Scribd Logo
Upload

Welcome to Scribd!

  • 83 views

    11605148

    Uploaded by

    HMiche93
    Sony patent about devices with multiple GPUs

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content

    You might also like

    11605148

    Uploaded by

    HMiche93
    83 views47 pages
    Sony patent about devices with multiple GPUs

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    Sony patent about devices with multiple GPUs

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    Download as pdf
    83 views47 pages

    11605148

    Uploaded by

    HMiche93
    Sony patent about devices with multiple GPUs

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    Download as pdf
    of 47
    ay United States Patent Cerny et al. USOL160514852 US 11,605,148 B2 *Mar. 14, 2023 (10) Patent No. 5) Date of Patent: os BY PRETESTING 4 REGIONS U INFORMATION INST SCREEN, ING PRIOR FRAME (71) Applicant: Sony Interuetive Entertainment Tne, Tokyo (IP) (72) Inventors: Mark B. Cerny, Los Angeles, CA (US): Florian Strauss, Sen Matoo, CA. (US): Tobias Berghoff, San Mateo, CA (us) (13) Assignee: Sony Interactive Entertainment In. Tokyo (IP) Notice: Subject to any diselaimes, the term of this patent is extended or adjusted under 35, USC. 154(5) by 0 days. This patent is subject to a terminal dis claimer QI) Appl. Now 17989311 (22) Filed: ut, 29, 2021 6) Prior Publication Data US 202100358075 A1 Now: 18,2 Related US. Application Data (63) Continuation of application No. 167780,722, filed on Feb, 3, 2020, now Pat, No, 11,080,814, 1) mec Goorin (2006.00) Goo 150 Gort 01) (Continued) ‘100A 701A 702A Previous Frame seeing rendeing cua LWA seus WHA ioe ceuc WY ceun WY 41308 ia 11008, G067 1720 2013.01): GOOF 9/3877 013.01): GU6E 98061 2013.01): GO6E 178 (2013.01) (Coatinved) (58) Fleld of Classiication Search (CPC sn: GOGP 9/5061: GOGE 9/3877: GOGP 17/18 GO6T 151005; GOST 15/10; GOBT 1/20, ‘ce application file for complete search history 66) References Cited US, PATENT DOCUMENTS (Continved) Primary Examiner 74) ntorney Phi Hoang, {gent or Firm — Peal TP, APC o ABSTRACT A method including rendering graphies for an application ‘sing gruphies processing units (GPUs). Responsibility for rendering of geometry is divided between GPUs based on freon regions, each GPU having a corresponding division ofthe esponsibility which is known Fist pieces of goon fry are rendered a the GPUs during a rendering phase of @ previous image frame, Sttsties are generated for the re fering. of the previous image frame. Second pisces of beometry ofa current image frame are assigned based on the Satis to the GPUs for geometry esting. Geometry testing fata curent image Irame on the second picoes of geomet js performed to generate information regarding cach piece of szonmetry and its relation to each sercen region, the geometry testing performed at each of the GPUs based on the assign ing. The information generated for the second pieces of eometry is used when rendering the peometry atthe GPU, 20 Claims, 19 Drawing Sheets 7018 7028 4) Guent Frame a cpeuA YY cpus WMA opuc Ye cpun GZ 11908 US 11,605,148 B2 Page 2 3) mec Gor 15/10 (ono) Gost 1770 (2006.01) Gor 9238 (01801) GU6E 9550 (2006.01) GOOF 1718 (2006.01) (2) US.cL ce Gu6r 150s (2013.01): GO6T 15/10 (2013.01): Gast 170 (2013.01) 60) References Cited US. PATENT DOCUMENTS 201400125682 AL* 572014 Rakai s6 5363 ‘955 201S!0371355 AL* 122015 Chen G6r 145833 basss 201910206029 AL® 72019 Diao ncn GOSF T2OED * cited by examiner US 11,605,148 B2 Sheet 1 of 19 Mar, 14, 2023 U.S. Patent mE s30uNOSSY NdO ia s30yNOS3u No MaNYSS ‘aWvO 1 v ANIONS ONISS300¥d SLL BWW 1 1 1 91901 | © @ @ | 31907 | | 21907 BWvO BWvo | BNO >RIOMLAN awy9 ano19 shi L Sls on on ‘BOING oe 30130 aNano ‘Nano 06 30IN30 Nano os U.S. Patent Mar. 14,2023 Sheet 2 of 19 US 11,605,148 B2 Legend Bus 220 Bus 200 us 240 2 8 So 2 as a w - State Rendering Ted of Draw 1 Geometry ms State ad Draw 2 ur State , Draw 3 FIG. 7A U.S. Patent Mar. 14,2023 Sheet 8 of 19 US 11,605,148 B2 700B 731 A B A B A : @ 732 Cc Cc D. Cc " |p a . A B A B A a c D c D c . FIG. 7B-1 Object Responsibility Object | Object | Object | Object o | 4 2 3 GPUA - pos x . GPUB mM ie x x a = x cpuD =X x x U.S. Patent Mar. 14,2023 Sheet 9 of 19 US 11,605,148 B2 oa 7564 eae Si tne rensonng \ GPU A fort 2 3 cpuB fo [7 2 3 GPu c [0 i 2, 3 E < 740 GPU D fo 1 2 3 700C-2 701" Speed Increase 702 745 Pretesting Rendering GPU A fo 2 GPU B 4 0 2 GPU C J2 1 GPU D f3| 1 2 Start of phase End of phase of rendering of rendering 750a 750b FIG. 7C U.S. Patent Mar. 14,2023 Sheet 10 of 19 US 11,605,148 B2 A m 1B a 