CN101908200A - Graphics processing system with power gating function and method - Google Patents

Graphics processing system with power gating function and method Download PDF

Info

Publication number
CN101908200A
CN101908200A CN2009101392911A CN200910139291A CN101908200A CN 101908200 A CN101908200 A CN 101908200A CN 2009101392911 A CN2009101392911 A CN 2009101392911A CN 200910139291 A CN200910139291 A CN 200910139291A CN 101908200 A CN101908200 A CN 101908200A
Authority
CN
China
Prior art keywords
tinter
picture
power gating
graphics processing
running
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.)
Granted
Application number
CN2009101392911A
Other languages
Chinese (zh)
Other versions
CN101908200B (en
Inventor
杨佳玲
王柏翰
郑育镕
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Institute for Information Industry
Original Assignee
Institute for Information Industry
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Institute for Information Industry filed Critical Institute for Information Industry
Priority to CN2009101392911A priority Critical patent/CN101908200B/en
Publication of CN101908200A publication Critical patent/CN101908200A/en
Application granted granted Critical
Publication of CN101908200B publication Critical patent/CN101908200B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Image Generation (AREA)
  • Power Sources (AREA)

Abstract

The invention provides a graphics processing system with a power gating function and a power grating method. The power grating method is applicable to a graphics processing unit which is composed of a shader integrating unit, wherein the shader integrating unit consists of multiple shaders. The power grating method comprises the following steps: drawing multiple front pictures; calculating the number of first operation shaders used for drawing all the front pictures and corresponding picture velocities; determining the number of second operation shaders used for drawing the next picture behind the front pictures according to the number of the first operation shaders used for drawing all the front pictures and the corresponding picture velocities; and starting the corresponding shaders by one or more power grating elements based on the number of the second operation shaders. The graphics processing system with an improved power grating function can achieve the purpose of saving power consumption based on the requirements of various graphics application programs.

