CA2546427A1 - Method and system for multiple 3-d graphic pipeline over a pc bus - Google Patents
Method and system for multiple 3-d graphic pipeline over a pc bus Download PDFInfo
- Publication number
- CA2546427A1 CA2546427A1 CA002546427A CA2546427A CA2546427A1 CA 2546427 A1 CA2546427 A1 CA 2546427A1 CA 002546427 A CA002546427 A CA 002546427A CA 2546427 A CA2546427 A CA 2546427A CA 2546427 A1 CA2546427 A1 CA 2546427A1
- Authority
- CA
- Canada
- Prior art keywords
- graphic
- gpus
- data
- hub
- gpu
- 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.)
- Abandoned
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
- G06F9/505—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering the load
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/14—Digital output to display device ; Cooperation and interconnection of the display device with other functional units
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/20—Processor architectures; Processor configuration, e.g. pipelining
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/005—General purpose rendering architectures
-
- G—PHYSICS
- G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
- G09G—ARRANGEMENTS OR CIRCUITS FOR CONTROL OF INDICATING DEVICES USING STATIC MEANS TO PRESENT VARIABLE INFORMATION
- G09G5/00—Control arrangements or circuits for visual indicators common to cathode-ray tube indicators and other visual indicators
- G09G5/36—Control arrangements or circuits for visual indicators common to cathode-ray tube indicators and other visual indicators characterised by the display of a graphic pattern, e.g. using an all-points-addressable [APA] memory
- G09G5/363—Graphics controllers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/14—Digital output to display device ; Cooperation and interconnection of the display device with other functional units
- G06F3/1423—Digital output to display device ; Cooperation and interconnection of the display device with other functional units controlling a plurality of local displays, e.g. CRT and flat panel display
- G06F3/1438—Digital output to display device ; Cooperation and interconnection of the display device with other functional units controlling a plurality of local displays, e.g. CRT and flat panel display using more than one graphics controller
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2200/00—Indexing scheme for image data processing or generation, in general
- G06T2200/28—Indexing scheme for image data processing or generation, in general involving image processing hardware
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Computer Hardware Design (AREA)
- Human Computer Interaction (AREA)
- Image Generation (AREA)
- Image Processing (AREA)
Abstract
Method and system for improving the parallelization of image processing, using one or more parallelization modes, wherein the4 image that is displayed on at least one computer screen by one or more Graphics Processing Units (GPUs) (130). Software applications (121) are provided for issuing graphics command and graphics libraries (122) are provided for storing data used to implement the graphics commands. A Software Hub Drivers (123) is provided for controlling a Hardware Hub (110), for performing real-time analysis of a data stream, from which frames of the image are generated, for determining the parallelization mode of each GPU, and for forwarding the data stream or a portion thereof to each GPU. GPU Drivers (124) are provided for allowing the GPUs to interact with the graphics libraries and an I/O module (160) is provided for interconnecting between the Software module and the Hardware Hub.
Description
NLETHGD AND SYSTE1VI F~R MULTIPLE 3-D GRAPHIC PIPELII~TE
Q'~7EI~, A PC BUS
Field of the Invention The present invention relates to a method and system for 3-D (three-dimensional) multiple graphic processing. More particularly, the invention relates to a method and system for improving the parallelization of image processing by Graphic Processing. Units (CPUs), based on unified framework of three parallelization methods, which are time division, image division and object division methods.
Definitions, Acronyms and Abbreviations Throughout this specification, the following definitions are employed:
GPLT: GPU (Graphic Processing Unit) like the CPU (Cents al Processing Unit), a GPU is a single-chip processor which is used primarily for computing 3-D
functions. This includes tasks such as, lighting effects, object transformations, and 3-D motion. These are mathematically-intensive tasks, which other wise, would put quite a strain on the CPU, but since the specialized GPU can handle most of the 3-D calculations, it helps the computer to perform these tasks more efficiently, and, of tour se, faster .
Polygon: Polygons in 3-D graphics are two-dimensional primitives, allowing generating and displaying of 3-D complex graphical objects. The polygons are the faces of the object and are composed from N vertices. Actually, a polygon is a closed plane figu r e, bounded by tbxee or mor a line segments.
Frame Buffer: a Fran2e Buffer (FB) is a buffer that stores the contents of an image, pixel by pixel. Generally, the portion of memory is reserved for holding the complete bit-mapped image that is sent to the monitor, for display.
Typically the frame buffer is stored in the memory chips on the video adapter.
In some instances; however, the video chipset is integrated into the motherboard design, and the frame buffer is stored in the general main memory.