8104 ‘6108 FIG. 8A A om |B a FIG. 8B US 11,605,148 B2 Sheet 11 of 19 Mar, 14, 2023 U.S. Patent 900A < o S iL 900B FIG. 9B 900C a 6 oO ¢ 1392 1991 FIG. 13A U.S. Patent Mar. 14,2023 Sheet 18 of 19 Rendering graphics for an application using a plurality of graphics processing units (GPUs) Y Dividing responsibilty for the rendering of geometry of the graphics between the plurality of GPUs based on a plurality of screen regions, each GPU having @ corresponding division of the responsibilty which is known to the plurality of GPUs Y Assigning a plurality of pieces of geometry of an image frame to the plurality of GPUs for geometry pretesting ’ Interleaving a first set of shaders to perform ‘geometry pretesting and rendering on a first set of, pieces of geometry with a second set of shaders to perform geometry pretesting and rendering on a second set of pieces of geometry FIG. 13B US 11,605,148 B2 1300B 1310 1320 1330 1340 U.S. Patent Mar. 14,2023 Sheet 19 of 19 US 11,605,148 B2 1400 re | ewory i cpu i | \ STORAGE oa i ! I 1408 \ T ee Se See - USER INPUTDEVICE aie jf eRPaCSWeENOR | | CACHE I ! I ru I | _ GRAPHICS SUBSYSTEM | 1409 Bee Y sata NETWORK INTERFACE ee 410 we AUDIO PROCESSOR FIG. 14 US 11,605,148 B2 1 SYSTEM AND METHOD FOR -MULILGPU RENDERING OF GEOMETRY BY PRETESTING AGAINST SCREEN REGIONS USING PRIOR FRAME INFORMATION CLAIM OF PRIORITY This application is a continuation of and claims proty to and the Benefit of commonly owned, patent application US. Ser. No. 161780,722, filed on Feb. 3, 2020, 2019, enttlad “System and Method for Eiiceat MultiGPU Rendering of Geometry by Preesting Against Sereen Regions Using Prior ame Information,” the dislosore of whichis incorporated herein in is entirety foe all purposes ‘CROSS REFERENCE TO RELATED APPLICATIONS ‘This application is related to commonly assigned, co- pending US. patent application Ser. No. 161780,680, fnitled “SYSTEM AND METHOD FOR EFFICIENT MULILGPU RENDERING OF GEOMETRY BY PRE: TESTING AGAINST INTERLEAVED SCREEN REGIONS BEFORE. RENDERING,” fled concuneatly with the present application, the disclosure of which is hereby incorporated by reference in its enltety. This appli cation is related 10 commonly assigned, co-pending US, patent application Ser No, 16/780,785, ened “SYSTEM AND METHOD FOR ECFICIENT MULTLGPU REN: DERING OP GEOMETRY BY PRETESTING AGAINST SCREEN REGIONS USING CONFIGURABLE SHAD- TERS,” filed concurrently with the present application, the alislosure of which is hereby incorporated by reference i its entirety TECHNICAL FIELD ‘The present disclosure it related 10 graphic processing, tnd more specifically for mulG-GPU collaboration when rendering an image for an application. BACKGROUND OF THE DISCLOSURE ‘In gocent years there has been continual push for online serviges tht allow fr online o cond gaming in streaming Toemat botween a cloud gaming server an client con- nected through a network. The streaming format has increas ingly become more papular because of the asilability of tame tiles on demand the ability to execute more complex ames, the ability 10 network between players for mult player gaming, sharing of assts between players, sharing of| Instant experiences between players andor ‘spectators, allowing friends to watcha frend play 2 video game, having ‘Fiend joi the on-going game play ofa Trend, ad the like The elood gaming server may be configured t provide resources one or more clients andlor applications, Tati, the cloud gaming server may be configired with resources capable of high throughput. For example, there ar limits 9 the performance that an individva graphs processing unit (GPU) can atain. To rendor even more complex seenes oF se even more complex algorithms (et. materials, Hihting, fe) when generating a seene, it may be desirable 10 use ‘multiple GPUs to render a single image. However, usage of those. graphics processing ‘units equally is dificult 10 achieve. Further, even though there are multiple GPUs 10 process an image for an application sing tational th 2 nologies, there is not the ability 10 support a corresponding ‘increase in both sercea pixel count and density of geometry (eg, four OPUs cannot write four times the pixels andor process four imes the vertices or primitives for an image). It isin this context that embodiments of the disclosare SUMMARY Embodiments of the present disclosure relate 1 using rltiple GPUs in collaboration to render «ingle image, such as mult-GPU rendering of geometry foram application by protesting agsinstsersen regions, which may be inter Jeaved, helore rendering Embodiments ofthe present disclosure disclose