CN107610039A - Image processing method and image processing apparatus - Google Patents
Image processing method and image processing apparatus Download PDFInfo
- Publication number
- CN107610039A CN107610039A CN201710555428.6A CN201710555428A CN107610039A CN 107610039 A CN107610039 A CN 107610039A CN 201710555428 A CN201710555428 A CN 201710555428A CN 107610039 A CN107610039 A CN 107610039A
- Authority
- CN
- China
- Prior art keywords
- frame
- image processing
- group
- event
- performance
- 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.)
- Withdrawn
Links
Landscapes
- Controls And Circuits For Display Device (AREA)
- Power Sources (AREA)
Abstract
The present invention provides a kind of image processing method and image processing apparatus.The image processing method includes：The first frame is handled by providing control setting to first group of equipment, to reach first performance measurement；After first frame, the scene information on the second frame is received from second group of equipment；Change between first frame and second frame is quantified；According to the comparison accommodation control setting between the change of the quantization and predetermined threshold value；And the control by providing the adjustment is set to first group of equipment to handle second frame.The image processing method and image processing apparatus of the present invention can optimize the performance level of image processing system.
Description
Technical field
The present invention relates generally to the video playback in electronic installation, more specifically, being the power supply on graphics processing unit
Management.
Background technology
Unless there are separately illustrating, the scheme of this part description is not the preceding case of foregoing Claims not because these are said yet
Bright be included in this part and to recognize it be preceding case.
Graphics processing unit (GPU) (also referred to as VPU, i.e. visual processing unit) is a special electricity
Sub-circuit, it is designed to fast operating and changes storage so as to accelerate to export the image wound in the frame buffering to display
Build.GPU is used in embedded system, cell phone, PC, work station and game machine.Modern GPU is for operating computer
All very effectively, the structure of its highly-parallel causes them than general CPU in the big block number of parallel processing for image and image procossing
According to algorithm in terms of more efficiently.
The content of the invention
Therefore, the power problemses of the invention to solve in image procossing, spy are provided at a kind of image processing method and image
Manage device.
The present invention provides a kind of image processing method and included：First is handled by providing control setting to first group of equipment
Frame, to reach first performance measurement；After first frame, the scene information on the second frame is received from second group of equipment；To this
Change between first frame and second frame is quantified；Adapted to according to the comparison between the change of the quantization and predetermined threshold value
Property adjust the control setting；And the control by providing the adjustment is set to first group of equipment to handle second frame.
The present invention separately provides a kind of image processing method, comprising：Frame is handled by providing control setting to one group of equipment,
To reach first performance measurement；When graphics processing unit is handling the frame, particular event is detected at the graphics processing unit；
The performance metric of event discrimination second based on the detection；And adjust the control of this group of equipment and set to reach second performance
Measurement.
The present invention separately provides a kind of image processing apparatus, comprising：One group of processing unit；Graphics processing unit；Display device；
And computer-readable storage medium, store multigroup instruction, wherein this it is multigroup instruction by the execution of this group of processing unit be by with
This group of processing unit is put to perform following action：When the frame that graphics processing unit processing will be shown on the display apparatus,
Control setting is provided to the graphics processing unit and the display device to reach first performance measurement；When the graphics processing unit exists
When handling the frame, a particular event is detected at the graphics processing unit；The performance of event discrimination second based on the detection
Measurement；And the control setting of the graphics processing unit is adjusted based on second performance metric.
The image processing method of the present invention can optimize the performance level of image processing system with image processing apparatus.
These and the other purposes of the present invention for those skilled in the art, is reading following be preferable to carry out
It is readily appreciated that and understands after the detailed description of example, the preferred embodiment is disclosed by several figures.
Brief description of the drawings
Fig. 1 display images processing system 100.
Fig. 2 shows that power supervisor 110 adjusts the schematic diagram of the setting of the performance of image processing system 100.
The conceptual flows 300 for being shown in management power supply in image processing system of Fig. 3.
Fig. 4 is the structural representation of the image processing system 100 comprising CPU, GPU, memory and display device.
Fig. 5 a-5b are shown with performance look-up table 150 to search the image processing system 100 of the performance setting thing of distinct device
In data flow.
Fig. 6 shows the timestamp of the event based on monitoring come to the fine setting of the performance setting of the frame in processing.
Fig. 7 is shown in the data flow that image processing system 100 performs fine setting performance setting based on Event Timestamp.
Fig. 8 conceptually shows the schematic diagram for the electronic system 800 that some embodiments of the present application are implemented.
Embodiment
Present specification and claims have used some specific components of word acute pyogenic infection of finger tip.Those skilled in the art can manage
Solution, manufacturer may use the different same components of title acute pyogenic infection of finger tip.This document passes through function not by the difference of name
Difference distinguish component.In the following description book and claims, word " comprising " is open, therefore it should be managed
Solve as " including, but are not limited to ... ".
Some embodiments disclose the method and device of the power supply for managing image processing system.Specifically, this method
The performance of image processing system is adaptively adjusted based on the scene information in every frame.In certain embodiments, such adjustment
Purpose is to make the performance level of image processing system to optimize.For some embodiments, the optimum performance of a frame is set
Surely it is the live load that image processing system just completes frame before the end of frame, the waste of such power supply is minimum.(when by
When complete live load performance level less than it is necessary horizontal when can abandon frame.Circuit is higher than necessary performance water operating in
Power supply can be wasted by putting down when carrying out completing task on time load.)
In certain embodiments, image processing system includes graphics processing unit (GPU) and a field for being couple to GPU
Scape perceives (scene-aware) power supervisor.Scene perception power supervisor is used to receive scene information and according to reception
Scene information carrys out Adaptive Control GPU.
Fig. 1 display images processing system 100, it includes scene perception power supervisor 110, for managing image procossing system
The performance of system.Scene information and phase are obtained in the data that each equipment is collected from image processing system 100 of power supervisor 110
The performance setting of image processing system should be adjusted.
Image processing system 100 is one comprising processing data and can producing the electricity of the component of the data of image display
Sub- equipment.Such a equipment can be in general computing device, e.g. desktop computer, notebook computer, tablet personal computer,
Smart mobile phone, etc., it includes CPU (CPU), storage assembly, input-output apparatus, network interface, Yong Hujie
Mouthful etc..Such a equipment is also provided with the hardware for having such as GPU to be used in particular for handling view data like that.As image
The electronic equipment of processing system 100 also can perform image-independent function.
Image processing system 100 includes power supervisor 110, frame analyzer 120, performance controller 130, event reporter
140 and performance look-up table (LUT) 150.Power supervisor 110 receives data as scene information and base from frame analyzer 120
The performance setting of performance controller 130 is determined in the scene information of reception.Power supervisor 110 detects from the scene information of reception
The level (level of scene change) of scene changes, for example, by by its scene information ratio that previous frame is received therewith
Compared with, and performance (estimated needed for an estimation is searched in performance look-up table 150 using the level of the scene changes of detection
required performance).Power supervisor 110 produces the performance of performance controller 130 using performance needed for estimation
Setting.