Obiect compositin~nit: performs re-composition of multiple three-dimensional rasters into final image. The merged data is resolved for the closest pixel to the viewer in 3-D space, based on the depth value of pixels.
The new method, based on autonomous associative decision, allows the use of multiple GPUs for any fr ame complexity.
Display List: a Display List is a description of the 3-D scene through a list of gr aphic pr imitives, such as polygons and attributes. The display list provides inter mediate image storage for quick image r etrieval.
Vertex array: a Vertex Array is an array of vertices describing the 3-D scene.
A Vertex Array provides intermediate image storage for quick image retrieval.
Alpha blending Alpha blending controls the way in which the graphic infor mation is displayed, such as levels of tr ansparency, or opacity.
Backer ound of the Invention The three dimensional graphic pipeline architecture breaks-down into segmented stages of CPU, Bus, GPU vertex processing and GPU fiagment (pixel) processing. A given pipeline is only as strong as the weakest link of one of the above stages, thus the main bottleneck determines the overall throughput. Enhancing performance is all that required for reducing or eliminating bottlenecks. The major bottleneck strongly depends on the application. Extreme cases are CA:D-Like (Computer Aided Design) applications, char acterized by an abundance of polygons (vertices), vs. video-game applications having a small polygon count but intensive fragment activity (e.g., texturing). The first class suffers from vertex processing bottlenecks, while the second class suffers from fr agment bottlenecks. Both ar a frequently jammed over the PC bus. Many applications have mixed characteristics, where bottlenecks may randomly alternate between extremes, on a single frame basis.
The only way to improve the performance of the GPU is by means of parallelizing multiple GPUs according to one of the bottleneck solving methods. There are two pr edominant methods for rendering gr aphic data with multiple GPUs. These methods include time division (time domain composition), in which each GPU renders the next successive frame, and image division (screen space composition), in which each GPU renders a subset of the pixels of each flame. The third one, much less popular, is the object division (polygon decomposition) method.
In the time division method each GPU render s the next successive frame. It has the disadvantage of having each GPU render an entire frame. Thus, the speed at which each frame is r endered is limited to the rendering rate of a single GPU. While multiple GPUs enable a higher fr ame rate, a delay can be imparted in the response time (latency) of the system to a user's input. This occur s because, while at any given time, only one GPU is engaged in displaying a rendered frame, each of the GPUs is in the process of rendering one of a series of frames in a sequence. To maintain the high frame rate, the system delays the user's input until the specific GPU, which fir st received the signal cycles through the sequence, is again engaged in displaying its rendered frame.
In practical applications, this condition serves to limit the number of GPUs that are used in a system. With large data sets, then a is another bottleneck, due to the fact that each GPU must be able to access all the data. This requires -d-either maintaining multiple copy operations of lar ge data sets or possible conflicts in accessing the single copy operation.
Image division method splits the screen between N GPUs, such that each one displays 1/N of the image. The entir a polygon set is transfers ed to each GPU
for processing, however, the pixel processing is significantly reduced to the vvindoi~~ size. Image division has no latency issues, but it has a similar bottleneck with large data sets, since each GPU must examine the entire database to determine which graphic elements fall within the portion of the scr een allocated to said GPU. Image division method suits applications with intensive pixel processing.
Object division method is based on distribution of data subsets between multiple GPUs. The data subsets are rendered in the GPU pipeline, and converted to Frame Buffer (FB) of fragments (sub-image pixels). The multiple FB's sub-images have to be merged (composited) to generate the final image to be displayed. Object division delivers parallel rendering on the level of a single frame of very complex data consisting of large amount of polygons. The input data is decomposed in the polygon level and re-composed in the pixel level. A
proprietary driver intelligently distributes data streams, which are generated ..
by the application, between aII GPUs. The rasters, generated by the GPUs, are composited into final raster, and moved to the display. The object division method well suits applications that need to render a vast amount of geometr ical data. Typically, these are CAD, Digital Content Cr eation, and comparable visual simulation applications, considered as "viewers," meaning that the data has been pre-designed such that their three-dimensional positions in space are not under the interactive control of the user. However, the user does have interactive control over the viewer's position, the direction of view, and the scale of the graphic data. The user also may have control over the selection of a subset of the data and the method by which it is rendered.
This includes manipulating the effects of image lighting, color ation, transparency and other visual characteristics of the underlying data.
In above applications, the data tends to be very complex, as it usually consists of massive amount of geometrical entities at the display List or vertex array.
