Application-directed DPM/DVS

For simplicity, we assume there is only one interactive application under consideration, which is usually true for handheld computers. Also, we assume application programming interfaces (APIs) are available to applications to change the system performance levels. The flowchart for inserting the performance level transition code is given in Fig. 3(a). First, the STD of the user interface is extracted, and major user interface event handlers/states are identified, through either post-implementation analysis or pre-implementation specifications. If the psychological model is used, the user interface content in each state is parsed to get the values of the parameters for user delay models, as described in the previous sections. The parameter values only depend on the current state. Based on the predicted delay, an appropriate performance level is selected for the state and the corresponding code is inserted at the end of the corresponding event handler.

The inserted code can either force the system back to a higher performance level after a time interval based on the predicted user delay, or just let the user input raise the performance level, depending on the delay overhead for performance level transition and user's delay tolerance.

Figure 3: (a) The flowchart for code insertion for DPM/DVS, and (b) power modes and mode transitions.