values...

Gettickcount returns system up time in milliseconds as an unsigned 32 bit integer, so can therefore only go up to 0xffffffff (4294967295 milliseconds, which as UL mentioned is about 47 days), after that it begins at 1 again.  So if between the two time samples the tick resets and you try to find "timeElapsed," you will get a negative number.

That should clear things up.

As UL mentioned, you may want to look into QueryPerformanceCounter and QueryPerformanceFrequency, from which uptime can be computed to 64 bit accuracy.

I now use QueryPerformanceCounter, and QueryPerformanceFrequency

[Adron]

Gettickcount returns system up time in milliseconds as an unsigned 32 bit integer, so can therefore only go up to 0xffffffff (4294967295 milliseconds, which as UL mentioned is about 47 days), after that it begins at 1 again.  So if between the two time samples the tick resets and you try to find "timeElapsed," you will get a negative number.

Actually you will not get a negative number. You will get the correct number for any timeElapsed interval shorter than 47 days. Whether the timer has wrapped during the interval or not does not matter.