CN105045367A - Android system equipment power consumption optimization method based on game load prediction - Google Patents
Android system equipment power consumption optimization method based on game load prediction Download PDFInfo
- Publication number
- CN105045367A CN105045367A CN201510024136.0A CN201510024136A CN105045367A CN 105045367 A CN105045367 A CN 105045367A CN 201510024136 A CN201510024136 A CN 201510024136A CN 105045367 A CN105045367 A CN 105045367A
- Authority
- CN
- China
- Prior art keywords
- game
- scene
- frequency
- android system
- cpu
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Power Sources (AREA)
Abstract
The invention discloses an Android system equipment power consumption optimization method based on game load prediction; the features are that the method comprises the following steps: 1, obtaining Android system equipment processor load in each game running phase, and predicting game running next frame processor load demand according to the obtained processor load; 2, obtaining present game access characteristics, present game running texture characteristics and game-user interaction degree information of the Android system equipment, and determining a present game running function scene; 3, carrying out CPU and GPU frequency and voltage adjustment through a preset DVFS frequency modulation module according to the present game running function scene. The method can reasonably distribute system resources according to different scenes and game performance demands of a user under the corresponding scene, thus reducing power consumption.
Description
Technical field
The invention belongs to android intelligent system equipment power dissipation optimisation technique field, particularly a kind of Android system equipment power dissipation optimization method based on game load estimation.
Background technology
Along with growth and the diversified development of function of application of hand-hold mobile device explosion type, power consumption, performance tradeoff are further outstanding.In order to alleviate this problem, solution was often based upon on the basis to system or task behavioural analysis in the past, but different application scenarioss and even use crowd can have a huge impact system or task behavior, therefore, the information fully excavating application scenarios carries out resource optimization to guidance system, reduce power consumption plays an important role.
In Samsung GalaxyS5, when electricity lower than 10% time screen will become black and white and close all nonessential services; The prior optimised power consumption mechanism be the introduction of for different scene, these mechanism all come from the power-saving technology PowerExtend of third party LucidLogix company.Its Part I is optimized for airmanship, be referred to as NavExtend, battery durable ability 25% can be extended under the scene using GPS navigation, after navigation way is determined, navigation behavior is predictable, the figure load required for GPS navigation can be regulated as required, for simple picture, frame refreshing frequency can be reduced; The second optimisation technique is WebExtend, can extend cruising time 25% equally when surfing the Net, its principle and the former is similar, is also the figure load of intelligent decision browser, adjustresources as required; Last then be for game contour load applications custom-designed GameExtend, battery durable 50% can be extended.
In the electronic product such as mobile phone, panel computer, the part that game has become in people's daily life, but game developer often when pursuing game visual effect and Consumer's Experience, not can take into account the reasonable distribution of software and hardware resources on opponent's machine platform, blindly improve frame per second.At IBMNotebook platform testing results game QuakeII (acquiescence opens soft playing up), under CPU599MHz and 1298MHz frequency, (X-axis is time shaft to its frame per second as shown in Figure 1, Y-axis is frames/sec), the frame per second of playing when operating in 1298MHz reaches as high as more than 70fps, the frame per second of playing when operating in 599MHz minimumly may reach 10fps, the game fluctuation of load is larger, to resource requirement change also greatly, but research [2] shows, after frame per second reaches certain critical value, improve frame per second and cannot improve Consumer's Experience, this will cause resource excessively to ask for, thus the power consumption/performance tradeoff on intelligent mobile phone platform is highlighted further.Domestic and international researcher is relatively many for the DVFS research of application-specific scene, but technique Application comparison maturation is at video playback decoding field [DietrichB, ChakrabortyS.Managingpowerforclosed-sourceandroidosgames bylightweightgraphicsinstrumentation [C] //NetworkandSystemsSupportforGames (NetGames), 201211thAnnualWorkshopon.IEEE, 2012:1-3.], DVFS for scene of game studies then relatively deficient and does not obtain comparatively quantum jump.DVFS research for game application is mainly divided into PC to hold and mobile terminal from research platform, is divided into and increases income game and close source game research object.
The existing DVFS for game application studies [GuY, ChakrabortyS, OoiWT.GamesareupforDVFS [C] //Proceedingsofthe43rdannualDesignAutomationConference.ACM, 2006:598-603.] multiaspect to be PC platform, concentrate on the research of game that windows platform increases income, the acquisition of game source code and to study difficulty carefully large, and from point of view of practicability, Research Significance is little.
For game of increasing income, because source code can obtain, so in the extraction having focused on the architectural feature in the game picture frame embodied in a frame code of research, the achievement that research obtains has developed irreplaceable effect for instructing game developer under awareness of saving energy.But for as third party, want to find out a kind of researcher for closing source game and reduce the universal method of power consumption by analyzing game behavior in operating system one-level, its research method has no way of using for reference.The present invention comes therefrom.
Summary of the invention
The object of this invention is to provide a kind of Android system equipment power dissipation optimization method based on game load estimation, solve Android system equipment game power consumption in prior art generally high, reduce the technical matterss such as device battery availability.The present invention is based on game load estimation Android system equipment power dissipation optimization method can according to different scenes and user under corresponding scene to the requirement of game performance, reasonable distribution system resource reduces the object of power consumption to reach.
In order to solve these problems of the prior art, technical scheme provided by the invention is as follows:
Based on a method for the reduction Android system equipment power dissipation of game load estimation, it is characterized in that said method comprising the steps of:
(1) obtain the processor load of each stage Android system equipment of game running, run the processor load demand of next frame according to processor load forecasting game;
(2) obtain the memory access characteristic of the going game of Android system equipment, texture features that going game runs, game and user interactive degree information, judge the function scene that going game runs;
(3) the function scene run according to going game carries out frequency and voltage adjustment according to predetermined DVFS FM module to CPU and GPU.
In preferred technical scheme, described method step (1) is poor by the execution time obtaining current and last the eglSwapBuffers () function called in bottom EGL of game process, quantizes to determine that CPU completes all calculating of present frame and the total load value called to GPU is played up in transmission.
In preferred technical scheme, the mistiming that described method step (1) is mapped by the number of times and current and last texture coordinate obtaining game process texture picture pinup picture, quantize to determine GPU load value.
In preferred technical scheme, in described method step (2), the function scene of game running divides into game loading scenario, scene is carried out in game, scene, Game Menu scene and game interaction scene are exited in game.
In preferred technical scheme, add up a period of time Memory Allocation number of requests identification game loading scenario by memory allocation function in Android system device core in described method step (2) and scene is exited in game:
1) when the distribution request amount of application program in measurement period to internal memory exceedes or equal predetermined allocation threshold, and releasing memory amount is lower than predetermined release threshold value, then judge that the current function scene of game running is as game loading scenario;
2) when application program in measurement period to the distribution request amount of internal memory lower than predetermined allocation threshold, and releasing memory amount exceedes or equals predetermined release threshold value, then judge that the current function scene of game running exits scene as game;
3) all the other situations, do not judge.