Therefore, the construction time of a single frame tends to be very long (e.g., typically 0.5 sec for 20 million polygons), which in turn slows down the overall system performance.
Therefore, then a is a need to provide a system which can guarantee the best system performance, being exposed to high traffic over the PC (Personal Computer) Bus.
It is an object of the present invention to provide an amplified strength of the GPU by means of parallelizing multiple GPUs.
It is another object of the present invention to provide a system, wherein the construction time of a single frame does not slow down the overall system response.
It is still another object of the present invention to provide a system and method, wherein the graphic pipeline bottlenecks of vertex processing and fragment processing are transparently and intelligently resolved.
It is still a fur then obj ect of the present invention to provide a system and method that has high scalability and unlimited scene complexity.
Tt is still a further object of the present invention to provide a process overcoming difficulties that are imposed by the data decomposition, which is partition of data and graphic commands between GPUs.
It is still a further object of the present invention to provide a method and s~Tstem for an intelligent decomposition of data and graphic commands, preserving the basic featur es of graphic libr aries as state machines and complying v~ith graphic standards.
Other objects and advantages of the invention will become apparent as the description proceeds.
Summary of the Invention '~'he present invention is directed to a system for improving the parallelization of image pr ocessing, using one or more par allelization modes, wherein the image that is displayed on at Ieast one computer screen by one or more Graphic Processing Units, which comprises:
one or more software applications, for issuing gr aphic commands;
one or more graphic libraries, for storing data used to implement the graphic commands;
one or more Software Hub Drivers, for controlling a Hardware Hub, for interacting with the operation system of the computer and the graphic libraries, for performing r eal-tine analysis of a data stream, from which frames of the image ar a generated, for deter mining the parallelization mode of each GPU, and for forwarding the data stream or a portion thereof to each GPU;
one or more GPU Drivers, for allowing the GPUs to interact with the graphic Iibr ar ies; and at least one I/O module for interconnecting between the Software module and the Hardware Hub, uTherein, the Hardware Hub distributes between the GPUs, for each frame, graphic commands and the data stream or a portion thereof, according to their relative complexity within the image, anal defines the complexity The Softv~Tare _7_ Hub Driver also composites a gr aphics output for display, using the outputs obtained from at least one GPU, while alternating, whenever required, the parallelization mode for the each fr ame.
Parallelization is based on an object division mode or on an image division mode or on a time division mode or on any combination thereof. The hardware hub comprises a compositing unit for composing a complete frame from processed portions of the data str cam. The hardware hub comprises a hub router for r outing polygonal data, for r outing gr aphic command stream, for routing pixel data and for routing the results of composition, while operating in the object division mode or in the image division mode or in the time division mode or in any combination thereof. The hardware hub comprises a control unit for receiving commands from the Software Hub Driver within the I/0 module. The hardware hub comprises a memory unit for storing intermediate processing results of one or more GPUs and data required for composition and transferring the processed data for display.
Preferably, the Software Hub Driver is capable of performing the following operations:
interception of the graphic commands from the standar d graphic library by means of the OS interface and utilities;
forwarding and creating graphic commands to the GPU Driver by means of the OS interface and utilities;
controlling the Hardware Hub, registry and installation operations by means of the OS interface and utilities:
maintaining the consistency of graphic machine states across the GPUs, based on the input graphic commands stream, while using state monitoring;
estimating the type of graphic load and overload in the executed application graphic context, while using application and gr aphic resources analysis;
_g_ load estimation of the GPUs load balance based on graphic commands stream and time measurements, while using application and gr aphic resources analysis;
adjusting the load distribution between GPUs according to feedback received from each GPU regarding the Ioad balance, while using application and graphic resources analysis;
performing manipulation in graphic functions accor ding to the current parallelization mode; and controlling the distributed graphic functions, while modifying the graphic commands and the data stream according to the current parallelization mode.
The present invention is directed to a method for improving the parallelization of image processing, using one or mare parallelization modes, wherein the image that is displayed on at least one computer screen by one or more Gr aphic Processing Units. Softwar a applications are provided for issuing gr aphic command and graphic libraries are provided for storing data used to implement the graphic commands. A Software Hub Drivers is provided for controlling a Hardware Hub, for interacting with the operation system of the computer and the graphic libraries, for performing real-time analysis of a data stream, from which frames of the image are generated, for determining the parallelization mode of each GPU, and for for war ding the data str eam or a portion thereof to each GPU. CPU Drivers ara provided for allowing the GPUs to interact with the 7.
graphic libraries and an I/0 module is provided for interconnecting between the Software module and the Har dware Hub. graphic commands and the data stream ar a portion thereof are distributed between the GPUs for each frame by the Hardware Hub, according to their relative complexity within the image, wherein the complexity is defined by the Software Hub Driver. The Software Hub Driver also composites a graphics output for display, using the outputs _g_ obtained from at least one GPU, while alternating, whenever required, the par~Ilelization made for the each fr ame.