Description

The graphics processing system of tool power gating function and method
Technical field
The invention relates to draws handles, and particularly relevant for graphics processing system with power gating function and power gating method, according to the variation of picture speed, comes the tinter quantity of the required running of performance prediction.
Background technology
Generally speaking, the drawing application program comprises the graphic plotting of complicated and high details, for example: three-dimensional (3D) draws, and in order to meet ever-increasing in this respect at present demand, set graphics processing unit (graphics processing unit in PC or mancarried device, GPU) be used to handle a large amount of calculating, in order to show various items, also therefore very power consumption.Further, for making battery-powered mancarried device, similarly be mobile phone, because power consumption is the problem of a particular importance, thereby the power consumption that graphics processing unit produced in the minimizing mobile phone is necessary.
In electronic component, the source of general power consumption mainly comprises: supply voltage and dynamic power consumption and because the static power consumption that the electric leakage consume is caused that frequency of operation produced.Along with the development of semiconductor process techniques, the caused static power consumption of electric leakage consume has become subject matter.The technology following with 65 nanometers (65nm) is example, and the power consumption above 40% results to leak electricity consumes.
Known, such as clock pulse lock control (clock-gating) technology or dynamic electric voltage frequency modulating (dynamic voltage and frequency scaling, DVFS) technology, be to be Power Cutback mode commonly used, both effect ground reduces dynamic power consumption, but be helpless to reduce the electric leakage consume, or help limited.In addition, other known way, power gating (power-gating) technology for example on whole graphics processing unit,, but lacks flexibility the power gating arrangements of components to the control of powering of whole graphics processing unit by the power gating element.Perhaps, with the power gating arrangements of components in each element of inside.When an element when idle, close the power supply of feed element by the power gating element of correspondence, and then reduce simultaneously dynamically and static power consumption.Yet this power gating mechanism needs extra control circuit, is used for opening or closing the power supply of supplying with each element, therefore still has power consumption.In addition, when carrying out power gating function, need the extra time (overhead) recover to supply with the power supply of each element, make the known consuming time and inefficiency of power gating mechanism.
Therefore, need a kind of graphics processing unit, have improved power gating function, be able to reach the purpose of saving power consumption according to the demand of various drawing application programs.
Summary of the invention
Embodiments of the invention provide a kind of graphics processing system with power gating function, comprise a graphics processing unit and a driver.Described graphics processing unit comprises an integration tinter unit and one or more power gating element.Described integration tinter unit comprises a plurality of tinters.Described these tinters are in order to draw a plurality of preceding pictures.Described one or more power gating element is coupled to described these tinters, in order to according to the described second running tinter quantity, starts corresponding tinter.Described driver is coupled to described graphics processing unit, calculate one first a running tinter quantity and a corresponding picture speed of drawing each preceding picture, and, draw one second running tinter quantity of described these preceding pictures next picture afterwards in order to decision according to the described first running tinter quantity of picture before each and the described picture speed of correspondence.
On the other hand, embodiments of the invention provide a kind of graphics processing system with power gating function, comprise a graphics processing unit and a driver.Described graphics processing unit comprises an integration tinter unit and one or more power gating element.Described integration tinter unit comprises a plurality of tinters.Described these tinters are in order to draw a plurality of preceding pictures.Described one or more power gating element is coupled to described these tinters, in order to according to the described second running tinter quantity, starts corresponding tinter.Described driver is coupled to described graphics processing unit, calculate one first a running tinter quantity and a corresponding picture speed of drawing each preceding picture, and, draw one second running tinter quantity of described these preceding pictures next picture afterwards in order to decision according to the described first running tinter quantity of picture before each and the described picture speed of correspondence.