In certain embodiments, these performance settings include frequency setting 131 and voltage setting 132.Voltage setting 132 refers to
Show image processing system 100 in the voltage needed for the operation at frequencies indicated by frequency setting 131.(higher voltage allows circuit
Operate in higher frequency, meeting is because of higher data throughput and/or more low latency causes higher performance to be measured for this, but also brings more
Big power consumption.)
The event accommodation performance controller 130 that image processing system 100 is reported based on event reporter 140
Performance setting.In certain embodiments, power supervisor 110 receives the identity of the event of report from event reporter 140
(identity) and the timestamp related to event (timestamp), and determine to be supplied to the performance of performance controller 130 to set
Whether enough (less fast or too slow) to determine.In certain embodiments, event reporter 140 is configurable, such as can be monitored
And/or the type of the event of report can adapt to the type of application run on CPU or GPU by user configuration.
Frame analyzer 120 is from the modules in image processing system 100 or circuit collection status, data or report and carries
For collection data to power supervisor.The module or circuit for being collected state can include following one or more assemblies：CPU、
GPU, storage device, display device, bus structures, and other kinds of circuit, they form image processing system together
100.The state of collection can include the signal directly sent by each circuit of image processing system 100 and/or be stored in
The data that can be read in storage organization by power supervisor 110.The data or status bits collected from these equipment are supplied to electricity
Source manager 110 determines performance setting (in performance controller 130 as the indicator (indicators) for obtaining scene information
In).
Power supervisor 110 uses some status datas to be used as " early stage indicator ", because they indicate frame at hand
Upcoming image processing load.The early stage indicator of frame be before the image procossing of frame starts or just start when predict
The available status data of the image processing load of frame is (such as before frame starts or incipient event, such as vertical synchronization
(vertical synchronization) or VSYNC).Such early stage indicator can include the status data related to CPU
And/or the memory access of frame at hand.
Performance controller 130 controls the one group of control data or signal to each module in image processing system 100 or circuit.
These modules or circuit can include one or more components below：CPU, GPU, storage device, display device, bus structures, with
And other kinds of circuit, they form image processing system 100 together.It is direct that control data can include power supervisor 110
Issue the signal of 100 each circuit of image processing system and/or power supervisor 110 is stored in control number in storage organization
According to.The control data or signal of the performance of circuit and/or equipment in the processing control image processing system of performance controller 130.
The control control clock frequency of performance controller 130 (for example, frequency setting 131) and operating voltage (for example, voltage setting 132)
Setting.Performance controller 130 also controls control display frame rate, the setting of the display response time of user interaction, or other can
Influence the performance of image processing system 100 or the setting of electricity usage.In certain embodiments, one group of performance setting can reach one
Individual specific performance metric (such as particular job frequency or particular data rate).
Event reporter 140 reports the operating procedure of some certain types of times or image processing system 140 to electricity
Source manager 110.Power supervisor 110 determines to be supplied to the performance setting of performance controller 130 be using the event of report
It is no enough.In order to report an event, event reporter 140 provide in certain embodiments the identity (identity) of event with
Mark the timestamp of time time of origin.The level of the scene changes of the corresponding revaluation event of power supervisor 110 simultaneously determines whether
To be set based on the scene changes level of revaluation with fine-tuning mode tuning performance.
Performance look-up table 150 is a look-up table, for mapping the scene information from frame analyzer 120 to Properties Control
The performance setting of device 130.Performance look-up table 150 can include direct mapping scenarios information to the entry of performance setting.Performance is searched
Table 150 also can containment mapping obtain parameter (derived parameters) arrive performance setting entry.For example, power supply
Manager 110 calculates the level of scene changes as a parameter to search performance is set from scene information.(level of scene changes is
The scene information of previous frame or the scene information or scene of scene and present frame between change measure.)
Power supervisor 110 is that the information provided based on frame analyzer 120 determines performance to performance controller 130
The horizontal module of setting.In certain embodiments, power supervisor 110 is one and run in image processing system 100
A software module in one group of processing unit.This group of processing unit operated in power supervisor can be CPU, GPU or
Other form the processing unit of image processing system.
In certain embodiments, scene information includes one group of early stage indicator.Power supervisor 110 compares the morning of present frame
The early stage indicator of phase indicator previous frame therewith determines the level of scene changes and according to the level point of the scene changes of determination
With one group of initial (initial) performance setting.Power supervisor is set in before frame starts using this group of initial performance of distribution and set
Determine the performance of image processing system 100.In certain embodiments, if sufficiently small (i.e. present frame is therewith for scene changes horizontal
Scene changes amount between previous frame is less than a threshold value), power supervisor 110 can from the performance setting of previous frame increase or decrease
The initial performance of one a small amount of (or retention property setting is constant) as new present frame.
In addition, if sufficiently large (i.e. the amount of present frame scene changes between previous frame therewith is more than for scene changes horizontal
One threshold value), power supervisor can provide the performance setting that can be obviously improved the performance of image processing system 100.Because
The level of scene changes is bigger, and the uncertainty for being actually needed the treating capacity for handling the frame is just bigger.Need to complete by over-evaluating
The performance of processing, risk of failure of the power supervisor in processing when the current frame can minimize.
Fig. 2 shows that power supervisor 110 adjusts the schematic diagram of the setting of the performance of image processing system 100.This figure is shown
It is being the setting of performance level in the processing of four successive frames (frame 1 to 4).As illustrated, in the beginning of each frame, power supply pipe
Reason device sets the performance of (dash area) offer image processing system according to initial performance.When image processing system is with the place of frame
When managing and running, power supervisor provides the fine setting (non-shaded portion) of performance.
As it was previously stated, the setting of initial performance setting is the ratio based on the early stage indicator between present frame therewith previous frame
Compared with.Fig. 2 conceptually shows the early stage indicator (the early stage indicator 211-214 of frame 1 to 4 respectively) of each frame.In this example
In son, the initial performance of frame 2 sets the comparison between the early stage indicator of the early stage indicator and frame 1 based on frame 2, frame 3 just
Comparison between the early stage indicator of early stage indicator and frame 2 of the beginning performance setting based on frame 3, the initial performance setting base of frame 4
Comparison between the early stage indicator of frame 4 and the early stage indicator of frame 3, etc..
As illustrated, power supervisor uses the performance setting of frame 1 to be set as the initial performance of frame 2.Because frame 1
Early stage indicator 211 and frame 2 early stage indicator 212 it is closely similar (cpu load for all including task A, B and C).Power supply pipe
Reason device also uses performance setting of the performance setting of frame 3 as frame 4, only a little increases.Because the early stage instruction of frame 3
The early stage indicator 214 of device 213 and frame 4 is closely similar (all to include task Y and Z cpu load, although early stage indicator 214
Include task X2 rather than X1 cpu load).For frame 2 and 4, the scene changes calculated it is horizontal sufficiently small so that power supply pipe
The performance setting for the previous frame that reason device can be continuing with is set as the initial performance of present frame.