a method for graphies processing. The method includes rendering traphies for an application using a plurality of graphics processing units (GPUs). The method includes dividing responsibilty for rendering of weometry of the graphics between the plraity of GPUs besed ona plurality of sreen regions, cach GPU having a corresponding division of the responsibility which is known to the plraliy of GPUs. Screen regions are interleaved, The method inchades assign ing plurity of pieces of geometry of an image frame to the plurality of GPUs for geometry testing, The method Includes assigning a GPU a piece of geometry ofan image fame generated by an application for geometry testing. The method includes performing geometry testing at the GPU 9 generate information regarding the piece of geometry and its felation to each of the phiraity of screen regions, ‘The method includes rendering the piece of geometry using the information at each of the plurality of GPUs, where vs the information can include, for example, skipping rendering entirely if thas boen determines that the pies of gometry does not overlap any sereen region assigned oa given GPU, In another embodiment a non-trasitory compsterread- able medium for performing a method is disclosed. The computer-teadable medium including program instroctions for rendering praphies for an application using plurality of raphies processing units (GPUs). The compurer-eadable ‘medium including program insirctions for dividing espon- Sibiiy for the rendering of geometry of the graphics between the plrality of GPUs based ona plurality of sreen regions, each GPU having a corresponding division af the responsibilty which is Known to the plurality of GPUs, wherein screen regions inthe plurality of sreen regions are Jnerleaved. The computer realable median) including pro- tram instructions for assigning a GPU a pieee of geometry fan image frame generated BY an application for geometsy protesting. The computer readable medium including pro: fram instructions for performing the geometry protesting st the GPU to generate infomation regarding the picoe of geometry and its oltion to each ofthe plurality af sreen regions. The computer readable medium inchuding program Instructions for using the infomation at each ofthe plurality of GPUs when reodering the image frame In sill another embodiment, 2 computer system is dis closed including « processor and memory coupled tothe processor and having stored therein instrctions that, if executed by the computer system, cause the compute sy tem to execute a method for graphies processing. The method including rendering. praphies for an application using a plurality’ of graphics processing nits (GPUs). The ‘method including dividing responsibility forthe rendering of fgoometry of the graphics between the plurity of GPUs based on a plurality of screen regions, cach GPU having a corresponding division of the responsibilty whichis known US 11,605,148 B2 3 ‘© the pluraiy of GPUs, wherein sereen regions ia the plurality of serven regions are interleaved. The method including assigning 9 GPU a piece of geometry ofan image fame generated by an application for eometry pretesting. ‘The method including performing the geometry prtesting st the GPU to generate information regarding the piece of geometry and its relation to each ofthe plurality of screen ‘regions. The method incloding using the information at each ff the plurality of GPUs when rendering the image fame. Embodiments of the present disclosure disclose a method {oe graphics processing. The method includes rendering triphics for an application using a plumiity of graphics processing units (GPUs). The method inclides dividing esponsibility for the rendering of geometry of the graphics between the plurality of GPUs based ona plurality of sreen regions, each GPU having a corresponding division of the responsibilty whichis Known tothe plurality of GPUs. The method includes performing geometry testing ata pretest GPU ona plurality of pieces of peometry of an imae frame encrated by an application to generate information regard- 2 Ing each picce of geometry and its relation to each of the plurality of sereen regions. The method includes rendering the plurality of pieces of geometry at each ofthe plurality of| GPUs using the information generated Tor each of the pluality of pieces of yeomety, where using the information Includes, for example, skipping rendering entirely if it has ben determined tht the piece of geometry does abt overlap ny screen region assigned toa given OPU, Tn another embodiment, « non-traasitory computerrea- able