In preferred technical scheme, before and after the texture mapping order of magnitude and the continuous a few frame texture quantity of playing time in described method step (2) by adding up game running in timing statistics, the accumulated value of difference judges that the current function scene of game running is that scene is carried out in Game Menu scene or game:
I) the predetermined texture order of magnitude is less than when adding up game texture mapping quantity in timing statistics, and continuous a few frame texture number change speed lower than the scene of predetermined variation speed, then judges that the current function scene of game running is as Game Menu scene in timing statistics;
II) other situations all judge that the current function scene of game running is as non-gaming menu scene.
In preferred technical scheme, in described method step (2), user interactions degree information is realized by the message mechanism of Android system equipment, wherein by message processing time dutycycle
as the index of task significance during measurement user interactions; In formula, cpu_time represents that the every CPU that to call for twice in DVFS frequency modulation algorithm carries out frequency adjustment time interval to CPU of Android system is for the treatment of the time of task; Msg_time represents that the every CPU that to call for twice in frequency modulation algorithm carries out frequency adjustment time interval to CPU of system is for the time of main thread processing messages; When ρ is greater than or equal to predetermined value, then judge that the current function scene of game running is game interaction scene; Otherwise judge that the current function scene of game running is not as game interaction scene.
In preferred technical scheme, in described method step (3), DVFS FM module adopts ondemand algorithm to carry out frequency modulation in accordance with the following steps:
A) when cpu busy percentage is greater than 90%, cpu frequency is risen to supported maximum frequency; Otherwise carry out B);
B) judge cpu busy percentage whether between 70 ~ 80%, if so, then continue to be recycled to next and calculate frequency modulation cycle; Otherwise carry out step C);
C) judge whether CPU ongoing frequency is less than current supported minimum frequency; If so, ongoing frequency is risen to supported minimum frequency; Otherwise reduction ongoing frequency, continues to be recycled to next and calculates the frequency modulation cycle.
In preferred technical scheme, in described method step (A) when cpu busy percentage is greater than 90%, if the current game scene obtained is that scene is carried out in game interaction scene or game, then cpu frequency is risen to supported maximum frequency; Otherwise judge whether current C PU frequency is more than or equal to 80% of supported maximum frequency, if current C PU frequency is more than or equal to 80% of supported maximum frequency, then continues to be recycled to next and calculate the frequency modulation cycle; Otherwise current C PU frequency upgrading is to 80% of supported maximum frequency.
In preferred technical scheme, in described method step (3) when ρ is greater than or equal to predetermined value, then improve the frequency and voltage of CPU; Otherwise reduce the frequency and voltage of CPU.
In preferred technical scheme, described method step (3) if in the statistics game texture mapping quantity in timing statistics that obtains be less than the predetermined texture order of magnitude, and continuous a few frame texture number change speed is lower than the scene of predetermined variation speed in timing statistics, then the target frame rate arranging GPU by DVFS FM module is 30fps; Otherwise be 60fps by the target frame rate being arranged GPU by DVFS FM module.
Technical solution of the present invention is absorbed in the low-power consumption research of Android intelligent platform closing source game.The present invention using scene of game as research object.Different from power consumption research both domestic and external in the past, the present invention portrays as theoretical foundation using game behavior, and then different conditions (or is referred to as function scene under instructing game application scenarios, different conditions is different for CPU, GPU resource demand characteristics) division, final by means of DVFS technological means, realize fine granularity game optimised power consumption, finally construct the power consumption control scheme of scene of game perception.
DVFS and dynamic voltage frequency adjustment, dynamic voltage frequency adjustment technology is that the application program run according to chip is to the different needs of computing power, the running frequency of dynamic adjustments chip and voltage are (for same chip, frequency is higher, required voltage is also higher), thus reach energy-conservation object.
As shown in Figure 2, by the analysis and summary to existing achievement in research, following three aspect main contents are comprised to the power consumption control of game: be first load analysis, quantize processor load during game running; Next identifies game state, because different game state (load phase, menu stage, play stage etc.) is different to the demand of frame per second, in other words, also different to the demand of resource; Finally using this two-part work as the input of power managed module to carry out load estimation, thus estimate the loading demand of game next frame, the corresponding relation between final quantization load and frequency also designs DVFS algorithm.
Load definition
According to smart mobile phone game application, for CPU and GPU, suitable load is defined respectively to the different characteristics that CPU and GPU resource require.Load definition carries out load estimation to instruct the prerequisite of optimised power consumption.
In android system, the display such as game view, work of playing up finally are realized by OpenGLES, and OpenGLES can realize on different platforms, be actually one group of API specification, therefore, the function of OpenGLES to be realized at concrete platform, need to carry out localization.In android system, localization work is realized by EGL.
The android system work of playing up has software simulating and hardware implementing two kinds of modes, and concrete which kind of implementation of selecting finally is realized by the hook Hook function in EGL.Therefore, realize can running at two kinds of mode lower modules, must start with from Hook Function, obtain the data needed, realize a series of activities such as prediction and frequency modulation with this.
Definition game load is in fact the behavior wishing to be reflected by this load game, and then by analyzing game behavior, the action after prediction, calculating in advance and the corresponding frequency that predicts the outcome according to predicting the outcome, then carrying out frequency modulation.Therefore, the final purpose of definition game load is reflection game behavioural characteristic.The execution time difference of the eglSwapBuffers function under EGL storehouse can be used as game load.The game process operating in DalvikVM virtual machine can call the C/C++ storehouse of Native layer, by the EGL function of C/C++ library call to bottom by JNI.And when completing the calculating of a frame, game process can call eglSwapBuffers function and backbuffer and frontbuffer in the double buffers in display system is exchanged at every turn.Each exchange realizes the refreshing of a frame.EglSwapBuffers is called by game process at every turn, and make process itself does not control frame per second, but fast as far as possible calls eglSwapBuffers to improve frame per second to ensure Consumer's Experience.Therefore, the mistiming that eglSwapBuffers performs at every turn all should as input data.
EglSwapBuffers, while being called by game process, can be called by system process SurfacFlinger in same frame time.The game load definition that the present invention adopts should be the mistiming that eglSwapBuffers is called for twice by game process.Game load is the time interval that eglSwapBuffers is called by game process for twice.Adopt the Android4.0 platform of pair buffers, calling of the eglSwapBuffers () function in bottom EGL makes GPU exchange front and back buffer memory, waits for that the content of this frame will be shown by VSync signal in the display.So twice eglSwapBuffers () function is that CPU completes all calculating of present frame and the total load value called to GPU is played up in transmission by the time interval that game process calls.The time interval of eglSwapBuffers () function is called in processing time i.e. twice game of definition game one frame is thus the load value of present frame.On technology realizes, as long as introduce Hook Function by the standardized interface of adjustment to eglSwapBuffers () function, in time calling this function at every turn, recording a precise time stamp, just can obtain load value by calculating current time with the last time difference called.