In addition, the performance of power supervisor distribution lifting is set as the initial performance of frame 3.Because the early stage of frame 3
The early stage indicator 212 of indicator 213 and frame 2 it is dramatically different (the early stage indicator of frame 2 includes task A, B and C cpu load,
And the early stage indicator of frame 3 includes task X1, Y and Z cpu load).For frame 3, the level of the scene changes calculated is too
Greatly so that the performance setting for the previous frame that power supervisor can not be continuing with is set as the initial performance of present frame.In fact,
Power supervisor not can determine that what is suitable initial performance setting, so its lifting initial performance is set to a level and made
Possibly for being enough for uncertainty.
In certain embodiments, power supervisor quantifies to scene changes.When the change of quantization is specific more than one
During threshold value, power supervisor lifted the frame initial performance set reach than previous frame the high a certain amount of lifting of performance metric
Performance metric.When the change of quantization is less than specific threshold, power supervisor by from previous frame performance setting adjustment one
Specified quantitative sets to set the initial performance of present frame.
In certain embodiments, the initial performance of lifting is a class value, and this class value is divided by the level based on scene changes
With (difference i.e. between early stage indicator).In certain embodiments, the initial performance of lifting is one group of predetermined value, and this group is predetermined
Value be totally independent of scene changes level and previous frame performance setting.In certain embodiments, the initial performance of lifting
Than previous frame the high specified quantitative of performance metric.
As illustrated, after the initial performance setting of each frame, power supervisor is further with fine-tuning mode scalability
Can setting.In certain embodiments, power supervisor is commented or the level of revaluation scene changes again in some event or operating procedure
(for example, reported by event reporter 140).When these steps or event occur, power supervisor compare scene information with
The scene information (or early stage indicator) of record carrys out the level of revaluation scene changes before one group.The scene letter recorded before this
Cease the scene information for the previous frame that can be, or event or operating procedure before.Power supervisor then uses the field of revaluation
The level of scape change determines new performance setting.In certain embodiments, the level of the scene changes of revaluation is by as index
Use, to search one group of performance setting from performance look-up table 150.
In certain embodiments, the event or operating procedure each reported are related to timestamp, the timestamp label event
Actually occur the time, power supervisor relatively actually occurs the time with it is expected time of origin to determine that the fine setting of performance setting is adjusted
It is whole.
The conceptual flows 300 for being shown in management power supply in image processing system of Fig. 3.In certain embodiments, power supply pipe
Manage device 110 and perform flow 300 when controlling the performance setting of image processing system.In certain embodiments, power supervisor is
One software module run by CPU or GPU, CPU or GPU perform flow 300.
Flow 300 starts (step 310) to extract the early stage indicator of upcoming frame/scene, and the frame includes one
Scene, the scene may or may not have previous frame significant change therewith.These early stage indicators are out of image processing system
The part for the scene information that each component receives.Flow also receives the instruction that (step 320) frame starts event, such as at hand
Frame VSYNC signals.When receiving VSYNC signals, upcoming frame turns into " present frame ".
After receiving frame starts event instruction, flow is by comparing the early stage indicator previous frame therewith of the present frame of extraction
Early stage indicator, to calculate (step 330) scene changes level.In certain embodiments, flow is by scene changes amount
Turn to a value or set of values.
The level of scene changes based on calculating, flow determine whether the level of scene changes is significantly to increase to (step 340)
Add still increases slightly, such as whether the quantization level of scene changes is more than a specific threshold.In certain embodiments, significantly
Scene changes can use summit (vertex) quantity, difference, which is drawn, calls (different draw call), renders varying number
Layer (different number of layers rendered), and additional events or program be activated.If scene becomes
Change significantly, flow proceeds to 345.If the quantization level of scene changes is not notable, such as the quantization level of scene changes is less than
Specific threshold, flow then proceed to 350.
In step 345, flow is based on one group predetermined, higher (being lifted) performance setting allocation performance setting.Power supply
Manager uses the performance setting of this lifting, because higher levels of scene changes meaning is more difficult to be based on existing performance setting
To predict optimum performance setting (the actual performance setting needed is much higher than existing performance setting).Power supervisor is therefore
Improving performance is set to higher than uncertain threshold value.In certain embodiments, the performance setting of lifting is predetermined value, these predetermined values
Set independently of current performance.In certain embodiments, power supervisor increases a predetermined lifting values and set to current performance
To reach the performance setting of lifting.After the performance setting of distribution lifting, flow proceeds to 360.
In step 350, flow reuses (reuse) existing performance setting or fine setting performance setting.In this operation, flow
Determining the frame, scene changes are very small between previous frame therewith, it is therefore more likely that existing performance setting (previous frame used in setting or
The setting used before present frame) it is still optimal setting.Therefore power supervisor reuses current performance setting, or with than micro-
Increase/reduce performance setting with adjusting less amount (fine setting threshold value is less than the uncertain threshold value of operation 345).Flow proceeds to
360。
In certain embodiments, the size of the fine setting of performance setting is the level based on scene changes.In some embodiments
In, fine setting is based on the whether enough inspections of performance.Flow receives the event reported or operating procedure (from event reporter
140) timestamp occurred with the event.Flow then compares expeced time of timestamp and event to determine that performance setting is too
It is high still too low, and therefore decide whether to increase or reduce performance setting.Using the timestamp of report event come fine setting property
Can set can further illustrate below with reference to Fig. 6-7.
In step 360, flow should determine that whether to set or tuning performance sets in timing determination frame and is in power supervisor
It is no to have another report event.First event of frame is that power supervisor is set or tuning performance setting is VSYNC events, and it is marked
Will the beginning of frame.But power supervisor can monitor other events or operating procedure in frame and when these events occur
Execution is commented and finely tuned again.One example of such event is when GPU completes to paint N number of triangle or pixel.If in frame
There is another so monitored event, flow proceeds to 370.No, flow just terminates.
In step 370, flow is when the event of monitoring occurs, by the scene information for comparing scene information and version before
Or by comparing the scene information of scene information previous frame therewith, carry out the level of revaluation scene changes.(this operation is similar to operation
330, it compares the early stage indicator of the early stage indicator of frame at hand previous frame therewith).Flow proceeds to 340.
As it was previously stated, in certain embodiments, image processing system includes CPU, GPU, storage stack, and display is set
It is standby.Power supervisor as scene information and controls the performance of these equipment to set by using the data as caused by these equipment
It is fixed to perform the power management of scene perception.
Fig. 4 is the structural representation of the image processing system 100 comprising CPU, GPU, memory and display device.At image
Reason system performs the power supply of scene perception by using the data of CPU, GPU, memory and display device as scene information
Management.
As illustrated, the circuit of image processing system 100 includes CPU410, GPU420, main storage 430, GPU memory
440, with display device 450.Storage stack controller 435 controls main storage 430 and GPU memory 440.Display controller
455 control display devices 450.Scene perception power supervisor 110 is shown as the software run on GPU420 or hard
Part module.But it may also be the software module run by CPU 410.These components are interconnected with different circuit elements, this
A little circuit elements are referred to as bus or bus structures (not shown).