Said method of the present invention comes the tinter quantity of the required running of performance prediction according to the variation of picture speed, and the demand according to various drawing application programs reaches the purpose of saving power consumption in addition.
Description of drawings
Fig. 1 is the power gating method synoptic diagram that shows according to graphics processing system of the present invention.
Fig. 2 is a graphics processing unit calcspar that shows according to the embodiment of the invention.
Fig. 3 is the graphics processing system with power gating function that shows according to the embodiment of the invention.
Fig. 4 is a power gating method flow diagram that shows according to the embodiment of the invention.
Fig. 5 is a graphics processing unit calcspar that shows according to another embodiment of the present invention.
Fig. 6 is a graphics processing unit calcspar that shows according to another embodiment of the present invention.
Drawing reference numeral:
302~graphics processing unit;
304~driver;
308~moderator;
310~command processor;
312~application program interface;
314,316~application program;
The corresponding I/O of 318~storer;
320~integration tinter unit;
320A, 320B, 320C, 320D~tinter;
328A, 328B, 328C, 328D~power gating element;
Vdd~voltage source; And
Ck~clock pulse.
Embodiment
For above-mentioned purpose of the present invention, feature and advantage can be become apparent, embodiment cited below particularly, and cooperate appended accompanying drawing, be described in detail as follows.
Fig. 1 is the power gating method synoptic diagram that shows according to graphics processing system of the present invention.
In the embodiment of Fig. 1, be to add the power gating control circuit in the graphics processing unit 102, cause described graphics processing unit 102 to have power gating function.Further, draw pictures (frame) before in described graphics processing unit 102, in advance by the required picture speed of the described picture of a driver 104 decisions (frame rate per second, FPS), in order to represent the drawing amount of described picture.Therefore, described driver 104 is according to the drawing amount of described picture, the electric power starting of controlling correlation function element in the described graphics processing unit 102 with close, using is not influencing the user and (for example: the fluency that picture presented) under the prerequisite, improve the overall power consumption of described graphics processing unit 102 is drawing usefulness.
Specifically, in an embodiment, when described graphics processing unit 102 was finished the drafting operation of a picture, described driver 104 calculated the pairing picture speed of described picture, in order to determine the required picture speed of next picture (shown in arrow 106).Then, before described graphics processing unit 102 carries out the drafting operation of described next picture, described driver 104 is controlled described graphics processing unit 102 just according to the picture speed of next picture, for example: set the electric power starting of correlation function element and close (shown in arrow 108).
In other embodiment, more can utilize a plurality of pairing picture speeds of preceding picture, determine the picture speed that next picture is required.
Fig. 2 is a graphics processing unit 202 calcspars that show according to the embodiment of the invention.
With reference to figure 2, described graphics processing unit 202 has an integration tinter unit 220 (unified shader unit), and described integration tinter unit 220 is a multi-processor, can handle a plurality of instructions in single clock pulse.Described integration tinter unit 220 comprises a plurality of tinters (shader processors or shader cores), as 220A, 220B, 220C etc.Each tinter can be very long instruction word (VLIW) collection (the very long instruction word of a kind of vector (vector) and scale (scalar) framework or a kind of most values (multiple scalar), VLIW) framework, and corresponding working storage archives and instruction memory cache are arranged.Moreover each tinter is carried out various coloration program, is responsible for carrying out the operation of vertex coloring (vertex shader) and pixel painted (pixel shader), in order to draw each picture.In addition, described graphics processing unit 202 more comprises: how much stages of fixed function (fixed-function geometry stages) 204, fixed function fragment stage (fixed-function fragment stages) 206, one moderator 208 and a command processor 210.
Particularly, how much stages 204 of described fixed function comprise a clipper (clipper) 212, a pel assembled unit (primitive assembly) 214 and a string stream receiver (streamer) 216.In this pipeline stage, described crossfire receiver 216 receives the vertex data of a 3D object, and is sent to tinter.Afterwards, described these tinters 220A, 220B, 220C carry out the vertex data character that relevant coloring process decides described 3D object, in order to convert described 3D object to be shown on the screen picture.Then, described pel assembled unit 214 is carried out geometrical combination, in order to the set of vertices composite polygon, similarly is triangle.Described clipper 212 is eliminated the triangle beyond the viewing area.