To sum up, for general frequency modulation algorithm, cpu load is defined as the proportion of CPU within certain time interval shared by the rush hour.And for scene of game, as supplementing former cpu load concept, the present invention chooses the time interval the supplementing as the cpu load value of playing between front and back two frame of game picture.
GPU load estimation
Along with the development of technology, current intelligent mobile phone platform to gradually adopt on sheet GPU to complete playing up of figure, thus form the collaborative work mode of CPU and GPU, namely CPU played by COMPREHENSIVE CALCULATING in AI, logic, physics and user input, then these rendering tasks are given GPU and are completed by the data structure comprised in decision next frame and rendering task.
By analyzing the graphics rendering pipeline of GPU, the following factor affecting GPU workload may be there is: resolution, model accuracy, illumination, texture and frame per second.Wherein GPU graph rendering pipeline as shown in Figure 3.Vertex processor is for running vertex shading program in each summit.Wherein complete geometric transformation and view transformation.Fragment processor read status information, polygon information and conversion vertex data, mainly complete rasterisation and texture mapping and be finally created on the display data in frame buffer.Mixed cell is responsible for the mixing of data and color value after fragment processor process.Rendering result write frame buffer the most at last in a frame is medium to be shown.
In these influence factors, and be not all suitable as the variable describing GPU load variation tendency: resolution and illumination between difference game relatively may be larger on the impact of GPU load, but for same game, they are almost constant in the whole process of game; Frame per second is that the amount jointly determined with machine performance write by game coding person code, is a dependent variable; The fine degree of 3D model in the accuracy representing game of model, show as the number of surfaces of model, can represent GPU loading condition, but it obtains difficulty, cost is high, and final texture features of selecting describes the variable of GPU load as definition.This mainly owing to following some: first, texture is the picture being attached to model surface, and the number of times of pinup picture can represent the face number of 3D model and the precision of model to a certain extent; Secondly, rasterisation and fragment processing stage, texture coordinate maps process inherently consuming time, is the important component part of GPU load.
Function scene
Function scene distinguishes the different scenes namely determined based on scene complexity concept, and the requirement characteristic of scene complexity game representation here to CPU/GPU resource itself, the parameters such as available memory access characteristic and texture features (quantity and change) are weighed.Scene distinguishes the complexity expecting to be weighed scene by suitable method, to reach the object customizing different chirping strategies for different scene.
By the lot of experiments of playing to difference and theoretical analysis, utilize the factor of influence affecting the scene characteristics of game own, i.e. memory access characteristic, texture features, user interactions degree define the complexity of scene of game.Wherein memory access characteristic is mainly used in the loading of identification game and exits scene; Texture features determines complexity (texture quantity) and the pace of change (texture variations) of the ongoing different scenic picture of game, is mainly used in distinguishing menu scene and scene of game; And message mechanism is for analyzing the level of interaction of user to scene of game, to depict the impact of user operation on game behavior.
(1) differentiation that memory access characteristic is carried out is utilized
Application program loading and stage of exiting picture own simply, are memory-intensive.From internal memory angle, the life cycle of application program can be divided into three phases: the outbreak period to memory requirements, the stationary phase to memory requirements and the rapid drawdown phase to memory requirements.Respectively corresponding application program startup, run and exit three phases, as in application program life cycle to shown in physical memory demand graph of a relation.
By to memory allocation function monitoring in kernel, statistics a period of time Memory Allocation number of requests, thus judge games unloading phase and exit the stage, based on the game loading scenario of memory access characteristic with to exit scene decision table as shown in table 1:
Table 1 is distinguished according to memory access characteristic and is loaded and exit scene decision table
| Numbering | 0 | 1 | 2 | 3 |
| Request internal memory is many | Y | Y | N | N |
| Releasing memory is many | Y | N | Y | N |
| Scene | — | Load | Exit | Other |
Memory Allocation characteristic when loading according to application program, to the applications of internal memory by analyzing application program in measurement period, gets rid of the impact that Dalvik virtual machine garbage collection strategy reclaims Installed System Memory, determines the loading procedure of playing; Monitor memory management in linux system simultaneously, when application program exits, function _ _ free_pages (structpage*page is discharged to internal memory, calling unsignedintorder), carry out internal memory release quantitative statistics, the two combine common as based on memory access characteristic to game loading and the differentiation foundation exiting scene.
(2) texture features is utilized to distinguish
By the analysis of different scene of playing for difference, find game except loading and exiting the stage, the variation characteristic of game texture can the complex situations of game representation scene: in menu scene, less based on the change of normally mutual texture scene; In scene of game, large based on normally mutual texture variations.The linear relationship of the strong correlation that texture mapping quantity and GPU utilization factor show demonstrates the relation between texture variations and GPU utilization factor: in menu scene, texture quantity is almost constant, corresponding GPU utilization factor also remains unchanged, in scene of game, texture number change is comparatively large, and corresponding GPU utilization factor follows change.In game play, statistics game common texture features as differentiation scene of the accumulated value of difference before and after the texture mapping order of magnitude and continuous a few frame texture quantity in timing statistics.
Texture variations characteristic in the continuous a few frame of game is utilized to distinguish scene, and different target frame rate is set, ensure Consumer's Experience to greatest extent, when scene partitioning, only texture quantity is few and that texture variations is slow scene Recognition is menu scene, and supposes that its target frame rate is 30; Target frame rate corresponding to other scenes is 60, and main object reduces frame per second that game developer arranges causes energy dissipation unreasonable frame per second higher than 60.The decision table formed is as shown in table 2.
Table 2 utilizes texture features to the differentiation of scene and target frame rate decision table
(3) message dutycycle portrays user's request
Due to the singularity of interactive system, first, in system, the importance of task is different along with the difference of the degree of relationship of it and Consumer's Experience.Secondly, the urgency level of user's request also plays vital effect for the distribution of resource, therefore, in principle should preferential fast processing to the relevant task of response user operation, the background task that user's degree of concern is low can make CPU run under lower frequency.
Therefore use user interactions degree to weigh the importance definition of task, and then instruct DVFS technology.Its principle improves the running frequency of CPU, otherwise reduce the running frequency of CPU.First this research contents needs the evaluation criterion quantizing task significance.This project proposes a kind of simple and effective, normalized task significance and evaluates mathematical model.The relation of Message Processing and user interactions that this model makes full use of Android is evaluated task significance.
