non-period timer

Quote:

---

Originally Posted by TheCPUWizard

Arjay,

WFSO is a valid alternate approach. But I do want to point out that multiple threads can each have their own message pump and avoid the interference you are talking about....

---

Agreed, you can use message pump in secondary threads. But unless the secondary threads need to process messages for other reasons (perhaps they load a STA COM object), I'm just suggesting why use a timer that requires message pumping when WFSO will do? :thumb:

The primary (potential) downside of WSFO is you are controlling the wait time and not the total loop time. Additional logic can take care of this, but can get increasingly complex if the processing time may exveed the loop time and you want to maintain long term accuracy.

Example: Assume you have some code that you want to run every 50mS. The machine is otherwise completely idle (Windows NEVER is). Most of the time your routine takes 1mS to execute, but every so often it takes 200mS. It is critical that after 24 hours, the routine has esxecuted 1728000 times (or as close as possible.... You need to play "catch up" if the timer falls behind.

My preference for true real-time timers such as desirect above is to use a hardward product to generate a hardware interrupt and execute the required logic (at least partially) in Kernel mode... but that is a whole 'nother story...

Thank you. I will try this in my project. :)