Further, the described fixed function fragment stage 206 comprises that a triangle sets up unit 222, a fragment generation unit 224, a hierarchy type advanced treatment unit 226, one degree of depth/template (Z/stencil) test cell 228, an interpolation unit 230 and a delineation unit 232.In this pipeline stage, described triangle is set up unit 222 and is carried out plane elimination (face-culling), the triangle that removal can't show, and calculate leg-of-mutton side equation preface.Described fragment generation unit 224 provides leg-of-mutton fragment systematic function, in order to calculate the pixel (pixel) that desire shows.Described hierarchy type advanced treatment unit 226 optionally was disposed in the described fixed function fragment stage 206, in order to giving up the fragment beyond triangle or the vision area, and can once give up the entire segment block.The described degree of depth/template test unit 228 is utilized depth buffer and stencil buffer, judges and gives up the fragment that is hidden.The triangle attribute that described interpolation unit 230 will carry out after the perspective correction carries out interpolation, in order to produce the attribute of fragment.Subsequently, carry out pixel by described delineation unit 232 again and describe operation.In an embodiment, the described degree of depth/template test unit 228 is also configurable after described delineation unit 232.
In operation, described command processor 210 is in order to receiving various drawing commands, and monitoring and set the power supply status of described these tinters.Described moderator 208 carries out Thread scheduling (thread scheduling) according to various drawing commands, and drawing command is dispensed to each tinter, in order to carry out 3D drawing computing.Because a large amount of drawing computings is responsible in described integration tinter unit 220, therefore become the bottleneck of whole graphics processing unit 202 power consumptions.Further, because the drawing amount difference that each picture is required, therefore can come to the control of powering of each tinter, for example by power gating: but respectively the power gating element of tinter 220A, 220B, 220C is carried out the action of conducting or cut-out.Thus, can effectively reduce the dynamic power consumption or the electric leakage consume of described graphics processing unit 202, make the overall power consumption of described graphics processing unit significantly reduce, and do not influence the execution usefulness of application program.
Fig. 3 is the graphics processing system calcspar with power gating function that shows according to the embodiment of the invention.
With reference to figure 3, described graphics processing system comprises a graphics processing unit 302 and a driver 304.In embodiment shown in Figure 3, described graphics processing unit 302 comprises an integration tinter unit 320, and it has 4 tinter 320A, 320B, 320C and 320D, in order to draw a plurality of pictures.Be similar to integration tinter unit 220 shown in Figure 2, described integration tinter unit 320 is a multi-processor, can handle a plurality of instructions in single clock pulse Ck.Further, described graphics processing unit 302 comprises 4 power gating element 328A, 328B, 328C and 328D, is respectively coupled to each tinter.Described these power gating elements are according to control signal 330A, 330B, 330C and the 330D of correspondence, in order to start or to close the tinter of correspondence.Described driver 304 is coupled to described graphics processing unit 302, from an application program interface (application programming interface, API) 312 receive and carry out various application programs, for example: one first application program 314, one second application program 316 etc., draw in order to drive described graphics processing unit 302 accordingly.The power gating method of described graphics processing system will cooperate the 3rd and 4 figure to be described in detail as follows.
Fig. 4 is a power gating method 40 process flow diagrams that show according to the embodiment of the invention.
As mentioned above, draw a set picture Frame at described graphics processing unit 302 N+1Before, described driver 304 can use a kind of based on historical (history-based) account form, with previous rendered frame Frame n, Frame N-1..., Frame N-m+1Picture speed FPS n, FPS N-1..., FPS N-m+1And the tinter quantity S that is operated n, S N-1..., S N-m+1Be the basis, draw described set picture Frame in order to predict described graphics processing unit 302 N+1The tinter quantity S that Shi Suoxu starts N+1Wherein, m represents the preceding number of pictures in order to prediction.Then,,, make the computing more efficiently of described graphics processing unit 302, and improve the overall power consumption of described graphics processing system in order to open or to close the power supply of corresponding tinter by described these power gating elements.