medivin for performing # method i= disclosed, The ‘computerteadable medium iachiding program instructions {or rendering graphies for an application using a plurality of| traphies processing units (GPUs). The computer-eadable ‘medium including program instroctions for dividing respon: ity for the rendering of geometry of the graphics between te plrality of GPUs based ona plurality of sercen regions, each GPU having a corresponding division of the responsibilty whichis known tothe plurality of GPUS. The computerrelable medium aching program insteetons or performing geometry testing at a pretest GPU on a plurality of pieces of geometry of un image frame penerated by an application fo generate information regarding each piece of geometry and its relation to each ofthe plurality of fereen regions. The computerreadable medium including program instictions for rendering the plurality of pisces of| eometry at each of the plurality of GPUs using the infor ‘mation generated for each of the plurality of pivcos of eometry, where using the information includes, for ‘example, skipping rendering entirely if i has been deee- mined that the piece of geometry does not overlap any screen region assigned to a given GPU. In slll another embodiment, a computer systema is dis closed including a processor and memory coupled 10 the processor and Tiving stored therein insirictions that, if executed by the computer system, cause the computer s95- fem to execute a method for graphies processing, The method including rendering. graphies for an application using a plurality of graphics processing units (GPUS). The tcthod includes dividing responsibility forthe rendering of| geometry of the graphics between the plurality of GPUs based on a plurality of screen regions, each GPU having a corresponding division of the responsibility whichis known ‘o the plurality of GPUs. The method includes peroeming eometry testing ala protest GPU on a plurality of pieces of fzonmetry of an image frame generated by an application to {senerate information regarding each piece of geometry and its relation to each of the plurality of sereen regions. The 4 method includes rendering the plurality of pieces of geont- cy a each ofthe plurality of GPUs using the information enerated for each of the plurality of pieces of peomety, ‘where using the information includes, fr example, skipping rendering entirely fit has been determined tht the piece of geometry does not averlap any screen region assighcd to a given GPU, Embosments ofthe present disclosure diselose 8 method for graphics processing. The method includes rendering tphies for an application using a plurality of graphics processing units (GPUs). The method includes dividing ‘esponsibility for the rendering of zeometry of the praphics between the plrality of GPUs based ona plurality of sreen regions, each GPU having a corresponding division of the responsibilty which is known tothe plurality of GPUs. The Iethod inchides rendering a first plurality of pieces of eometry a the plurality of GPUs during a rendering phase fa previous imape frame generated by an application. The ‘method inches penerating statisti forthe rendering ofthe previous image frame. The method includes assigning based fn the statistics # socond plurality of picees of geometry of 4 current image frame generated by the application tothe Plrality of GPUs for geometry testing, The method ineludes performing geometry testing ata current image fame on the Second plurality of pieces of peometry to penerat informa tion rparding each piece ofthe second plurality of pieces of goometry and its elation to each ofthe plurality of screen regions, wherein the geometry testing is performed at each ofthe plurality of GPUs based on the assigning, The method ‘nchides rendering the second plrality of pisces of goon ety at each ofthe plurity of GPUs using the information trenerted for each of the second plurality of pieces of peometry, where using the information can include, for example, skipping rendering entirely if i has been deter mined that the pice of geometry does not overlap any sercen region assigned to a given GPU, In another embodiment, 8 non-transitory computer-rea- able medium for performing method fs disclosed. The omputerreadable medium including program instructions To rendering prophies for an application using plurality of niphics processing units (GPUs). The computer-eadable ‘medium including program insnictions for dividing respon: sibility for the rendering of geometry of the graphics between the plurality of GPUs based ona plurality of sreen regions, each GPU