Namely the importance of portraying task use the degree of correlation of a kind of mathematical model quantification tasks and Consumer's Experience.In this definition " message processing time dutycycle ρ " as the index weighing task significance, ρ is defined as follows:
In formula, cpu_time and msg_time represents system every time of calling CPU in frequency modulation algorithm carries out frequency adjustment time interval to CPU for twice and being used for main thread processing messages for the treatment of time (non-Idle time) of task and CPU respectively.Message processing time dutycycle ρ not only can represent CPU for the treatment of the number of the time of message within a period of time, and provides a kind of normalized reference quantity being convenient to compare.
DVFS FM module
1 ondemand algorithm chirping strategies
Frequency modulation algorithm of the present invention is at Android existing ondemand frequency modulation algorithm, namely to use at present to modify in the basis of maximum ondemand algorithms.First a simple analysis introduction is done to ondemand algorithm.By analyzing the source code of ondemand algorithm, the process flow diagram carrying out frequency modulation of ondemand algorithm can be obtained as Fig. 6.
The Linux carrying out labor once Android bottom according to Fig. 6 carries the principle of work of frequency modulation algorithm ondemand algorithm.
(1) frequency modulation of ondemand algorithm is according to being cpu load, i.e. system load or cpu busy percentage, and the computing method of load are the idle time ratio that accounts for CPU T.T. in the past period.The frequency modulation of different levels is carried out according to the different ondemand of ratio.Core objective keeps cpu busy percentage at 70%-80%.
(2) ondemand algorithm carries out frequency modulation according to cpu load, and its principal feature is raising frequency rapidly, slow frequency reducing.When present load is greater than 90, when namely cpu busy percentage is more than 90%, ondemand by frequency adjustment to maximal value.1.2G is under this platform.If cpu busy percentage is between 70%-80%, then do not operate.If cpu busy percentage is less than 70%,
If ongoing frequency be less than ondemand support low-limit frequency, be 200M under this platform, so then by frequency upgrading to 200M.Otherwise ondemand will turn down cpu frequency slowly.
Ondemand algorithm carries out frequency modulation for whole system, carries out frequency modulation from the angle of cpu busy percentage, therefore, and the less pertinence of ondemand algorithm when carrying out frequency modulation for scene of game.Be mainly manifested in two aspects:
(1) poor to game frequency modulation specific aim, there is blindness
Because the frequency modulation foundation of ondemand bases oneself upon on the basis of cpu busy percentage, and play, owing to can utilize GPU to play up etc., and have the feature of interactivity, therefore, the utilization factor of CPU cannot reflect the feature of game.
In addition, android system does not pursue the fairness of task process, but pursues Consumer's Experience, so, as long as the foreground that cpu frequency can meet when game is carried out is experienced, can slowly process for background program.In this case, it is inequitable for utilizing cpu busy percentage to carry out frequency modulation concerning game, has blindness.Carry out frequency modulation according to game characteristic and then become very necessary.
(2) chirping strategies is too conservative, when cpu busy percentage is greater than 90%, set of frequency is too high
Experimentally, when cpu busy percentage is greater than 90%, ondemand is too conservative to the strategy of maximum frequency by frequency upgrading.When cpu busy percentage more than 90%, ondemand under linux intelligent platform directly by frequency upgrading to maximum 1.2G.And by amendment ondemand algorithm, under scene of game, when cpu busy percentage is more than 90%, be not frequency is directly risen to 1.2G, but reduce this standard, only directly rise to the frequency of 1G or 800M.In this case, according to the ruuning situation of the game obtained, find that situation of frame losing or card machine does not occur for it.The ruuning situation of game is still good, and Consumer's Experience is also without any impact.Therefore, can illustrate, ondemand chirping strategies for game time too conservative.
Based on the ondemand optimisation strategy of game behavior
According to the analysis of a upper joint, the algorithm of ondemand is to the less pertinence of game, and too conservative, and therefore DVFS chirping strategies is optimized according to game behavior.Main optimization two aspects, one is that ondemand is changed into frequency modulation based on game characteristic based on the strategy of cpu busy percentage frequency modulation, namely utilizes the predicted data of game load/based on the target frame rate of game state as frequency modulation foundation.Two is when play load and predicted data/target frame rate are greater than 90%, frequency is not directly set to mxm. 1.2G, but is set to 1G or 800M or lower.Show herein by experiment, the optimization based on the ondemand algorithm of game behavior is effective.
Relative to scheme of the prior art, advantage of the present invention is:
The present invention is absorbed in the optimised power consumption based on game load estimation on android system equipment, abandons forefathers' undue dependence for game source code in research process; By analyzing the characteristics of demand of CPU and GPU computational resource under difference in functionality scene game, and then the load definition of CPU/GPU is expanded.On the basis that function scene is distinguished, devise the DVFS frequency modulation algorithm of CPU and GPU respectively, the final CPU/GPU voltage frequency modulation scheme forming game function scene perception.The program achieves the identification to scene different during game running.The demand of different scenes to resource of game is different; User is under different scene of game, also different to the sensitivity of game performance performance, after successfully identifying the different scenes of game, just can according to different scenes and user under corresponding scene to the requirement of game performance, reasonable distribution system resource reduces the object of power consumption to reach.
Accompanying drawing explanation
Below in conjunction with drawings and Examples, the invention will be further described:
Fig. 1 is the frame rate of game under 599MHz and 1298MHz;
Fig. 2 is the block diagram of the reduction Android system equipment power dissipation that the present invention is based on game load estimation;
Fig. 3 is GPU graph rendering piping drawing;
Fig. 4 is to physical memory demand graph of a relation in application program life cycle.
Fig. 5 is the Android system equipment power dissipation optimization architecture figure that the present invention is based on game load estimation.
Fig. 6 is ondemand frequency modulation algorithm flow chart.
Embodiment
Below in conjunction with specific embodiment, such scheme is described further.Should be understood that these embodiments do not limit the scope of the invention for illustration of the present invention.The implementation condition adopted in embodiment can do further adjustment according to the condition of concrete system, and not marked implementation condition is generally the condition in normal experiment.
Embodiment
1, experiment porch
What adopt in experiment is ODROID development board, and its CPU adopted is SAMSUNGEXYNOS4412, and it is the Cortex-A9QuadCore based on ARM company, supports DVFS, and support the frequency band of frequency modulation be 200M, 400M, 600M, 800M, 1G, 1.2G.There is the GPU of Mali-400QuadCore simultaneously, support firmly to play up.The version of android system is Android-4.04.
2 optimum configurations
The platform put up is tested jetpack game, checks that its forecast result of model, game state distinguish effect and energy-saving effect.
