CN103955266B - The low power consumption design method of Sink load estimation is moved based on Android - Google Patents

Abstract

The mobile Sink node low power consumption design method that the low power consumption design method of Sink load estimation belongs in wireless sensor network is moved, especially by the low power consumption design method of cpu load prediction based on Android；The method obtains the load load_cur of current CPU by the CPU driver of Android, update the historic load value in history_load [3] array, judge that the first two time period cpu load history value history_load [1] and the history_load [2] of current slot in history_load [3] array are the most equal, and calculate current best initial weights α respectively_best, then by being calculated cpu load value load_next of subsequent time period, finally call _ _ cpufreq_driver_target function, according to the cpu load value load_next regulation CPU operating frequency of prediction；The present invention moves the low power consumption design method method of Sink load estimation based on Android, it is possible to achieve the dynamic adjustment of weights, improves load estimation precision further, thus selects more suitably CPU to calculate frequency and reduce power consumption.

Description

The low power consumption design method of Sink load estimation is moved based on Android

Technical field

The movement that the low power consumption design method of Sink load estimation belongs in wireless sensor network is moved based on Android Sink node low power consumption design method, especially by the low power consumption design method of cpu load prediction.

Background technology

At present, in terms of cpu load prediction, system based on Android uses the most simple and practical PAST and calculates Method, but the cpu load forecast error of this algorithm is bigger, thus most suitable CPU work can not be selected according to the load of prediction Frequency, so power consumption cannot ensured to reduce while performance；In order to improve load estimation precision, the index of Weight number adaptively Averagely (Dynamic Exponential average, DEXP) algorithm is suggested and carries out cpu load prediction, it was predicted that precision is Improve, but this algorithm predicts be supported at peak value that to change ratio shallower, but error is the biggest, although reduce power consumption with sacrificial Domestic animal performance is cost；In order to improve load estimation precision further, linear prediction (Linear Prediction, LP) algorithm quilt It is contemplated, although this algorithm is effectively improved load estimation precision, but its weights are fixing, it is impossible to load Change and dynamically adjust, it is impossible to ensure that its weight is current optimum.

Summary of the invention

In order to solve the problems referred to above, the invention discloses a kind of low-power consumption moving Sink load estimation based on Android Method for designing, the method can realize the dynamic adjustment of weights, improve load estimation precision further, thus select more suitably CPU calculates frequency and reduces power consumption.

The object of the present invention is achieved like this:

Move the low power consumption design method of Sink load estimation based on Android, comprise the following steps:

Step one: obtained the load load_cur of current CPU by the CPU driver of Android, enters step 2；

Step 2: the cpu load load_cur obtained by step one updates the history in history_load [3] array and bears Load value, update mode is assigned to for history_load [1] is assigned to history_load [2], history_load [0] History_load [1], load_cur are assigned to history_load [0], make to preserve all the time in history_load [3] array Current slot and the cpu load value of the first two time period, enter step 3；

Step 3: judge the first two time period cpu load history value of current slot in history_load [3] array History_load [1] and history_load [2] is the most equal, if:

No, seek current best initial weights by below equation,

It is to setValue be 0.5,

Enter step 4；

Step 4: utilize step 3 to obtain, by the load value of below equation prediction CPU subsequent time period,

After obtaining cpu load value load_next of subsequent time period, enter step 5；

Step 5: call _ _ cpufreq_driver_target function, adjusts according to cpu load value load_next of prediction Joint CPU operating frequency.

The present invention moves the low power consumption design method of Sink load estimation based on Android, it is possible to achieve weights dynamic Adjust, improve load estimation precision further, thus select more suitably CPU to calculate frequency and reduce power consumption.

Accompanying drawing explanation

Fig. 1 is that the present invention moves the low power consumption design method flow chart of Sink load estimation based on Android.

Detailed description of the invention

Below in conjunction with the accompanying drawings the specific embodiment of the invention is described in further detail.