Further, described driver 304 is that request according to each application program drives described graphics processing unit 302 and carries out various mapping operations.Therefore, described driver 304 also can be judged the beginning and the end of pattern drafting by the request of application program, in order to carry out the operation of power gating.For example, described first application program 314 can comprise an instruction SwapBuffer, and in order to the end of expression pattern drafting, and application program 316 can comprise an instruction ClearBuffer, in order to the end of expression pattern drafting.The term of execution of above-mentioned instruction, carry out the power gating operation, just can not have influence on the usefulness of drawing.
With reference to the 3rd and 4 figure, when described driver 304 when described application program interface 312 receives described first application program 314, described driver 304 produces corresponding order package, is sent to a command processor 310 (step S402) of described graphics processing unit 302 by the corresponding I/O of storer (memory-mapped I/O) 318.
In response to the execution of described instruction SwapBuffer, that is represent described set picture Frame N+1Previous picture Frame nComplete, described driver 304 calculates described these preceding picture Frame immediately n, Frame N-1..., Frame N-m+1The tinter quantity S that is operated n, S N-1..., S N-m+1And corresponding picture speed FPS n, FPS N-1..., FPS N-m+1(step S404).
By way of example, when m=5, represent that described driver 304 calculates described set picture Frame N+1Preceding 5 rendered frame Frame n, Frame N-1..., Frame N-4The tinter quantity S that is operated n, S N-1..., S N-4And corresponding picture speed FPS n, FPS N-1..., FPS N-4
In addition, driver 304 is according to these preceding picture Frame n, Frame N-1..., Frame N-m+1Running tinter quantity S n, S N-1..., S N-m+1And corresponding picture speed FPS n, FPS N-1..., FPS N-m+1, draw described set picture Frame in order to decision N+1The tinter quantity S of required running N+1(step S406).
More specifically, described driver 304 can be drawn described set picture Framen according to the following formula decision N+1Required tinter quantity S N+1:
Figure B2009101392911D0000081
Wherein, m is quantity, the S of described these preceding pictures n, S N-1..., S N-m+1For drawing described these preceding picture Frame n, Frame N-1..., Frame N-m+1The tinter quantity, the FPS that are operated n, FPS N-1..., FPS N-m+1Be described these preceding picture Frame m, Frame N-1..., Frame N-m+1Pairing picture speed, Target_FPS are a controlled variable and n 〉=m for a target picture speed, the α according to the adjustment of demonstration demand.
Afterwards, when described driver 304 when described application program interface 312 receives described second application program 316 (step S408), in response to the execution of described instruction ClearBuffer, that is expression begins to carry out described set picture Frame N+1Drafting operation, described driver 304 is according to described set picture Frame N+1The tinter quantity S of required running N+1And the power supply status of present each tinter 320A, 320B, 320C and 320D, produce a corresponding order package, in order to the electric power starting that disposes described these tinters 320A, 320B, 320C and 320D and close, and described corresponding order package is sent to described command processor 310 (step S410).In an embodiment, suppose described set picture Frame N+1A picture Frame before nEmployed tinter quantity S nGreater than drawing described set picture Frame N+1The tinter quantity S of required running N+1, then close the power supply supply of idle (inactive) tinter accordingly.Otherwise, then open the power supply supply that needs running (active) tinter accordingly.
Next, in response to described order package, described command processor 310 produces in order to conducting or cuts off control signal 330A, 330B, 330C and the 330D of each power gating element 328A, 328B, 328C and 328D, use the electric power starting of setting described these tinters 320A, 320B, 320C and 320D or close, and notify a moderator 308 (step S412).
Afterwards, described moderator 308 can carry out drawing command and distribute according to the tinter that has started, carries out described set picture Frame N+1Drafting operation (step S414), for example: the mode of can the segment formula drawing (tile-based rendering) is drawn described set picture Frame N+1
In an embodiment, each power gating element comprises a transistor.As shown in Figure 3, each power gating element comprises a nmos pass transistor, be coupled between a voltage source V dd and each tinter, and its gate receives the control signal that described command processor 310 is sent.Whether therefore, each transistor is switched on or cuts off according to the control signal of correspondence, described voltage source V dd is provided to each tinter in order to decision.