In certain embodiments, CPU410, GPU420, main storage 430, GPU memory 440, Memory Controller 435
And display controller 455 is implemented with the hardware circuit module of the integrated circuit of one or more electronic installations.For example,
In some embodiments, main storage 430 is implemented with GPU memory 440 with physical memory devices, and Memory Controller 435,
Display controller 455, CPU 410, GPU420 and power supervisor 110 are implemented in an IC.
As illustrated, CPU410 direct communications mutual with GPU or being communicated by memory 430 with 440.The group stores
The control memory of device controller 435 equally also performs the direct storage of memory construction to the access of main storage 430 and 440
Device access (DMA) operation.The data transfer that these dma operations can be included between main storage 430 and GPU memory 440,
Data transfer between GPU420 and GPU memory 440, the data transfer between CPU and main storage 430, and in master
Data transfer between memory 430 and display 450 (there is display buffer to be used to store the pixel data to be shown).
These equipment of image processing system 100, which perform to calculate to draw with computing, provides the data to be shown to display device
450.For example, image processing system receives data from big storage device or network by I/O equipment 405, and data are stored to master
Memory 430.Based on the data of these storages, CPU410 performs various calculating tasks and/or load, produces the data of processing
The image that display device 450 to be given is shown is processed into GPU 420.The use of power supervisor 110 on by CPU410 with
The load and the information of task and the information for the data for being produced by CPU 410 and GPU420 and/or being handled that GPU420 is performed
As scene information.
Although not showing, in certain embodiments, image processing system 100 is the part of a camera arrangement, and
The view data as caused by image processing system 100 is supplied to an image or video encoder.
Scene information is used for the optimum level of the power settings of prognostic chart picture processing system, because they indicate to need to perform
To produce the size of the live load of the necessary data of display or camera record.The scene information never collected with flow is put
It is analyzed together and jointly the level that (such as summation) goes out scene changes.Some power supervisors are followed by from image procossing system
The example for the scene information that the distinct device of system is collected：
Application program/game engine/game physics calculate (Game physics calculation of previous
scene loading)；
GPU context quantity (context number)；
Summit/pel quantity (Vertex/primitive number)；
Drawing command quantity (Draw command number)；
Vertex coloring run time and complexity (Vertex shading run-time and complexity)；
Inlay run time and complexity (Tessellation run-time and complexity)；
Summit is distributed and covering number of blocks (Vertex distribution and covered tile numbers)；
Post-processing object layer number (Rendering target layer number)；
Every layer renders resolution and number of blocks (Rendering resolution and tile number of each
layer)；
Pixel shader run time and complexity (Pixel Shading run-time and complexity)；
Texture size/type/layer/complexity (Texture size/type/layer/complexity)；
General GPU event counters (General GPU event counter), i.e. block, pel, summit, pixel, line
Reason, instruction ... etc.；
Application programming interfaces (API, application programming interface) type；
Chip temperature in image processing system；
Cpu load (before next scene, pretreatment, API Calls)；
Bandwidth/DRAM delays/cache hit rate；
VSYNC events (vertical synchronization of segmentation video field)；And
External user event.
As it was previously stated, power supervisor determines that each frame is opened using the scene information that some are collected as early stage indicator
The one group of initial performance setting begun.It is followed by the example for being used as the scene information of early stage indicator：
The cpu load calculated using/game engine/game physics；
GPU renders/the API tracks (API trace) of calculating standard (OpenGL, OpenCL, Vulkan, etc.), comprising
The attribute of each API funcalls, state and parameter；
Vertex coloring run time and complexity；
Inlay run time and complexity；
The quantity of the block of block list-covering；
The quantity of post-processing object layer；
Every layer of resolution and total number of blocks；
API types；
Pixel shader run time and complexity；
Texture type, size, layer, run time, and complexity；
User interface event；And
Display quantity.
In certain embodiments, power supervisor provides each component of the performance setting to image processing system, to reach
Specific performance is measured, such as particular job frequency (in order to reach a particular data rate or delay).One group of performance setting can wrap
Performance setting containing different components, module or circuit in image processing system.In other words, one group of performance setting can include CPU
410 frequency is set with voltage, and GPU 420 frequency and voltage are set, frequency and voltage setting of bus structures, etc..
In some embodiments, the particular module of image processing system 100 or the performance setting of circuit include the setting of other influences performance.
For example, the performance setting of display 450 (or display controller 455) can include control frame per second, the display response of user interaction
The setting of time (also crying " display is final (display deadline) "), because they also have influence on the work(of image processing system
Consumption；CPU410 performance setting can also include the particular content that use several cores.
The example of following performance setting is controlled by scene perception power supervisor：(as initial performance setting or
Fine setting)
Switch GPU or the power supply of its subprocess (sub-instances)；
Deceleration/acceleration GPU/CPU and its subprocess frequency and voltage；
(CPU, GPU etc.) early stage of equipment wakes up (Early wake-up) or early stage accelerates (early speed-up)；
The adjustment (such as main storage 405 and/or GPU memory 440) of bandwidth of memory and resolving strategy；And
Display frame rate and final strategy (deadline strategy).
In certain embodiments, the fine setting of performance setting includes budget and corrected with stepping.Such budget is corrected with stepping
It may be used on some or all in being set below image processing system：
Switch extrinsic stain device/submodule/SRAM power supplys PMIC/LDO/MTCMOS；
Deceleration/acceleration work tinter/submodule/SRAM frequencies and average voltage (even voltage)；
Waken up by the early stage of prediction or accelerate early stage to reduce performance reduction；
The cpu load distribution of flow；
DRAM bandwidth distributes；And
Display strategy and final strategy (Display strategy and deadline policy).
As it was previously stated, power supervisor monitors (i.e. after frame starts) other events or operating procedure in frame and at that
A little events perform when occurring and comment and finely tune again.It is subsequently power supervisor in order to which tuning performance sets monitored event or operation
The example (such as by event reporter 140) of step：
The event occurred at CPU:
The cpu load of GPU applications；
In the event that the vertex coloring stage (in GPU) occurs:
Pel process performance；
The pel stage quantity (Number of Primitive Phase kicked) of execution；
Colour location counter；
In the event that the pixel shader stage (in GPU) occurs:
The quantity for the layer to be rendered；
The block process performance of each layer；
Colour location counter；And
Multiple sampling antialiasing (MSAA) type.
As it was previously stated, in certain embodiments, power supervisor is based on scene information (comprising early using a look-up table
Phase indicator) carry out search performance setting.Fig. 5 a-b are shown with performance look-up table 150 to search the performance setting thing of distinct device
Data flow in image processing system 100.
As shown in Figure 5 a, power supervisor 110 from CPU410, Memory Controller 435, display controller 455,
GPU420, and the other equipment 490 comprising bus structures component receives scene information and/or indicator (is included and referred in early days
Show device).Power supervisor 110 using the scene information of reception come from look-up table 150 search performance set.It is initial in order to produce
Performance setting, the early stage indicator that power supervisor compares the early stage indicator of frame at hand previous frame therewith (are shown in
Stored in storage 510) and be quantified as difference " levels of scene changes ".