The low power consumption design method moving Sink load estimation based on Android of the present embodiment, flow chart such as Fig. 1 institute Show.The method comprises the following steps:

Step one: obtained the load load_cur of current CPU by the CPU driver of Android, enters step 2；

Step 2: the cpu load load_cur obtained by step one updates the history in history_load [3] array and bears Load value, update mode is assigned to for history_load [1] is assigned to history_load [2], history_load [0] History_load [1], load_cur are assigned to history_load [0], make to preserve all the time in history_load [3] array Current slot and the cpu load value of the first two time period, enter step 3；

Step 3: judge the first two time period cpu load history value of current slot in history_load [3] array History_load [1] and history_load [2] is the most equal, if:

No, seek current best initial weights by below equation,

It is to setValue be 0.5,

Enter step 4；

Step 4: utilize step 3 to obtain, by the load value of below equation prediction CPU subsequent time period,

After obtaining cpu load value load_next of subsequent time period, enter step 5；

Step 5: call _ _ cpufreq_driver_target function, adjusts according to cpu load value load_next of prediction Joint CPU operating frequency.

Above step is done finer explanation: the frequency regulation work of android system is all in CPUFreq subsystem Complete, conservative, ondemand, userspace, powersave and performance five defined in this subsystem Planting chirping strategies (cpufreq_policy), every kind of strategy has the frequency modulator (cpufreq_governor) of oneself, and these are adjusted Frequently device all can call dbs_timer_init () function and initialize a work queue postponed, every identical delay just meter The work calculating cpu load joins in this queue, and cpu load calculates and called dbs_check_cpu by do_dbs_timer () () realizes, and the realization of SAWDLP algorithm is also in dbs_check_cpu () function, the current CPU's got in this function Load value is cur_load, then load Weight number adaptively linear prediction algorithm code is as follows:

/ * renewal historic load */

for(i=1;i>=0;i--)

history_load[i+1] = history_load[i];

history_load[0] = cur_load;

/ * ask best initial weights */

if(history_load [1] != history_load [2])

alpha= 10*( history_load [0]-history_load [2]) /

(history_load [1]-history_load [2]);

else

alpha = 5;

if(alpha > 10)

alpha %= 10;

/ * predict next load */

next_load = (alpha*history_load [0] + (10 - alpha)*history_load [1]) /10;

In order to avoid real arithmetic, amplified 10 times when calculating best initial weights, then reduced 10 when prediction load Times.Next_load is the cpu load value of the subsequent time period of prediction, selects a suitable frequency then by this load Call _ _ cpufreq_driver_target function realize frequency modulation operation.

Contrasting with existing method, result shows: predicted by the linear prediction algorithm of Weight number adaptively of the present invention Cpu load value percentage error compared with actual cpu load is 20.03%, than PAST algorithm 55.01%, LP algorithm The 22.65% of 20.24% and DEXP algorithm will be low.

Claims (1)

1. move the low power consumption design method of Sink load estimation based on Android, it is characterised in that comprise the following steps:

Step one: obtained the load load_cur of current CPU by the CPU driver of Android, enters step 2；

Step 2: the cpu load load_cur obtained by step one updates the historic load in history_load [3] array Value, update mode is assigned to for history_load [1] is assigned to history_load [2], history_load [0] History_load [1], load_cur are assigned to history_load [0], make to preserve all the time in history_load [3] array Current slot and the cpu load value of the first two time period, enter step 3；

Step 3: judge the first two time period cpu load history value of current slot in history_load [3] array History_load [1] and history_load [2] is the most equal, if:

No, seek current best initial weights by below equation,

It is to setValue be 0.5,

Enter step 4；

Step 4: utilize step 3 to obtain, by the load value of below equation prediction CPU subsequent time period,

After obtaining cpu load value load_next of subsequent time period, enter step 5；

Step 5: call _ _ cpufreq_driver_target function, according to the cpu load value load_next regulation of prediction CPU operating frequency.