Further, in operation, each tinter 320A, 320B, 320C and 320D can dispose a texture cell (texture unit) separately, perhaps share one or more texture cell.Therefore, each tinter 320A, 320B, 320C and 320D can receive data texturing from texture cell by texture access path 332,334,336 and 338 separately.In in the case, also can flexibly adjust the distribution of power gating element according to the configuration of texture cell.Thus, can significantly promote the efficient of power management.Above-mentioned power gating mechanism will cooperate Fig. 5 and Fig. 6 to be described in detail as follows.
Fig. 5 is a graphics processing unit 502 calcspars that show according to another embodiment of the present invention.
With reference to figure 5, described graphics processing unit 502 comprises an integration tinter unit 520, regional shared storage 512 and 514, texture cell 508 and 510, universe shared storage 516 and Thread processing unit 518.
In this embodiment, described integration tinter unit 520 has a plurality of tinters.Described these tinters comprise 2 tinters gather together (cluster) 504 and 506, use described regional shared storage 502 and 514 to draw separately.In addition, tinter is gathered together and 504 and 506 is coupled to 2 texture cells 508 and 510 separately, and described universe shared storage 516 is shared by texture cell 508 and 510.Particularly, each tinter is gathered together and is comprised 8 tinters separately.Described Thread processing unit 518 comprises 2 Thread sequencers 522 and 524, in order to carry out the distribution of Thread.
In in the case, 504 and 506 each self-configuring one power gating element but each tinter is gathered together.Thus, the conducting of each power gating element or cut-out will start or close corresponding tinter and gather together.In addition, the control of powering of regional shared storage under also can be simultaneously gathering together and texture cell to each tinter.Not only reduce the cost of power gating control circuit, also can save the power consumption that tinter periphery related elements is caused.
Fig. 6 is a graphics processing unit 602 calcspars that show according to another embodiment of the present invention.
With reference to figure 6, described graphics processing unit 602 comprises an integration tinter unit 620, how much control modules 604, a tinter control module 606 and texture cells 608.
In this embodiment, described integration tinter unit 620 has a plurality of tinters.Described these tinters comprise 2 multiple painted processing units (shader multi-processor) 610 and 612, and described 2 multiple painted processing units 610 and 612 form a tinter and gather together, and use described texture cell 608 to draw jointly.Described how much control modules 604 and described tinter control module 606 are in order to receive data and to distribute drawing work.In Fig. 6, each multiple painted processing unit comprises 8 tinter SP, I and C memory cache (cache), multiple Thread release unit MT (multi-thread issue), 2 special function unit SFU (Special Function Unit) and shared storage MEM, in order to the computing of drawing.Under this framework, multiple painted processing unit 610 and 612 is shared described texture cells 608, therefore, these 2 multiple painted processing units 610 and 612 can be considered as a power management unit, by the control of powering of same power gating element.When described power gating element cut off, whole tinter was gathered together, that is multiple painted processing unit 610 and 612, and described texture cell 608 will be closed together.Reduce unnecessary power gating element and power consumption further.
Therefore, by graphics processing system of the present invention and power gating method thereof, when drawing, can be according to the picture speed variation relation of each picture, dynamically control needs the tinter quantity of running, thereby reduces unnecessary power consumption.
Method of the present invention, or specific kenel or its part can exist with the kenel of procedure code.Procedure code can be contained in tangible media, get (getting) Storage Media as floppy disk, CD, hard disk or any other machine readable as computer-readable, also or be not limited to the computer program product of external form, wherein, when procedure code by machine, when being written into and carrying out as computer, this machine becomes in order to participate in device of the present invention.Procedure code also can pass through some transfer mediums, transmit as electric wire or cable, optical fiber or any transmission kenel, wherein, when procedure code by machine, as computer receive, when being written into and carrying out, this machine becomes in order to participate in device of the present invention.When the general service processing unit is done in fact, procedure code provides a class of operation to be similar to the unique apparatus of using particular logic circuit in conjunction with processing unit.
Though the present invention discloses as above with preferred embodiment; right its is not in order to limit the present invention; anyly have the knack of those skilled in the art; without departing from the spirit and scope of the present invention; when can being used for a variety of modifications and variations, so protection scope of the present invention defines and is as the criterion when looking the claim scope.