As shown in Figure 5 b, power supervisor uses the horizontal index as lookup look-up table 150 of the scene changes, and takes
One group of performance setting is returned, includes frequency, voltage and frame per second.In an example shown, the quantization level of scene changes is " 3 ", and
And look-up table one group of performance setting of corresponding generation, it is 400MHz, voltage 2.4V comprising frequency, and frame per second is per second 27
Frame.As it was previously stated, one group of performance setting can include many parameters, such as the different class frequencys of multiple different circuits or module and electricity
Pressure, and such as display response are final (display response deadline), memory access resolving strategy etc.
Parameter.
As it was previously stated, power supervisor is based not only on the initial performance setting of each frame of early stage indicator offer of frame,
The fine setting set after the processing of the frame starts to the frame execution performance in processing.In certain embodiments, these adjustment are
Implemented in the particular event during GPU handles frame.Power supervisor is to evaluate performance setting using these events
It is no enough and adjust accordingly.In certain embodiments, image processing system includes the event reporter of such as event reporter 140
To report these events, for example, by the identity and the timestamp of event generation of reporting each event.The phase of power supervisor 110
The expected time of the event should be distinguished with timestamp using the event of report, to determine that performance setting is too high or too low.For example,
Power supervisor 110 in some embodiments monitor GPU finish watching into the calculating of the 10,000 of a frame triangles will be how long.
Power supervisor 110 is using the timestamp related to the event come the timestamp based on the event and when being expected to run of the event
Between comparison determine how soon GPU can end task and whether to increase or decrease performance.
Fig. 6 shows the timestamp of the event based on monitoring come to the fine setting of the performance setting of the frame in processing.The figure shows two
The adjustment of performance setting during individual successive frame 601 and 602, particularly when the two frames are to be handled with GPU 420 to show or shine
Cameras record.In this embodiment, the period that frame 601 and frame 602 are shown on screen is all 16.6ms.
As illustrated, when handling frame 601, GPU operation is in frequency 525MHz.This frequency be from frame inheritance before and
Come, because previous frame is same or like therewith for its early stage indicator.This frequency may also be the performance setting of one group of lifting, because field
The level of scape change is considered as too greatly.
Power supervisor monitors multiple GPU events, comprising event " X " and " Y " (this can be corresponding, such as GPU completions 10,
000 triangle renders).(after frame starts) and GPU events can occur at 4.1ms marks for GPU events " X " expection
" Y " at 7.0ms marks it is contemplated that occur.GPU the frequencies 525MHz's and GPU being all based on these expeced times in frame 601
What load was predicted.The GPU events X time that actually occurs is 4.1ms, and the GPU events Y time that actually occurs is 7.0ms, this
Individual and their expeced time is identical (or very close).Power supervisor is it is thus determined that load is just optimal fast to approach
Rate is processed, and retention property is set in 525MHz.Task/load 1-A, 1-B, with 1-C almost terminates at the end of frame, confirms
The GPU of frame 601 performance setting is almost optimal.
For GPU with the processing of frequency 525MHz start frames 602, this frequency is to be inherited from the processing of frame 601, because its early stage
Indicator is same or similar with frame 601.GPU can handle multiple loads (task 2-A, 2-B, and 2-C).Based on these loads
And frequency 525MHz, performance manager determine that event X expeced time is 4.1ms and event Y expeced time is 7.0ms.
With GPU handle frame 601, event X actual event result is 2.5ms, it is meant that GPU run must than need must
(" 525MHz amendment estimation " display load 2A-2C completes too early) hurry up so speed can be reduced to reduce power consumption.Power supply
Therefore manager reduces GPU frequencies to 400MHz.GPU is then carried out with 400MHz frequency processings until it encounters thing in 9.8ms
Part Y, it was contemplated to be originally earlier to be reached in 7.0ms.In other words, GPU is handled too slow and can not be used 400MHz frequency
Rate completes load in time (" 400MHz amendment estimation " prediction load 2B and 2C is not timely completed).Therefore power supervisor carries
The performance setting for rising GPU is 700MHz to be timely completed task.
Fig. 7 is shown in the data flow that image processing system 100 performs fine setting performance setting based on Event Timestamp.As schemed
Show, what event reporter 140 was detected by sending the timestamp 702 of an event identity (ID) 701 and event to report this
Event (such as GPU completes 10,000 triangles) gives power supervisor 110.(in certain embodiments, power supervisor is receiving
The timestamp is provided during to an event reported).The timestamp of event allows power supervisor to distinguish the reality of the event
Time.Then power supervisor 110 (is shown as from lookup using the event id 701 of reception to search the expeced time of the event
The expeced time 711 that table 710 receives).Power supervisor 110 compares expeced time 711 and real time based on timestamp 702
702, to determine the event whether in tolerance interval within the scheduled time.If it is not, power supervisor sends the performance of adjustment
Each circuit to image processing system 100 is set, includes CPU 410, GPU 420, Memory Controller 435, display control
Device 455 and other equipment 490.In certain embodiments, when the amount of fine setting is by actual based on event id 701 and the event
Between difference between 702 and expeced time 711 search and obtain from performance look-up table 150.
In certain embodiments, the content of each look-up table (includes performance look-up table 150 and look-up table LUT expeced time
710) it is all based on scene information and dynamic is adjustable.It is such as micro- when the various combinations for performing scene information or early stage indicator
Timing, power supervisor 110 can be with better performance settings come the content of more new capability look-up table 150.Power supervisor also exists
After being informed in the particular characteristic setting lower real time that it needs to reach particular event, renewal look-up table expeced time 710 it is interior
Hold.
Electronic system example
Many foregoing features and application are implemented with the mode of software flow, and these softwares are specifically real with one group of instruction
Existing, these instructions are recorded in computer-readable storing media (also referred to as computer readable medium).When these instructions are by one or more
Individual calculating or processing unit (for example, one or more processors, processor core or other processing units), they to handle
Indicated action in unit execute instruction.The example of computer readable medium include, but are not limited to, and CD-ROM, flash memory, deposit at random
Access to memory (RAM) chip, hard disk, the read-only storage (EPROM) of erasable programmable, the read-only of electrically erasable are deposited
Reservoir (EEPROM) etc..Carrier wave and electric signal is not wirelessly or non-wirelessly included in computer readable medium.
In this manual, term " software " means comprising the firmware in read-only storage or is stored in magnetic storage
It is middle can be read into memory in allow processor handle application.Moreover, in certain embodiments, multiple software inventions can be implemented
The software inventions of difference are remained simultaneously for the subdivision of a bigger program.In certain embodiments, multiple software inventions
Different programs can be embodied as.Finally, any combinations for being embodied as software inventions together of distinct program described herein all exist
In scope of the present application.In certain embodiments, when loading software program in one or more electronic systems come when operating,
Definition is one or more to be run and performs the specific its embodiments of the operation of software program.
Fig. 8 conceptually shows the schematic diagram for the electronic system 800 that some embodiments of the present application are implemented.Electronic system
800 can be a computer (such as desktop computer, PC, tablet personal computer, etc.), phone, PDA, or other kinds of
Electronic equipment.Such electronic system includes various computer readable mediums with being connected connecing for other kinds of computer readable medium
Mouthful.Electronic system 800 includes bus 805, processing unit 810, graphics processing unit (GPU) 815, system storage 820, network
825th, read-only storage 830, permanent storage appliance 835, input equipment 840, and output equipment 845.