having a corresponding division of the responsibilty which is knosn tothe plurality of GPUs, The computer readable medium ineluding program insteetions for rendering a fist plurality of pigees of geometry at the plurality of GPUs during a rendering phase of a previous ‘mage frame generated by an application. The computer: readable medium including program instnctions for gener Sing statins forthe renderingof the previows image lame. ‘The computer-readable medium including program instruc- tions for assigning ased on the statsies a second plurality of pieces of geomet ofa current image frame generated by theapplcation tothe plurality of GPUs for geometry testi The computer readable mesium including program insiie tions for performing geometry testing at ® current image fame on the second plurality of pisces of geometry 19 generate information reganding each piece of the second Plralityof pieces of geometry and its relation to each ofthe luality of screen reyions, wherein the geometry testing is performed at each of the plurality of GPUs based om the assigning. The computer-readable medium including pro: fim insirctions for rendering the second plurality of Picces of geometry at each ofthe plarality of GPUs using the Information generated for each of the second plurality of US 11,605,148 B2 5 pieces of geometry, where using the information can Include, for example, skipping rendering ently i thas been determined that the piece of poometry does not overlap ny screen region assigned toa given OPU, Tn sill another embodiment, a computer system ie dis closed including « processor and memory coupled t0 the processor and having sored therein instrctions that, executed by the computer system, case the computer sy5- fem to execute a method for graphies processing. The ‘method includes rendering grophies for an application using plurality of graphics processing units (GPUs). The method includes dividing esponsibility forthe rendering of geom- ‘try of the graphics heswcon the plurality of GPUs hased on plurality of screen regions, each GPU having a eomrespond- ing division of the responsibility which is known to the plurality of GPUs. The method includes rendering a fist plurality of pieces of geometry atthe plurality of GPUs during a rendering phase ofa previous image frame genes ‘ted by an application. The method inchides genersting ‘attics forthe rendering ofthe previous image frame. The tcthod includes assigning based onthe waists a sosond plurality of pisces of geometry of a current image frame teneated by the application t the plurality of GPUs for eometry testing, The method includes performing geoa- (y testing aa curent image frame onthe sevond plurality of pieces of geometry to generate information regarding cach piece ofthe second plurality of pieces of geometry and its elation to each othe plurality of screen regions, wherein the geometry testing is performed at each ofthe plurality of| GPUs based on the assigning. The method incies render ing the second plurality of pieces of geometry at each ofthe pluality of GPUs using the information generated for each Of the second plurality of pieces of peometry, where using the infomnation can incl, for example, skipping rendering ctirely fit has been determina that the pieve of seometry ‘doesnot overlap any sreen region assigned io given GPU [Embodiments of the present disclosure disclose a method. {or graphies processing. The method includes rendering triphies for sm application using 2 plurality of graphics processing units (GPUs). The method includes dividing responsibilty forthe rendering of geometry of the graphies between te plurality of GPUs based ona plurality of serven regions, cach GPU having a corresponding division of the responsibilty whichis known tothe plurality of GPUs. The ‘method inchides assigning a plurality of pies of gsometty of an image frame to the plurality of GPUs for geometry testing. The method inehudes sting fest sate configuring ‘one or more shades to perform the geometry testing. The sthod includes performing geometry testing at the plurality lf GPUs on the plurality of pieces of geometry to generate {information regarding each poce of geometry and ts rel foto each ofthe plurality of screen regions, The method Includes sotng a second state configuring the one or more shaders to perform rendering. The method includes render- ing the plurality of pieces of geometry at each of the phirality of GPUs using the information generated foreach Df the plurality of pieces of geometry, where using the information includes, for example, skipping. rendering entirely ifit