Claims (18)

1. a power gating method is applicable to a graphics processing unit, it is characterized in that, described graphics processing unit has an integration tinter unit, and described integration tinter unit comprises a plurality of tinters, and described power gating method comprises:
Draw a plurality of preceding pictures;
Calculate one first a running tinter quantity and a corresponding picture speed of drawing each preceding picture;
According to the described first running tinter quantity and the corresponding described picture speed of picture before each, draw one second running tinter quantity of described these preceding pictures next picture afterwards in order to decision; And
According to the described second running tinter quantity, start corresponding tinter by one or more power gating element.
2. power gating method as claimed in claim 1 is characterized in that, described method more comprises:
Utilize the tinter that has started, in order to draw described next picture.
3. power gating method as claimed in claim 1 is characterized in that, described method more comprises:
Carry out one first application program; And
In response to the execution of described first application program, calculate the described first running tinter quantity and the corresponding described picture speed of each preceding picture, wherein, described first application program finishes corresponding to one of picture before each.
4. power gating method as claimed in claim 1 is characterized in that, determines the step of the described second running tinter quantity to comprise:
Before the described second running tinter quantity of decision, obtain a power supply status of each tinter.
5. power gating method as claimed in claim 4 is characterized in that, the step that starts described these tinters comprises:
Carry out one second application program; And
In response to the execution of described second application program, according to the described power supply status of described second running tinter quantity and each tinter, produce an order package, in order to controlling described one or more power gating element,
Wherein, described second application program one initial corresponding to described next picture.
6. power gating method as claimed in claim 5 is characterized in that, described method more comprises:
In response to described order package, produce one or more control signal,
Wherein, each power gating element is control signal conducting or the cut-out by a correspondence.
7. power gating method as claimed in claim 6, it is characterized in that, each power gating element comprises a transistor, is coupled between a voltage source and at least one tinter, and whether provide in order to the control signal decision according to described correspondence provides described voltage source to described at least one tinter.
8. power gating method as claimed in claim 1, it is characterized in that, described these tinters comprise that a plurality of tinters gathers together, be coupled to a plurality of texture cells, and wherein, each texture cell is coupled at least one tinter and gathers together, with so that each power gating element starts a texture cell simultaneously and described at least one tinter of coupling is thereon gathered together.
9. power gating method as claimed in claim 1 is characterized in that, the described second running tinter quantity is to produce according to following formula:
Wherein, S N+1Described second of described next picture operates tinter quantity, m is quantity, the S of described these preceding pictures in order to draw n, S N-1..., S N-m+1For drawing the described first running tinter quantity, the FPS of picture before each n, FPS N-1..., FPS N-m+1For the pairing described picture speed of picture before each, Target_FPS are that a target picture speed, α are a controlled variable and n 〉=m.
10. the graphics processing system with power gating function is characterized in that, described system comprises:
One graphics processing unit, comprise an integration tinter unit and one or more power gating element, wherein, described integration tinter unit comprises a plurality of tinters, in order to draw a plurality of preceding pictures, described one or more power gating element is coupled to described these tinters, in order to according to the described second running tinter quantity, starts corresponding tinter; And
One driver, be coupled to described graphics processing unit, calculate one first a running tinter quantity and a corresponding picture speed of drawing each preceding picture, and, draw one second running tinter quantity of described these preceding pictures next picture afterwards in order to decision according to the described first running tinter quantity of picture before each and the described picture speed of correspondence.
11. graphics processing system as claimed in claim 10 is characterized in that, the tinter that described graphics processing unit utilization has started is in order to draw described next picture.
12. graphics processing system as claimed in claim 10, it is characterized in that, when one first application program is carried out, execution in response to described first application program, described driver calculates the described first running tinter quantity and the corresponding described picture speed of each preceding picture, and wherein, described first application program finishes corresponding to one of picture before each.
13. graphics processing system as claimed in claim 10 is characterized in that, described driver is obtained a power supply status of each tinter before the described second running tinter quantity of decision.
14. graphics processing system as claimed in claim 13, it is characterized in that, when one second application program is carried out, execution in response to described second application program, described driver produces an order package, in order to control described one or more power gating element according to the described power supply status of the described second running tinter quantity and each tinter, and wherein, described second application program one initial corresponding to described next picture.
15. graphics processing system as claimed in claim 14 is characterized in that, described graphics processing unit comprises:
One command processor is coupled to described driver, produces one or more control signal, in order in response to described order package,
Wherein, each power gating element is control signal conducting or the cut-out by a correspondence.
16. graphics processing system as claimed in claim 15, it is characterized in that, whether each power gating element comprises a transistor, is coupled between a voltage source and at least one tinter, described voltage source is provided to described at least one tinter in order to the control signal decision according to described correspondence.
17. graphics processing system as claimed in claim 10, it is characterized in that, described these tinters comprise that a plurality of tinters gathers together, be coupled to a plurality of texture cells, and wherein, each texture cell is coupled at least one tinter and gathers together, with so that each power gating element starts a texture cell simultaneously and described at least one tinter of coupling is thereon gathered together.
18. graphics processing system as claimed in claim 10 is characterized in that, the described second running tinter quantity is to produce according to following formula:
Figure F2009101392911C0000041
Wherein, S N+1Described second of described next picture operates tinter quantity, m is quantity, the S of described these preceding pictures in order to draw n, S N-1..., S N-m+1For drawing the described first running tinter quantity, the FPS of picture before each n, FPS N-1..., FPS N-m+1For the pairing described picture speed of picture before each, Target_FPS are that a target picture speed, α are a controlled variable and n 〉=m.
CN2009101392911A 2009-06-05 2009-06-05 Graphics processing system with power gating function and method Active CN101908200B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2009101392911A CN101908200B (en) 2009-06-05 2009-06-05 Graphics processing system with power gating function and method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2009101392911A CN101908200B (en) 2009-06-05 2009-06-05 Graphics processing system with power gating function and method