Bus 805 briefly represents system, periphery and the core of all communication connection 800 each internal units of electronic system
Piece bus.For example, bus 805 communicatedly connects processing unit 810 and GPU 815, read-only storage 830, system storage 820
And permanent storage appliance 835.
From these each memory cells, processing unit 810 takes out the instruction to be run and data to be processed, so as to hold
The flow of row the application.Processing unit can be single processor or polycaryon processor in different embodiments.Some instruction hairs
Performed to GPU 815.The various calculating that GPU815 can provide processing unit 810 unload or helped processing unit 810 to provide
Image procossing.
Read-only storage (ROM) 830 stores processing unit 810 and static data and finger needed for electronic system other modules
Order.On the other hand, permanent storage appliance 835, it is a read-write memory equipment.This equipment is nonvolatile memory list
Member, it can even store instruction and data when electronic system 800 is closed.Some embodiments of the present application use massive store
Equipment (such as magnetic disc or laser disc and corresponding disk drive) is used as permanent storage appliance 835.
Other embodiment using a dismountable storage device (such as floppy disk, flash memory device, etc. and correspondingly
Disk drive) it is used as permanent storage appliance.As the system storage 820 of permanent storage appliance 835 is that a read-write memory is set
It is standby.But unlike storage device 835, system storage 820 is the read-write memory of a volatibility, such as at random
Access memory.System storage 820 stores some instructions required during processor operation and data.In certain embodiments,
Flow in the application is stored in system storage 820, permanent storage appliance 835, and/or read-only storage 830.Example
Such as, the instruction for being used for handling multimedia video is included according to some embodiments, different memory cells.Different deposited from these
Storage unit, processing unit 810 fetch the instruction to be performed and data to be processed to perform the flow in some embodiments.
Bus 805 is also connected to input and output equipment 840 and 845.Input equipment 840 causes user to be handed over electronic system
Change information and select command.Input equipment 840 includes alphanumeric (alphanumeric) keyboard and sensing equipment (is called
" cursor control device "), camera (such as IP Camera), receive the microphone of voice command or similar equipment, etc..
Output equipment 845 shows image or output data caused by electronic system.Output equipment 845 includes printer and display device,
Such as cathode-ray tube display (CRT) or liquid crystal display (LCD), and loudspeaker or similar audio output apparatus.Some
The equipment that embodiment includes such as Touch Screen, it can be used as input equipment and output equipment simultaneously.
Finally, as shown in figure 8, bus 805 couples electronic system 800 by network adapter (not shown) arrives network
825.In this way, computer can turn into computer network a part (be, for example, LAN (" LAN "), wide area network (" WAN ") or Intranet,
Or the network (network of networks) of network, such as Internet.The component root of any or all electronic system 800
It is used cooperatively according to the application.
Some embodiments include electronic building brick, such as microprocessor, and library computer program is instructed in machine readable or computer
Memory (can also be called computer-readable storage medium, machine readable medium or machine-readable storage media) in readable medium.
Some examples of such computer readable medium include RAM, ROM, read-only compact disk (CD-ROM), compact disc recordable (CD-
R), solid state drive (CD-RW), read-only digital versatile disc (such as DVD-ROM, DVD-dual layer-ROM), it is various it is recordable/
Rewritable DVD (such as DVD-RAM, DVD-RW, DVD+RW etc.), flash memories (such as SD card, mini-SD cards, micro-
SD card, etc.), it is magnetic and/or solid state hard disc, read-only with recordable Blu-Ray CDs, high density compact disc (ultra
Density optical discs), any other optically or magnetically medium, and floppy disk.Computer readable medium can be stored by extremely
The computer program of few processing unit execution simultaneously is used to perform various operations comprising one group of instruction.Computer program or computer code
Example can include machine code, such as the code as caused by compiler, comprising with computer, electronic component or as the micro- of translater
The code of the higher level of computing device.
Above discussion be substantially on perform software microprocessor or polycaryon processor, some embodiments by one or
Performed by multiple integrated circuits, e.g. application specific integrated circuit (ASIC) or field programmable gate array (FPGA).In some implementations
In example, the instruction that these integrated circuits perform is stored in circuit itself.Patrolled in addition, some embodiments perform to be stored in weave into
Collect the software in equipment (PLD), ROM or RAM device.
As used in present specification and claims, term " computer ", " server ", " processor ",
And " memory " all referring to electronics or other technologies equipment.These terms eliminate people or crowd.For illustrative purposes, art
Language " display " or " show tools " mean and shown on an electronic equipment.As in present specification and claims institute
As use, term " computer readable medium ", " computer readable medium " and " machine readable medium " is integrally all defined in storage
Information is deposited in the tangible physical objects of the form of a computer-readable.These terms eliminate wireless signal, wired download signal
And any other transient signal.
Although the description of the present application has been related to the description of various details, those skilled in the art can
The solution present invention can be implemented with other particular forms, without departing from the spirit of the present invention.In addition, multiple diagrams (including Fig. 3) are general
Illustrate flow to the property read.The concrete operations of these flows can strictly as shown in the figure and illustrate order perform.Concrete operations can
Without being performed in a series of continuous operations, different embodiments can perform different specific operations.Moreover, flow can use several subflows
Journey is implemented, or the part as bigger flow.Therefore, those skilled in the art it will be appreciated that the present invention be not limited to it is foregoing
Details, but to be defined by claim above.
Those skilled in the art will be noted that, after the guidance of the present invention is obtained, described device and method can be entered
The substantial amounts of modification of row and conversion.Correspondingly, above disclosure is construed as, and only passes through the boundary of attached claim
To limit.
Claims (25)
1. a kind of image processing method, comprising：
The first frame is handled by providing control setting to first group of equipment, to reach first performance measurement；
After first frame, the scene information on the second frame is received from second group of equipment；
Change between first frame and second frame is quantified；
According to the comparison accommodation control setting between the change of the quantization and predetermined threshold value；And
Control by providing the adjustment is set to first group of equipment to handle second frame.
2. image processing method as claimed in claim 1, it is characterised in that to the change between first frame and second frame
The step of being quantified includes：Compare one group of early stage indicator of first frame and one group of early stage indicator of second frame, its
In a frame one group of early stage indicator include can before frame event starts be used for predict the frame image processing load shape
State data.
3. image processing method as claimed in claim 1, it is characterised in that
When the change of the quantization is more than a specific threshold, the control setting to first group of equipment is adjusted, to reach the
Two performance metrics, wherein second performance metric measure a big specified quantitative than the first performance；And
When the change of the quantization is less than the specific threshold, the control to first group of equipment is adjusted based on the 3rd performance metric
Setting, wherein the 3rd performance metric and the difference of first performance measurement are less than a specified quantitative.
4. image processing method as claimed in claim 1, it is characterised in that control setting is supported comprising frequency with voltage
Operation of the first group of equipment in the frequency.