The accuracy rate fiducial interval of setting model prediction is 95%, and namely predicated error is less than 5% judgement prediction accurately.During setting game state is distinguished, the target frame rate 60FPS set under game state is then presented at this state, the target frame rate 30FPS under menu state then display menu state.Simultaneously, set and in different situations saving energy test is carried out to jetpack game, setting load arranges the different highest frequency (600M risen when being greater than 90%, 800M, 1G) test, simultaneously target setting frame per second is 60FPS, tests according to different settings, with the frequency modulation method comparison of frequency modulation algorithm of the present invention and original ondemand.
3 experimental results
3.1 model predictions and state distinguish result
In model prediction module, print the accuracy rate of the online real-time estimate of model, result shows, predictablity rate of the present invention is very close with its degree of fitting, and predictablity rate is in about 90% fluctuation.And effect is distinguished for state, also achieve and distinguish effect more accurately.
Confirm through experiment, when game state, frame per second is 60FPS, is 30FPS at menu state.When current state is game state, its target frame rate is 60FPS; Just clicked menu when occurring, now detect and enter menu state by game state, therefore, target frame rate is reduced to 30FPS from 60FPS.Show the correctness that game function scene state is distinguished.
6.3.2 energy-saving effect
The frequency modulation algorithm of frequency modulation algorithm of the present invention (different maximum frequencies is set) and original ondemand is used to compare according to different optimum configurations.Obtain result as table 3.
Table 3 energy-saving effect table
Analytical table 6, average fractional energy savings can reach 10%.And, display data can be known to utilize display APP to judge, in frequency-modulating process, when target setting frame per second is 60, when game state, actual frame per second remains on about 60FPS always, therefore, utilizing the system of this power consumption optimization method, the energy consumption of 10% can be saved when playing than ondemand, and the operation of modules does not affect the runnability of game substantially, Consumer's Experience can be ensured simultaneously.
This experiment continues through multiple game and tests the loading scenario, menu scene and the scene of game that confirm effectively to identify game.The game tested at present has: cut rope, AngryBirds, Zu Ma, mad jet plane, Flyboys, racing car etc.The conversion of actual play test Scene comprise game from scene of game to menu scene conversion, to play from menu scene to scene of game conversion and game change from menu scene to scene of game from scene of game to menu scene conversion and game.
Above-described embodiment, only for technical conceive of the present invention and feature are described, its object is to person skilled in the art can be understood content of the present invention and implement according to this, can not limit the scope of the invention with this.All equivalent transformations of doing according to Spirit Essence of the present invention or modification, all should be encompassed within protection scope of the present invention.
Claims (10)
1., based on a method for the reduction Android system equipment power dissipation of game load estimation, it is characterized in that said method comprising the steps of:
(1) obtain the processor load of each stage Android system equipment of game running, run the processor load demand of next frame according to processor load forecasting game;
(2) obtain the memory access characteristic of the going game of Android system equipment, texture features that going game runs, game and user interactive degree information, judge the function scene that going game runs;
(3) the function scene run according to going game carries out frequency and voltage adjustment according to predetermined DVFS FM module to CPU and GPU.
2. Android system equipment power dissipation optimization method according to claim 1, it is characterized in that described method step (1) is poor by the execution time obtaining current and last the eglSwapBuffers () function called in bottom EGL of game process, quantize to determine that CPU completes all calculating of present frame and the total load value called to GPU is played up in transmission.
3. Android system equipment power dissipation optimization method according to claim 1, it is characterized in that quantizing to determine GPU load value the mistiming that described method step (1) is mapped by the number of times and current and last texture coordinate obtaining game process texture picture pinup picture.
4. Android system equipment power dissipation optimization method according to claim 1, is characterized in that the function scene of game running in described method step (2) divides into game loading scenario, scene is carried out in game, scene, Game Menu scene and game interaction scene are exited in game.
5. Android system equipment power dissipation optimization method according to claim 1, is characterized in that adding up a period of time Memory Allocation number of requests identification game loading scenario by memory allocation function in Android system device core in described method step (2) and scene is exited in game:
1) when the distribution request amount of application program in measurement period to internal memory exceedes or equal predetermined allocation threshold, and releasing memory amount is lower than predetermined release threshold value, then judge that the current function scene of game running is as game loading scenario;
2) when application program in measurement period to the distribution request amount of internal memory lower than predetermined allocation threshold, and releasing memory amount exceedes or equals predetermined release threshold value, then judge that the current function scene of game running exits scene as game;
3) all the other situations, do not judge.
6. Android system equipment power dissipation optimization method according to claim 1, when to it is characterized in that in described method step (2) by adding up game running in timing statistics, before and after the game texture mapping order of magnitude and continuous a few frame texture quantity, the accumulated value of difference judges that the current function scene of game running is that scene is carried out in Game Menu scene or game:
I) the predetermined texture order of magnitude is less than when adding up game texture mapping quantity in timing statistics, and continuous a few frame texture number change speed lower than the scene of predetermined variation speed, then judges that the current function scene of game running is as Game Menu scene in timing statistics;
II) other situations all judge that the current function scene of game running is as non-gaming menu scene.
7. Android system equipment power dissipation optimization method according to claim 1, is characterized in that in described method step (2), user interactions degree information is realized by the message mechanism of Android system equipment, wherein by message processing time dutycycle
as the index of task significance during measurement user interactions; In formula, cpu_time represents that the every CPU that to call for twice in DVFS frequency modulation algorithm carries out frequency adjustment time interval to CPU of Android system is for the treatment of the time of task; Msg_time represents that the every CPU that to call for twice in frequency modulation algorithm carries out frequency adjustment time interval to CPU of system is for the time of main thread processing messages; When ρ is greater than or equal to predetermined value, then judge that the current function scene of game running is game interaction scene; Otherwise judge that the current function scene of game running is not as game interaction scene.
8. Android system equipment power dissipation optimization method according to claim 1, is characterized in that in described method step (3), DVFS FM module adopts ondemand algorithm to carry out frequency modulation in accordance with the following steps:
A) when cpu busy percentage is greater than 90%, cpu frequency is risen to supported maximum frequency; Otherwise carry out B);
B) judge cpu busy percentage whether between 70 ~ 80%, if so, then continue to be recycled to next and calculate frequency modulation cycle; Otherwise carry out step C);
C) judge whether CPU ongoing frequency is less than current supported minimum frequency; If so, ongoing frequency is risen to supported minimum frequency; Otherwise reduction ongoing frequency, continues to be recycled to next and calculates the frequency modulation cycle.