Publications (2)

Publication Number Publication Date
CN101908200A true CN101908200A (en) 2010-12-08
CN101908200B CN101908200B (en) 2012-08-08

Family

ID=43263653

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2009101392911A Active CN101908200B (en) 2009-06-05 2009-06-05 Graphics processing system with power gating function and method

Country Status (1)

Country Link
CN (1) CN101908200B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016090641A1 (en) * 2014-12-12 2016-06-16 上海兆芯集成电路有限公司 Graphics processing system and power gating method thereof

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI354239B (en) * 2006-06-08 2011-12-11 Via Tech Inc Decoding system unit
CN101216932B (en) * 2008-01-03 2010-08-18 威盛电子股份有限公司 Methods of graphic processing arrangement, unit and execution triangle arrangement and attribute arrangement

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016090641A1 (en) * 2014-12-12 2016-06-16 上海兆芯集成电路有限公司 Graphics processing system and power gating method thereof
US10209758B2 (en) 2014-12-12 2019-02-19 Via Alliance Semiconductor Co., Ltd. Graphics processing system and power gating method thereof

Also Published As

Publication number Publication date
CN101908200B (en) 2012-08-08

Similar Documents

Publication Publication Date Title
TWI393067B (en) Graphics processing system with power-gating function, power-gating method, and computer program products thereof
US10783698B2 (en) Point cloud operations
US11049266B2 (en) Point cloud viewpoint and scalable compression/decompression
CN106537447B (en) Dynamic scaling of graphics processor execution resources
CN101740004B (en) Picture processing using a hybrid system configuration
US10162405B2 (en) Graphics processor power management contexts and sequential control loops
US10560892B2 (en) Advanced graphics power state management
EP3886036A1 (en) High-speed resume for gpu applications
US20170199558A1 (en) Flexible and scalable energy model for estimating energy consumption
CN106662999A (en) Method And Apparatus For Simd Structured Branching
CN106575430A (en) Method and apparatus for pixel hashing
WO2017160383A1 (en) Facilitating execution-aware hybrid preemption for execution of tasks in computing environments
US20190204894A1 (en) Resource load balancing based on usage and power limits
Kim et al. A reconfigurable SIMT processor for mobile ray tracing with contention reduction in shared memory
DE102020105902A1 (en) HARDWARE INDEX ASSIGNMENT MECHANISM
CN108734627A (en) Determine size by the adaptable caching of live load
CN102866765A (en) Graphics processing unit and management method thereof
CN101908200B (en) Graphics processing system with power gating function and method
CN107168457A (en) A kind of low-power consumption GPU SOC methods
DE102020108215A1 (en) Control surface access by means of flat memory allocation
US10839490B2 (en) Apparatus and method for runtime training of a denoising machine learning engine
CN113994363A (en) Method and apparatus for wave slot management
US20200311948A1 (en) Background estimation for object segmentation using coarse level tracking
Chang et al. A 172.6 mW 43.8 GFLOPS energy-efficient scalable eight-core 3D graphics processor for mobile multimedia applications
Lysecky et al. Efficient reconfiguration methods to enable rapid deployment of runtime reconfigurable systems

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant