I have a system that needs to spend the majority of its time in a sleep mode (using internal 30kHz oscillator), wake to respond to some interrupts (either still running at 30kHz or 6MHz external crystal) and occasionally perform some high speed tasks (using the 6MHz external with a PLL --->
50MHz). I've neglected my power management to this point and now I am at the tail end of the project looking to comeback through and add clock management.I am having a problem coming up with an elegant way to manage this. My current scheme (which I have yet to implement because I feel it could be done cleaner) is like so:
struct{ char * pcFlag; char cState; }g_sPowerMangement[10];
char g_cCurrentClock;
char systemPowerChange(char * pcFlag,char cState){
//Set flag = 1 //Add request to g_sPowerManagement //If request is greater than current speed, change clock speed and all peripherals dependant upon it
}void systemPowerMangement(){
//Loop through g_sPowerManagement to see if any pcFlags are cleared, if cleared remove from g_sPowerManagement
//If no more requests for a current clock speed, fall back to next lowest and change all peripherals dependant upon it
//If no items remain in g_sPowerManagement go to sleep
}Basically the module that requires the extra speed will create a global flag and call the power change function with a pointer to the flag and the desired speed. The clock will then be adjusted. The power management function will be called repeatedly from an infinite loop in main. It ensures that the system is running at the lowest possible clock based on the requests held within the power management structure. When no more requests exist in the structure the system goes to sleep.
For the most part I think this is decent except in a module that has many exit points where I could potentially miss the clearing of the flag (keeping the system at the requested speed forever).
Does this sound like a reasonable approach? How do others handle this?
--------------------------------------- Posted through