9. Android system equipment power dissipation optimization method according to claim 8, to it is characterized in that in described method step (A) when cpu busy percentage is greater than 90%, if the current game scene obtained is that scene is carried out in game interaction scene or game, then cpu frequency is risen to supported maximum frequency; Otherwise judge whether current C PU frequency is more than or equal to 80% of supported maximum frequency, if current C PU frequency is more than or equal to 80% of supported maximum frequency, then continues to be recycled to next and calculate the frequency modulation cycle; Otherwise current C PU frequency upgrading is to 80% of supported maximum frequency.
10. Android system equipment power dissipation optimization method according to claim 1, it is characterized in that described method step (3) if in the statistics game texture mapping quantity in timing statistics that obtains be less than the predetermined texture order of magnitude, and continuous a few frame texture number change speed is lower than the scene of predetermined variation speed in timing statistics, then the target frame rate arranging GPU by DVFS FM module is 30fps; Otherwise be 60fps by the target frame rate being arranged GPU by DVFS FM module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510024136.0A CN105045367A (en) | 2015-01-16 | 2015-01-16 | Android system equipment power consumption optimization method based on game load prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510024136.0A CN105045367A (en) | 2015-01-16 | 2015-01-16 | Android system equipment power consumption optimization method based on game load prediction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105045367A true CN105045367A (en) | 2015-11-11 |
Family
ID=54451966
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510024136.0A Pending CN105045367A (en) | 2015-01-16 | 2015-01-16 | Android system equipment power consumption optimization method based on game load prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105045367A (en) |
Cited By (35)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106547544A (en) * | 2016-10-17 | 2017-03-29 | 上海传英信息技术有限公司 | The method and device of game image buffering during a kind of low battery |
| CN106681833A (en) * | 2016-12-27 | 2017-05-17 | 宇龙计算机通信科技(深圳)有限公司 | Method and device for regulating CPU frequency |
| WO2017206916A1 (en) * | 2016-05-31 | 2017-12-07 | 广东欧珀移动通信有限公司 | Method for determining kernel running configuration in processor and related product |
| CN107577537A (en) * | 2017-09-06 | 2018-01-12 | 广东欧珀移动通信有限公司 | Resource allocation method and Related product |
| CN107577533A (en) * | 2017-08-31 | 2018-01-12 | 广东欧珀移动通信有限公司 | Resource allocation method and Related product |
| CN107621981A (en) * | 2017-09-06 | 2018-01-23 | 广东欧珀移动通信有限公司 | Resource allocation method and Related product |
| CN107783803A (en) * | 2017-11-21 | 2018-03-09 | 广东欧珀移动通信有限公司 | System optimization method, device, storage medium and the intelligent terminal of intelligent terminal |
| CN107844187A (en) * | 2016-09-21 | 2018-03-27 | 龙芯中科技术有限公司 | Power consumption management method, device and electronic equipment |
| CN108919938A (en) * | 2018-06-29 | 2018-11-30 | 中国科学技术大学苏州研究院 | A kind of power consumption of processing unit optimization method suitable for Android game |
| CN109308246A (en) * | 2017-07-27 | 2019-02-05 | 阿里巴巴集团控股有限公司 | Optimization method, device and the equipment of system parameter, readable medium |
| CN109343902A (en) * | 2018-09-26 | 2019-02-15 | Oppo广东移动通信有限公司 | Operation method, device, terminal and the storage medium of audio processing components |
| CN109413480A (en) * | 2018-09-30 | 2019-03-01 | Oppo广东移动通信有限公司 | Picture processing method, device, terminal and storage medium |
| CN109685726A (en) * | 2018-11-27 | 2019-04-26 | Oppo广东移动通信有限公司 | Scene of game processing method, device, electronic equipment and storage medium |
| CN109947569A (en) * | 2019-03-15 | 2019-06-28 | Oppo广东移动通信有限公司 | Bind method, apparatus, terminal and the storage medium of core |
| CN110286710A (en) * | 2019-07-01 | 2019-09-27 | 联想(北京)有限公司 | A kind of control method, processor and electronic equipment |
| CN110339567A (en) * | 2019-07-17 | 2019-10-18 | 三星电子（中国）研发中心 | System resource configuration, scene prediction model training method and device |
| CN111124824A (en) * | 2019-12-18 | 2020-05-08 | 北京智明星通科技股份有限公司 | Mobile phone game frame rate recommendation method and device and mobile terminal |
| CN111158454A (en) * | 2019-12-20 | 2020-05-15 | 青岛海尔科技有限公司 | Low-power consumption method and device for real-time embedded operating system |
| CN111324198A (en) * | 2020-03-18 | 2020-06-23 | 深圳市亿晟科技有限公司 | Method for dynamically adjusting frequency and voltage of android device according to load |
| CN111480336A (en) * | 2018-03-26 | 2020-07-31 | 华为技术有限公司 | Frame rate adjusting method and terminal |
| CN111666140A (en) * | 2020-05-28 | 2020-09-15 | 北京百度网讯科技有限公司 | Resource scheduling method, device, equipment and storage medium |
| CN111902790A (en) * | 2018-08-30 | 2020-11-06 | 华为技术有限公司 | Frequency modulation method and device and computer readable storage medium |
| CN113608978A (en) * | 2021-07-15 | 2021-11-05 | 荣耀终端有限公司 | Test method, test device and computer readable storage medium |
| CN110209501B (en) * | 2019-06-03 | 2022-02-08 | Oppo广东移动通信有限公司 | Frequency adjusting method and device of graphic processor, terminal and storage medium |
| WO2022089512A1 (en) * | 2020-10-29 | 2022-05-05 | 华为技术有限公司 | Load control method and apparatus, and device |
| CN114443256A (en) * | 2022-04-07 | 2022-05-06 | 荣耀终端有限公司 | Resource scheduling method and electronic equipment |
| CN114632329A (en) * | 2022-05-16 | 2022-06-17 | 荣耀终端有限公司 | Terminal equipment performance adjusting method and related device |
| CN114840859A (en) * | 2021-12-21 | 2022-08-02 | 许磊 | Streaming media system of general data |
| CN115840595A (en) * | 2022-12-12 | 2023-03-24 | 摩尔线程智能科技(北京)有限责任公司 | Configuration method and device for chip, electronic equipment and storage medium |
| WO2023048716A1 (en) * | 2021-09-23 | 2023-03-30 | Google Llc | Automated frame pacing based on gameplay state |