5. image processing method as claimed in claim 1, it is characterised in that first group of equipment includes graphics processing unit.
6. image processing method as claimed in claim 1, it is characterised in that second group of equipment include CPU,
Memory Controller, graphics processing unit, the scene information are included and produced by the graphics processing unit and the CPU
One group of raw data.
7. image processing method as claimed in claim 1, it is characterised in that the scene information include it is following in it is at least one：
The CPU load calculated using/game engine/game physics；
Graphics processing unit renders/calculated the application programming interfaces track of standard；
Vertex coloring run time and complexity；
Inlay run time and complexity；
The quantity of the block of block list-covering；
The quantity of post-processing object layer；
Every layer of resolution and total number of blocks；
Application programming interfaces type；
Pixel shader run time and complexity；
Texture type, size, layer, run time, and complexity；
User interface event；And
Display quantity.
8. image processing method as claimed in claim 1, it is characterised in that also include：
When the graphics processing unit is handling second frame, a particular event is detected in the graphics processing unit；
The performance metric of event discrimination 1 the 4th based on the detection；And
Adjust and set to the control of first group of equipment to reach the 4th performance metric.
9. image processing method as claimed in claim 8, it is characterised in that the event of the detection is related to timestamp, wherein
The step of detecting first performance measurement is comprising the expeced time for comparing the timestamp and the particular event.
10. image processing method as claimed in claim 8, it is characterised in that the step of detecting the particular event includes monitoring
The step of particular event, includes：The event occurred when CPU is loaded is monitored, at vertex coloring stage graphics processing unit
The event occurred and the event occurred at pixel shader stage graphics processing unit.
11. image processing method as claimed in claim 1, it is characterised in that more wrap the step of second frame for handling display
Contain：Receive the scene information on second frame from second group of equipment, and the scene information based on the reception adjust this first
The control setting of group equipment.
12. image processing method as claimed in claim 1, it is characterised in that second performance metric is one independently of this
The predetermined value of the change of quantization.
13. image processing method as claimed in claim 1, it is characterised in that second performance metric is based on the quantization
Change rather than the first performance measure distinguished predetermined value.
14. a kind of image processing method, comprising：
Frame is handled by providing control setting to one group of equipment, to reach first performance measurement；
When graphics processing unit is handling the frame, particular event is detected at the graphics processing unit；
The performance metric of event discrimination second based on the detection；And
The control for adjusting this group of equipment is set to reach second performance metric.
15. image processing method as claimed in claim 14, it is characterised in that carried to reach first performance measurement
The control setting supplied is one group of early stage indicator based on the frame, and the group early stage indicator, which includes, to be used to start it in frame event
The available status data of the preceding image processing load for predicting the frame.
16. image processing method as claimed in claim 14, it is characterised in that the event of the detection is relevant with timestamp, its
The step of middle discrimination second performance metric, includes the expeced time for comparing the timestamp and the particular event.
17. image processing method as claimed in claim 14, it is characterised in that control setting includes frequency and voltage, comes
Support the operation of this group of equipment under the frequency.
18. image processing method as claimed in claim 14, it is characterised in that first performance measurement is one based on the frame
Group scene information and it is discernable, this group of scene information includes following at least one：
The CPU load calculated using/game engine/game physics；
Graphics processing unit renders/calculated the application programming interfaces track of standard；
Vertex coloring run time and complexity；
Inlay run time and complexity；
The quantity of the block of block list-covering；
The quantity of post-processing object layer；
Every layer of resolution and total number of blocks；
Application programming interfaces type；
Pixel shader run time and complexity；
Texture type, size, layer, run time, and complexity；
User interface event；And
Display quantity.
19. image processing method as claimed in claim 14, it is characterised in that control setting is comprising following at least one：
Switch graphics processing unit or the power supply of its subprocess；
The frequency and voltage of deceleration/acceleration graphics processing unit/CPU and its subprocess；
The early stage of equipment comprising CPU and graphics processing unit wakes up or early stage accelerates；
The adjustment of bandwidth of memory and resolving strategy；And
Display frame rate and final strategy.
20. image processing method as claimed in claim 14, it is characterised in that the step of detecting the particular event includes monitoring
The step of particular event, includes：The event occurred when CPU is loaded is monitored, at vertex coloring stage graphics processing unit
The event occurred and the event occurred at pixel shader stage graphics processing unit.
21. a kind of image processing apparatus, comprising：
One group of processing unit；
Graphics processing unit；
Display device；And
Computer-readable storage medium, store multigroup instruction, wherein this it is multigroup instruction by the execution of this group of processing unit be by with
This group of processing unit is put to perform following action：
When the frame that graphics processing unit processing will be shown on the display apparatus, there is provided the image procossing list is given in control setting
Member reaches first performance measurement with the display device；
When the graphics processing unit is handling the frame, a particular event is detected at the graphics processing unit；
The performance metric of event discrimination second based on the detection；And
The control setting of the graphics processing unit is adjusted based on second performance metric.
22. such as the image processing apparatus of claim 21, it is characterised in that the event of the detection is relevant with timestamp, wherein distinguishing
Not second performance metric the step of include the expeced time for comparing the timestamp and the particular event.
23. such as the image processing apparatus of claim 21, it is characterised in that control setting includes frequency and voltage, to support
The operation of the graphics processing unit under the frequency.
24. such as the image processing apparatus of claim 21, it is characterised in that the control sets the frame per second control for including the display device
System.
25. such as the image processing apparatus of claim 24, it is characterised in that further include CPU, wherein this group of scene
Information on load of the packet containing the CPU.
