Power management method that can provide frequency frame by frame to drive video decoder
A power management method includes: receiving an input video signal, wherein the input video signal comprises multiple frames; extracting at least one video feature from each of the multiple frames; performing a prediction operation according to the at least one video feature to generate a coarse prediction result, wherein the coarse prediction result is indicative of a value of a required power of the each of the multiple frames; and for the each of the multiple frames, adjusting a frequency of a video decoder according to a refined prediction result to drive the video decoder, wherein the refined prediction result is an error calibrated version of the coarse prediction result.
1 . A power management method, comprising:
receiving an input video signal, wherein the input video signal comprises multiple frames;
extracting at least one video feature from each of the multiple frames, wherein the at least one video feature comprises codec type, frame size, resolution, frame rate, profile, level, frame type, or a combination thereof;
performing a prediction operation according to the at least one video feature to generate a coarse prediction result; and
for the each of the multiple frames, adjusting a frequency of a video decoder according to a refined prediction result to drive the video decoder, wherein the refined prediction result is an error calibrated version of the coarse prediction result, and is indicative of a value of a required power of the each of the multiple frames;
wherein the power management method further comprises:
comparing a value of an actual power corresponding to the each of the multiple frames with the value of the required power to generate a comparison result; and
performing an error calibration operation upon the coarse prediction result according to the comparison result, to generate the refined prediction result.
2 . The power management method of claim 1 , further comprising:
after the frequency of the video decoder is adjusted, storing the value of the actual power corresponding to the each of the multiple frames into a storage device.
3 . A non-transitory machine-readable medium for storing a program code, wherein when loaded and executed by a processor, the program code instructs the processor to execute software modules comprising:
a video driver, arranged to receive an input video signal, and extract at least one video feature from each of multiple frames, wherein the input video signal comprises the multiple frames, and the at least one video feature comprises codec type, frame size, resolution, frame rate, profile, level, frame type, or a combination thereof;
a prediction module, comprising a prediction model, wherein the prediction model is arranged to perform a prediction operation according to the at least one video feature to generate a coarse prediction result; and
a resource controllable module, arranged to adjust a frequency of a video decoder according to a refined prediction result for the each of the multiple frames, to drive the video decoder, wherein the refined prediction result is an error calibrated version of the coarse prediction result, and is indicative of a value of a required power of the each of the multiple frames;
wherein the prediction module further comprises:
an error detection module, arranged to compare a value of an actual power corresponding to the each of the multiple frames with the value of the required power to generate a comparison result; and
a calibration module, arranged to perform an error calibration operation upon the coarse prediction result according to the comparison result, to generate the refined prediction result.
4 . The non-transitory machine-readable medium of claim 3 , wherein after the frequency of the video decoder is adjusted, the video decoder stores the value of the actual power corresponding to the each of the multiple frames into a storage device.