| CN116027878A (en) * | 2022-05-16 | 2023-04-28 | 荣耀终端有限公司 | Power consumption adjustment method and electronic equipment |
| CN116028207A (en) * | 2022-05-16 | 2023-04-28 | 荣耀终端有限公司 | Scheduling policy determination method, device, equipment and storage medium |
| CN116095221A (en) * | 2022-08-10 | 2023-05-09 | 荣耀终端有限公司 | Frame rate adjusting method in game and related device |
| WO2023151644A1 (en) * | 2022-02-10 | 2023-08-17 | Mediatek Inc. | Adaptive loading-aware system management for balancing power and performance |
| US11803221B2 (en) | 2020-03-23 | 2023-10-31 | Microsoft Technology Licensing, Llc | AI power regulation |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102830785A (en) * | 2011-06-17 | 2012-12-19 | 上海果壳电子有限公司 | Quick response system for equipment end input interaction and implementation method thereof |
| CN102902349A (en) * | 2012-11-08 | 2013-01-30 | 山东大学 | Method for reducing power consumption of embedded system by contact dynamic sampling technique |
| WO2013081602A1 (en) * | 2011-11-30 | 2013-06-06 | Intel Corporation | Reducing power for 3d workloads |
| CN103324270A (en) * | 2013-06-25 | 2013-09-25 | 东莞宇龙通信科技有限公司 | Mobile terminal and method for adjusting CPU (Central Processing Unit) frequency thereof |
-
2015
- 2015-01-16 CN CN201510024136.0A patent/CN105045367A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102830785A (en) * | 2011-06-17 | 2012-12-19 | 上海果壳电子有限公司 | Quick response system for equipment end input interaction and implementation method thereof |
| WO2013081602A1 (en) * | 2011-11-30 | 2013-06-06 | Intel Corporation | Reducing power for 3d workloads |
| CN102902349A (en) * | 2012-11-08 | 2013-01-30 | 山东大学 | Method for reducing power consumption of embedded system by contact dynamic sampling technique |
| CN103324270A (en) * | 2013-06-25 | 2013-09-25 | 东莞宇龙通信科技有限公司 | Mobile terminal and method for adjusting CPU (Central Processing Unit) frequency thereof |
Non-Patent Citations (2)
| Title |
|---|
| BENEDIKT DIETRICH等: "Lightweight graphics instrumentation for game state-specific power management in Android", 《MULTIMEDIA》 * |
| 朱宗卫: "基于系统时空行为特征的内存功耗优化研究", 《万方数据库》 * |
Cited By (56)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017206916A1 (en) * | 2016-05-31 | 2017-12-07 | 广东欧珀移动通信有限公司 | Method for determining kernel running configuration in processor and related product |
| CN107844187A (en) * | 2016-09-21 | 2018-03-27 | 龙芯中科技术有限公司 | Power consumption management method, device and electronic equipment |
| CN107844187B (en) * | 2016-09-21 | 2020-06-12 | 龙芯中科技术有限公司 | Power consumption management method and device and electronic equipment |
| CN106547544A (en) * | 2016-10-17 | 2017-03-29 | 上海传英信息技术有限公司 | The method and device of game image buffering during a kind of low battery |
| CN106681833A (en) * | 2016-12-27 | 2017-05-17 | 宇龙计算机通信科技(深圳)有限公司 | Method and device for regulating CPU frequency |
| CN106681833B (en) * | 2016-12-27 | 2021-03-09 | 宇龙计算机通信科技(深圳)有限公司 | Method and device for adjusting CPU frequency |
| CN109308246A (en) * | 2017-07-27 | 2019-02-05 | 阿里巴巴集团控股有限公司 | Optimization method, device and the equipment of system parameter, readable medium |
| CN107577533A (en) * | 2017-08-31 | 2018-01-12 | 广东欧珀移动通信有限公司 | Resource allocation method and Related product |
| CN107577533B (en) * | 2017-08-31 | 2020-12-15 | Oppo广东移动通信有限公司 | Resource allocation method and related product |
| WO2019047709A1 (en) * | 2017-09-06 | 2019-03-14 | Oppo广东移动通信有限公司 | Resource configuration method and related product |
| CN107577537A (en) * | 2017-09-06 | 2018-01-12 | 广东欧珀移动通信有限公司 | Resource allocation method and Related product |
| CN107621981A (en) * | 2017-09-06 | 2018-01-23 | 广东欧珀移动通信有限公司 | Resource allocation method and Related product |
| CN107783803A (en) * | 2017-11-21 | 2018-03-09 | 广东欧珀移动通信有限公司 | System optimization method, device, storage medium and the intelligent terminal of intelligent terminal |
| CN107783803B (en) * | 2017-11-21 | 2021-04-06 | Oppo广东移动通信有限公司 | System optimization method and device of intelligent terminal, storage medium and intelligent terminal |
| CN111480336A (en) * | 2018-03-26 | 2020-07-31 | 华为技术有限公司 | Frame rate adjusting method and terminal |
| US11404023B2 (en) | 2018-03-26 | 2022-08-02 | Huawei Device Co., Ltd. | Frame rate adjustment method and terminal |
| CN111480336B (en) * | 2018-03-26 | 2021-08-13 | 华为技术有限公司 | Frame rate adjusting method and terminal |
| CN108919938A (en) * | 2018-06-29 | 2018-11-30 | 中国科学技术大学苏州研究院 | A kind of power consumption of processing unit optimization method suitable for Android game |
| US11460905B2 (en) | 2018-08-30 | 2022-10-04 | Huawei Technologies Co., Ltd. | Frequency scaling responding to a performance change method and apparatus and computer-readable storage medium |
| CN111902790B (en) * | 2018-08-30 | 2022-05-31 | 华为技术有限公司 | Frequency modulation method, frequency modulation device and computer readable storage medium |
| CN111902790A (en) * | 2018-08-30 | 2020-11-06 | 华为技术有限公司 | Frequency modulation method and device and computer readable storage medium |
| CN109343902A (en) * | 2018-09-26 | 2019-02-15 | Oppo广东移动通信有限公司 | Operation method, device, terminal and the storage medium of audio processing components |
| CN109343902B (en) * | 2018-09-26 | 2021-08-03 | Oppo广东移动通信有限公司 | Audio processing assembly operation method and device, terminal and storage medium |
| CN109413480B (en) * | 2018-09-30 | 2021-10-08 | Oppo广东移动通信有限公司 | Picture processing method, device, terminal and storage medium |
| CN109413480A (en) * | 2018-09-30 | 2019-03-01 | Oppo广东移动通信有限公司 | Picture processing method, device, terminal and storage medium |
| CN109685726A (en) * | 2018-11-27 | 2019-04-26 | Oppo广东移动通信有限公司 | Scene of game processing method, device, electronic equipment and storage medium |
| CN109685726B (en) * | 2018-11-27 | 2021-04-13 | Oppo广东移动通信有限公司 | Game scene processing method and device, electronic equipment and storage medium |
| CN109947569A (en) * | 2019-03-15 | 2019-06-28 | Oppo广东移动通信有限公司 | Bind method, apparatus, terminal and the storage medium of core |
| CN109947569B (en) * | 2019-03-15 | 2021-04-06 | Oppo广东移动通信有限公司 | Method, device, terminal and storage medium for binding core |