has been determined that the plece of geometry ‘doesnot overlap any sreen region assigned toa given GPU. Th another embodiment, @ non-transitory compter-read- able medium for performing a method is disclosed, The ‘omputerteadable medium including program insievedons {or rendering graphs for an application using & plurality of mphics processing units (GPUs). The computer-eadable ‘ncdum including program instrsctions for dividing respon sibility for the rendering of geometry of the graphics 6 berwoon the plurality of GPUS based ona plurality of sereen regions, each GPU having a corresponding division of the responsibility whichis known tothe plurality of GPUs, The computerreadable medium including program instrctions fr assigning a plurality of picces of geometry of an image frame to the plurality of GPUs for geometry testing. The computer-eadable medium including program instrctions for seting a fst state configuring one or more shaders 10 perform the geometry testing. The computer-readable medium including program instructions for performing eomety testing atthe plurality of GPUs on the plurality of Pieces of geometry 10 kenerte information raring each Pisce of geometry and its elation to each of the plurality of | fercen regions, The computerreodable medium including program insirictons for seting a second state configuring the one or more shaders 10 perform rendering. The com puterseadable medium including program instructions for rendering the plurality of pieces of geometry at each of the plurality of GDUs using the information generated foreach of the plurality of pieces of geometry, where using the {information includes, for example, skipping. rendering entirely fit has boen determined tha the pee of geometry does not overlap any screen rion assigned oa given GPU, Tn sill another embodiment, a computer system is dis: closed including a processor and memory coupled 10 the processor and having stored therein instractons that, if executed by the computer system, cause the computer sys- tem to execute a method for graphics processing. The ‘method includes rendering graphics for an application ws plurality of graphies processing units (GPUS). The motel Includes dividing responsibility for the rendering of geon- cy of the praphies between the plurality of GPUs base on plurality of seen regions, each GPU having a eorrespond- ing division of the responsibility which is known to the pluraly of GPUs. The method includes assigning «plurality Df pieces of geometry ofan image frame to the plarlity of | GPUS for geometry testing. The method includes setting a first state configuring one or more shaders to perform the eometry testing, The method includes performing geom- ‘ey testing atthe plurality of GPUs on the phality of pieces bf geometry to generate information regring each piece of toometry snd is elation to each ofthe plurality of sreen regions. The method includes seting a second state config~ tring the one or more shaders to perform rendering. The hethod includes rendering the plurality of pisces of geom ey a each ofthe plurality of GPUs using the information enerated for each of the plurality of pigees of geometry, ‘where using the information includes, fr example, skinning rendering entirely it has been determined thatthe piece of eometry docs not overlap any sereen region assigned to a iven GPU, "Embodiments ofthe present disclosure disclose a method for graphics processing. The method includes rendering trophies for an application using a plurality of graphics processing units (GPUs). The method includes divi responsibilty for the rendering of geometry of the graph between the plraityof GPUs based ona pluraiy of sreen regions, cach GPU having a corresponding division of the responsibilty which is known tothe plurality of GPUs, The rethod includes assigning a plurality of pees of geometry fof an image frame to the plurality of GPUs for geometty testing. The method inehides interleaving a first set of slurs to peeform geometry testing and rendering om a fst set of pieces of geometry with a second set of shaders 10 perform scometry testing and rendering on a second set of | Pisces of goomeiry. The geometry testing generates corte- ‘ponding information regarding each pioce of geometry in US 11,605,148 B2 7 the fist set oF socond set and is relation to exch of the plurality of screen egions. The corresponding information fs used by the pluraty of GPUs to render exch pice of eometry in fis set oF sevond se, where using the infor: ‘mation ineldes, for example, skipping rendering entirely if ‘thas boon determined thatthe ploe of geometry