Applications Claiming Priority (4)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US201662361039P | 2016-07-12 | 2016-07-12 | |
| US62/361,039 | 2016-07-12 | ||
| US15/606,132 US20170262955A1 (en) | 2017-05-26 | 2017-05-26 | Scene-Aware Power Manager For GPU |
| US15/606,132 | 2017-05-26 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107610039A true CN107610039A (en) | 2018-01-19 |
Family
ID=61059734
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710555428.6A Withdrawn CN107610039A (en) | 2016-07-12 | 2017-07-10 | Image processing method and image processing apparatus |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN107610039A (en) |
| TW (1) | TWI633517B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108648259A (en) * | 2018-03-27 | 2018-10-12 | 广东欧珀移动通信有限公司 | image drawing method, device, storage medium and intelligent terminal |
| CN108765531A (en) * | 2018-03-27 | 2018-11-06 | 广东欧珀移动通信有限公司 | Image rendering method, device, storage medium and intelligent terminal |
| CN109165103A (en) * | 2018-10-15 | 2019-01-08 | Oppo广东移动通信有限公司 | Frame rate control method, device, terminal and storage medium |
| WO2019153993A1 (en) * | 2018-02-09 | 2019-08-15 | Huawei Technologies Co., Ltd. | Video playback energy consumption control |
| WO2021052070A1 (en) * | 2019-09-19 | 2021-03-25 | 华为技术有限公司 | Frame rate identification method and electronic device |
| 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 |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109918125B (en) * | 2019-03-20 | 2022-06-03 | 浪潮商用机器有限公司 | GPU configuration method and device based on OpenPOWER architecture |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101126954A (en) * | 2002-08-22 | 2008-02-20 | 辉达公司 | Method and apparatus for adaptive power consumption |
| CN103870213A (en) * | 2012-12-18 | 2014-06-18 | 辉达公司 | Triggering performance event capture via pipelined state bundles |
| CN104956321A (en) * | 2012-12-29 | 2015-09-30 | 英特尔公司 | Identification of GPU phase to determine GPU scalability during runtime |
| US20150317762A1 (en) * | 2014-04-30 | 2015-11-05 | Qualcomm Incorporated | Cpu/gpu dcvs co-optimization for reducing power consumption in graphics frame processing |
| US20160054782A1 (en) * | 2014-08-19 | 2016-02-25 | Nikos Kaburlasos | Dynamic scaling of graphics processor execution resources |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6820209B1 (en) * | 1999-07-15 | 2004-11-16 | Apple Computer, Inc. | Power managed graphics controller |
| US7401242B2 (en) * | 2005-09-27 | 2008-07-15 | International Business Machines Corporation | Dynamic power management in a processor design |
| TW200720881A (en) * | 2005-11-16 | 2007-06-01 | Micro Star Int Co Ltd | Efficiency optimization method of hardware components by use of adjusting working frequencies |
| CN101281415A (en) * | 2007-04-06 | 2008-10-08 | 上海摩飞电子科技有限公司 | Method for regulating dynamic voltage frequency in power supply management technique |
| US8458497B2 (en) * | 2007-10-11 | 2013-06-04 | Qualcomm Incorporated | Demand based power control in a graphics processing unit |
| US8271812B2 (en) * | 2010-04-07 | 2012-09-18 | Apple Inc. | Hardware automatic performance state transitions in system on processor sleep and wake events |
| CN102520754B (en) * | 2011-12-28 | 2013-10-23 | 东南大学 | Dynamic voltage scaling system-oriented on-chip monitoring circuit |
| CN103019367B (en) * | 2012-12-03 | 2015-07-08 | 福州瑞芯微电子有限公司 | Embedded type GPU (Graphic Processing Unit) dynamic frequency modulating method and device based on Android system |
| US9606605B2 (en) * | 2014-03-07 | 2017-03-28 | Apple Inc. | Dynamic voltage margin recovery |
| US9905199B2 (en) * | 2014-09-17 | 2018-02-27 | Mediatek Inc. | Processor for use in dynamic refresh rate switching and related electronic device and method |
-
2017
- 2017-07-10 CN CN201710555428.6A patent/CN107610039A/en not_active Withdrawn
- 2017-07-11 TW TW106123178A patent/TWI633517B/en not_active IP Right Cessation
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101126954A (en) * | 2002-08-22 | 2008-02-20 | 辉达公司 | Method and apparatus for adaptive power consumption |
| CN103870213A (en) * | 2012-12-18 | 2014-06-18 | 辉达公司 | Triggering performance event capture via pipelined state bundles |
| CN104956321A (en) * | 2012-12-29 | 2015-09-30 | 英特尔公司 | Identification of GPU phase to determine GPU scalability during runtime |
| US20150317762A1 (en) * | 2014-04-30 | 2015-11-05 | Qualcomm Incorporated | Cpu/gpu dcvs co-optimization for reducing power consumption in graphics frame processing |
| US20160054782A1 (en) * | 2014-08-19 | 2016-02-25 | Nikos Kaburlasos | Dynamic scaling of graphics processor execution resources |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019153993A1 (en) * | 2018-02-09 | 2019-08-15 | Huawei Technologies Co., Ltd. | Video playback energy consumption control |
| CN111684485A (en) * | 2018-02-09 | 2020-09-18 | 华为技术有限公司 | Video playback energy consumption control |
| CN111684485B (en) * | 2018-02-09 | 2023-09-29 | 华为技术有限公司 | Video playback energy consumption control |
| CN108648259A (en) * | 2018-03-27 | 2018-10-12 | 广东欧珀移动通信有限公司 | image drawing method, device, storage medium and intelligent terminal |
| CN108765531A (en) * | 2018-03-27 | 2018-11-06 | 广东欧珀移动通信有限公司 | Image rendering method, device, storage medium and intelligent terminal |
| CN108648259B (en) * | 2018-03-27 | 2022-07-15 | Oppo广东移动通信有限公司 | Image drawing method and device, storage medium and intelligent terminal |
| CN109165103A (en) * | 2018-10-15 | 2019-01-08 | Oppo广东移动通信有限公司 | Frame rate control method, device, terminal and 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 |
| WO2021052070A1 (en) * | 2019-09-19 | 2021-03-25 | 华为技术有限公司 | Frame rate identification method and electronic device |
Also Published As
| Publication number | Publication date |
|---|---|
| TW201802768A (en) | 2018-01-16 |
| TWI633517B (en) | 2018-08-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107610039A (en) | Image processing method and image processing apparatus | |
| CN111045814B (en) | Resource scheduling method and terminal equipment | |
| US20170262955A1 (en) | Scene-Aware Power Manager For GPU | |
| CN108337358B (en) | Application cleaning method and device, storage medium and electronic equipment | |
| US9286084B2 (en) | Adaptive hardware reconfiguration of configurable co-processor cores for hardware optimization of functionality blocks based on use case prediction, and related methods, circuits, and computer-readable media | |
| US10831384B2 (en) | Memory device with power management | |
| CN110308782A (en) | Power consumption prediction, control method, equipment and computer readable storage medium | |
| CN104335175A (en) | Methods and systems to identify and migrate threads among system nodes based on system performance metrics | |
| US20180373562A1 (en) | Controlling operation of a gpu | |
| US20210247987A1 (en) | Algorithm program loading method and related apparatus | |
| CN113222403B (en) | Big data-based power regulation method and device, storage medium and electronic equipment | |
| CN103701934A (en) | Resource optimal scheduling method and virtual machine host machine optimal selection method | |
| CN109275011A (en) | The processing method and processing device of smart television motor pattern switching, user equipment | |
| CN110795400B (en) | File management method, device, equipment and medium | |
| CN114936085A (en) | ETL scheduling method and device based on deep learning algorithm | |
| CN103942807A (en) | Real-time processing fast image block fusion system and fusion method | |
| CN114063885A (en) | Method, apparatus and computer program product for managing storage space | |
| CN106649654A (en) | Data updating method and device | |
| CN112711387A (en) | Method and device for adjusting capacity of buffer area, electronic equipment and readable storage medium | |
| CN105988823B (en) | Memory allocation method and its system when in terminal using starting | |
| CN112256462B (en) | Service life prediction method, device and medium of NAND Flash memory | |
| CN104050100B (en) | A kind of data flow memory management method and system suitable for big data environment | |
| CN116521350B (en) | ETL scheduling method and device based on deep learning algorithm | |
| CN108681480B (en) | Background application program control method and device, storage medium and electronic equipment | |
| CN105224389B (en) | Based on the virtual machine resource integration method that linear dependence and segmenting vanning are theoretical |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20180119 |
|
| WW01 | Invention patent application withdrawn after publication |