QueryPerformanceCounter() vs. GetTickCount()
Author
QueryPerformanceCounter()
After a very exhaustive search I've looked at a way to make a timer for two things;
-To time a one-time event like a menu fading in.
-To time a "metronome" event like sprite animations.
So.. I looked at the QueryPerfomanceCounter() and found ALOT of C# code examples. Only after a while did I find this. (How to use QueryPerformanceCounter() to get very accurate timed events.)
GetTickCount()
The other method when I searched for gave a nice simple reference. (As well as some links to QueryPerformanceCounter, and QueryPerformanceFrequency ><)
My questions is this;
Are either of these accurate to run maybe around.. 33.3333 ms or less? Is the whole "make sure to check if QueryPerformanceCounter exists!" really important since the pentium is like.. olddd? What happens when someone uses a AMD processor with my game/timer? Finally.. I've heard alot of "bad" comments about QueryPerformanceCounter.. are they true? What about overloading a timer? I'm sorta new to Win32.
Thanks for any help!
QueryPerformanceCounter has in general more than enough precision, which is definately a good thing. The problem however with QueryPerformanceCounter is that it's accuracy is increasingly becoming subject to degradation on modern processors that use power saving features, and this includes many modern desktop cpu's.
I've never used GetTickCount for timing purposes so I can't comment on how suitable it is for this purpose. Instead of GetTickCount, I normally use timeGetTime, along with timeBeginPeriod and timeEndPeriod. For most purposes in the games I've made, a resolution of 5 ms has been more than suitable, and timeGetTime can be set to that or lower.
Perhaps the best solution is to use a combination of the methods, one as a rough check against the other.
I've never used GetTickCount for timing purposes so I can't comment on how suitable it is for this purpose. Instead of GetTickCount, I normally use timeGetTime, along with timeBeginPeriod and timeEndPeriod. For most purposes in the games I've made, a resolution of 5 ms has been more than suitable, and timeGetTime can be set to that or lower.
Perhaps the best solution is to use a combination of the methods, one as a rough check against the other.
There have been a couple threads floating arround the forums recently about this.
One problem I can attest to about QPC, is that AMD64x2 systems (Pentium D too?) seem to have problems with it. This comes from dual-core systems having 2 high performance timers, and so the application gets conflicting timing data when the thread is switched between processors.
-I have heard there are fixes for it, but I have yet to find one that removes the problem in all cases (assuming external attempt to fix the problem).
There still exist programs using QPC and RSDC that get buggy results.
Though you can still use QPC if you make sure that whatever thread your timer code is running in, you make your program set its own affinity to a single processor.
Answering some of your questions
-QPC is acurate to less than 1ms.
-GetTickCount() seems to vary. But has a max resolution of 1ms, but is often several ms.
-timeGetTime() can be forced to have a 1ms resolution.
One problem I can attest to about QPC, is that AMD64x2 systems (Pentium D too?) seem to have problems with it. This comes from dual-core systems having 2 high performance timers, and so the application gets conflicting timing data when the thread is switched between processors.
-I have heard there are fixes for it, but I have yet to find one that removes the problem in all cases (assuming external attempt to fix the problem).
There still exist programs using QPC and RSDC that get buggy results.
Though you can still use QPC if you make sure that whatever thread your timer code is running in, you make your program set its own affinity to a single processor.
Answering some of your questions
-QPC is acurate to less than 1ms.
-GetTickCount() seems to vary. But has a max resolution of 1ms, but is often several ms.
-timeGetTime() can be forced to have a 1ms resolution.
Author
1 millisecond = 0.001 seconds
1/30 = "maximum" draw rate = 0.0333 (repeating)
33 ms resolution, timeGetTime seems attractive to me.. I'm going to try to use it. Thanks for all the info! ^^;
1/30 = "maximum" draw rate = 0.0333 (repeating)
33 ms resolution, timeGetTime seems attractive to me.. I'm going to try to use it. Thanks for all the info! ^^;
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