does aot overlap any sereen region assigned to a piven GPU. In ancther embodiment, a non-ansitory computerzead- able medium for perfoming a method is disclosed. The Compiterrendable medium incliding program instmctions {or rendering saphis for an application using & plurality of mphies processing units (GPUs). The computer readable ‘medium including program instructions for dividing respon: sibility for the rendering of geometry of he graphics between the plraity of GPUs based ona plurality of ereen regions, each GPU having 2 corresponding division of the responsibility whichis knowa tothe plurality of GPUs. The ‘ompitrteadable medium including program instroetions for assigning 2 plurality of pieces of poomey of an image fame w the plurality of GPUs for geometry testing. The computerreadable medium including program iastnctions fo intrlaving a fst sot of shaders to perform geometry testingand rendering oma fist set of pieces of goomicry with a second set of shaders to perform geometry testing and fendering on a second set of pisces of goometry. The geometry testing generates comesponding infor ‘egading each poe of geometry in the fist et or second set and it elation fo each ofthe plurality of sereen regions. The ‘corresponding information is used by the plurality of GPUs to render each piece of goomety i first Set or second Se, ‘where using the information inludes, fr example, skipping rendering entirely it has been determined that the piece oF * geometry does not overlap any’ sereen region assigned to a wiven GPU. nail another embodiment, a computer system is diss closed inciding « processor and memory coupled 10 the processor and having stored therein insirictions that, if executed by the computer system, cause the computer $9 fem to execute a method for graphies processing. The ‘method includes rendering graphics for an aplication using plurality of graphics processing units (GPUS). The method Includes dividing responsiblity lor the rendering of geom- ‘ty ofthe graphics berween the plurality of GPUs based on ‘plurality of sercen regions, each GPU having comrespond- ing division of the responsiblity which is kaown to the phuality of GPUs. The method includes assigning a plurality Dfpieces of geometry of an image frame to the plurality of GPUs for goometry testing. The meta’ inclidesinterlav- ing a fist se of shaders to perform geometry testing and rendering om a first set of piotes of geometry with» second Sf shaders to perform geometry testing and rendering on ‘second set of pices of goomety. The geometry testing tenerates corresponding information reganling each piece ‘of geometry in the fist sel or second set and its relation 10 toch of the plunity of screen regions. The corresponding {information Is used by the plurality of GPUs to render each picce of goomtny in first set or sovond st, where wsing the Information includes, for example, skipping rendering entirely fit has been determined that the plece of geometry ‘docs not overlap any screen region assigned to given GPU. ‘Other aspects of the disclosure will become apparent fom the following detailed description, taken in conjunction with the aecompanying drawings, illustrating by way of example the principfos of the disclosure [BRIEF DESCRIPTION OF THE DRAWINGS ‘The disclosure may best be understood by reference to the following description taken ia conjunction with the accom ppanying drawings in which: 8 FIG. 1isadiagram ofa system for providing gaming over 8 nctwork between one or more clot gaming servers configured for implementing multiple GPUs in collabora to render a single image, including mult-GPU (raph processing nit) rendering of geometry foran application by Protesting the geometry against screen regions, which may be interleaved, in accondance with embodiments of the present disclosure FIG. 2 isa diagram of a multi-GPU architecture wherein ‘aultiple GPUs collaborate to render a single image, in scoardance with one embodiment ofthe present dislosure FIG. 3 is diagram of multiple graphs processing unit resources configured for mult-GPU rendering of geometry ‘or an application by protesting the geometry against screen regions, which may’ be interleaved, in accordance with one embodiment ofthe present disclosure, FIG, 4 is a diagram of a rendering architecture imple senting a grphics pipeline that is configured for mul-CPU processing scl that multiple GPUs collaborate to render a Single image, in seeordance with one embodiment of the present disclosure FIG. Siva flow diagram illustrating a method for graphics processing including molti

    You might also like