| CN110209501B (en) * | 2019-06-03 | 2022-02-08 | Oppo广东移动通信有限公司 | Frequency adjusting method and device of graphic processor, terminal and storage medium |
| CN110286710B (en) * | 2019-07-01 | 2021-05-18 | 联想(北京)有限公司 | Control method, processor and electronic equipment |
| CN110286710A (en) * | 2019-07-01 | 2019-09-27 | 联想(北京)有限公司 | A kind of control method, processor and electronic equipment |
| CN110339567B (en) * | 2019-07-17 | 2023-05-26 | 三星电子（中国）研发中心 | System resource allocation and scene prediction model training method and device |
| CN110339567A (en) * | 2019-07-17 | 2019-10-18 | 三星电子（中国）研发中心 | System resource configuration, scene prediction model training method and device |
| CN111124824A (en) * | 2019-12-18 | 2020-05-08 | 北京智明星通科技股份有限公司 | Mobile phone game frame rate recommendation method and device and mobile terminal |
| CN111158454A (en) * | 2019-12-20 | 2020-05-15 | 青岛海尔科技有限公司 | Low-power consumption method and device for real-time embedded operating system |
| CN111324198B (en) * | 2020-03-18 | 2023-08-18 | 深圳市亿晟科技有限公司 | Method for dynamically adjusting frequency and voltage of android device according to load |
| CN111324198A (en) * | 2020-03-18 | 2020-06-23 | 深圳市亿晟科技有限公司 | Method for dynamically adjusting frequency and voltage of android device according to load |
| US11803221B2 (en) | 2020-03-23 | 2023-10-31 | Microsoft Technology Licensing, Llc | AI power regulation |
| CN111666140A (en) * | 2020-05-28 | 2020-09-15 | 北京百度网讯科技有限公司 | Resource scheduling method, device, equipment and storage medium |
| WO2022089512A1 (en) * | 2020-10-29 | 2022-05-05 | 华为技术有限公司 | Load control method and apparatus, and device |
| CN113608978A (en) * | 2021-07-15 | 2021-11-05 | 荣耀终端有限公司 | Test method, test device and computer readable storage medium |
| WO2023048716A1 (en) * | 2021-09-23 | 2023-03-30 | Google Llc | Automated frame pacing based on gameplay state |
| CN114840859A (en) * | 2021-12-21 | 2022-08-02 | 许磊 | Streaming media system of general data |
| WO2023151644A1 (en) * | 2022-02-10 | 2023-08-17 | Mediatek Inc. | Adaptive loading-aware system management for balancing power and performance |
| CN114443256B (en) * | 2022-04-07 | 2022-08-30 | 荣耀终端有限公司 | Resource scheduling method and electronic equipment |
| CN114443256A (en) * | 2022-04-07 | 2022-05-06 | 荣耀终端有限公司 | Resource scheduling method and electronic equipment |
| CN114632329B (en) * | 2022-05-16 | 2022-10-25 | 荣耀终端有限公司 | Terminal equipment performance adjusting method and related device |
| CN116028207A (en) * | 2022-05-16 | 2023-04-28 | 荣耀终端有限公司 | Scheduling policy determination method, device, equipment and storage medium |
| CN116027878A (en) * | 2022-05-16 | 2023-04-28 | 荣耀终端有限公司 | Power consumption adjustment method and electronic equipment |
| CN116027878B (en) * | 2022-05-16 | 2023-10-20 | 荣耀终端有限公司 | Power consumption adjustment method and electronic equipment |
| CN114632329A (en) * | 2022-05-16 | 2022-06-17 | 荣耀终端有限公司 | Terminal equipment performance adjusting method and related device |
| CN116028207B (en) * | 2022-05-16 | 2024-04-12 | 荣耀终端有限公司 | Scheduling policy determination method, device, equipment and storage medium |
| CN116095221A (en) * | 2022-08-10 | 2023-05-09 | 荣耀终端有限公司 | Frame rate adjusting method in game and related device |
| CN116095221B (en) * | 2022-08-10 | 2023-11-21 | 荣耀终端有限公司 | Frame rate adjusting method in game and related device |
| CN115840595A (en) * | 2022-12-12 | 2023-03-24 | 摩尔线程智能科技(北京)有限责任公司 | Configuration method and device for chip, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105045367A (en) | Android system equipment power consumption optimization method based on game load prediction | |
| KR102490908B1 (en) | Resource scheduling method and terminal device | |
| EP3333668B1 (en) | Virtual machine power consumption measurement and management | |
| US20120192200A1 (en) | Load Balancing in Heterogeneous Computing Environments | |
| US20190146842A1 (en) | Method and Apparatus for Allocating Computing Resources of Processor | |
| US20210208935A1 (en) | Method for Scheduling Multi-Core Processor, Terminal, and Storage Medium | |
| Qian et al. | Jade: Reducing energy consumption of android app | |
| AU2011254039A1 (en) | Dynamically adjusting the complexity of worker tasks in a multi-threaded application | |
| CN107844187A (en) | Power consumption management method, device and electronic equipment | |
| CN111708642A (en) | Processor performance optimization method and device in VR system and VR equipment | |
| CN110795238B (en) | Load calculation method and device, storage medium and electronic equipment | |
| CN102866765B (en) | Graphics processing unit and management method thereof | |
| Peters et al. | Frame-based and thread-based power management for mobile games on HMP platforms | |
| CN111506434A (en) | Task processing method and device and computer readable storage medium | |
| Han et al. | Performance improvement of Linux CPU scheduler using policy gradient reinforcement learning for Android smartphones | |
| Li et al. | An adaptive cpu-gpu governing framework for mobile games on big. little architectures | |
| Lu et al. | Mildip: An energy efficient code offloading framework in mobile cloudlets | |
| Zhang | DVFS-based energy-aware scheduling of imprecise mixed-criticality real-time tasks | |
| Padoin et al. | Managing power demand and load imbalance to save energy on systems with heterogeneous CPU speeds | |
| Kim et al. | Energy-efficient real-time multi-core assignment scheme for asymmetric multi-core mobile devices | |
| Park et al. | Synergistic CPU-GPU frequency capping for energy-efficient mobile games | |
| Cheng et al. | Behavior-aware integrated CPU-GPU power management for mobile games | |
| CN115617494B (en) | Process scheduling method and device in multi-CPU environment, electronic equipment and medium | |
| US20130031391A1 (en) | Multi-core processor system, electrical power control method, and computer product | |
| CN113918002A (en) | Frequency modulation method, frequency modulation device, storage medium and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20151111 |
|
| RJ01 | Rejection of invention patent application after publication |