Whenever the parallelization mode is an Object di~nsion parallelization mode, the following steps are performed;
for each frame, generating a stream of graphic operations and polygonal data;
marking the polygonal data and gr aphic commands by means of the Software Hub Driver for distribution between multiple GPUs;
sending the marked data to the Hardware Hub;
distributing the marked data via the Hub Routes to the multiple GPUs;
rendering the data by means of GPUs;
retrieving the data fr om the Frame Buffer s and forwarding the r etrieved data to the compositing unit via the Hub Routes;
compositing the content of the Frame Buffers into a single Frame Buffer; and forwarding the content of the single Frame Buffer to at least one designated GPU for display.
Wl~.enever the par allelization mode ~is an Image di~nsion parallelization mode, the following steps are performed:
sub dividing the screen to portions and assigning differ ent viewpo.r is to GPUs by means of the Software Hub Driver;
moving the entire polygonal data and graphic commands to the Hub Routes;
tr ansmitting the entir a polygonal data and graphic commands to GPUs, wherein each GPU receives the same data;
rendering the data by means of GPUs;
forwarding a portion of the content stored in the Frame Buffers to compositing unit in Hardwar a Hub for the complete image creation; and -IO-forwarding the image to at least one designated GPU for display.
Whenever the parallelization mode is a Time division parallelization mode, the following steps are performed:
forwarding to each one of the multiple GPUs the entire amount of polygons for r endering;
redirecting the entire polygonal data and graphic commands by means of Software Hub Driver to all GPUs, while alternating between them;
rendering the data by means of GPUs;
transferring rendered data from at least one GPU via the Hub Router; and r edirecting the resulting content of the Frame Buffer via Hub Router to at least one designated GPU for display.
The distribution of polygons between multiple GPUs is performed by distributing blocks of data between multiple GPUs and by testing each graphic operation for blocking mode, in which one or more parallelization modes are carried out, thereafter. The data is redirected in regular non-blocking path to at least one designated GPU, This process is repeated until a blocking operation is detected. Then GPUs are synchronized by performing a flush operation in order to terminate rendering and clean up the internal pipeline in each GPU;
performing a composition open ation for merging the contents of the Frame Buffers into a single Frame Buffer and byte ansmitting the single Frame Buffer back to all GPUs. Then the composited complete frame is terminated at all .
GPUs, except one or mor a designated GPUs, whenever a Swap operation is detected and displaying the image by means of the one or more designated GPUs. the same data is processed by all GPUs, as long as the blocking mode is active and the Swap operation is not detected. whenever the blocking mode is inactive, the designated data is further processed by multiple GPUs,.
Brief Description of the Drawings In the dr awings:
Fig. 1 is a block diagram of a multiple GPU architecture system, according to an embodiment of the pr esent invention;
Fig. 2 is a block diagram of Hardwar a Hub components, actor ding to an embodiment of the pr went invention;
Fig. 3 is a block diagram of Object division parallelization mode, according to an embodiment of the present invention;
Fig. 4 is a block diagram of Image division parallelization mode, actor ding to an embodiment of the present invention;
Fig. 5 is a block diagram of Time division parallelization mode, according to an embodiment of the pr esent invention;
- . Fig. 6 is a schematic block diagram of a possible integration of the Software Hub Driver into the operating system environment, according to an embodiment of the pr esent invention;
- Fig. 'l is a functional block diagr am presenting the main tasks of the Software Hub Driver, according to an embodiment of the present invention;
- Fig. 8 is a flow chart presenting an process for distribution of the polygons between the multiple GPUs, according to an embodiment of the pr esent invention; and Fig. 9 discloses a sample configuration of the system, employing 8 GPUs, according to an embodiment of the present invention.
Tt will be appreciated that for simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn to scale. For example, the dimensions of some of the elements may be exaggerated relative to other elements for clarity. Further, where considered appropriate, reference _ 12_ numerals may be repeated between the figures to indicate corresponding or analogous elements.
Detailed Description of the Preferred Embodiments The Multiple 3-D Graphic Pipeline The current invention calls for the introduction of an extended PC graphic architecture including novel operational component, the 3-D pipeline Hub.
Fig. 1 presents multiple GPU architecture system I00, according to an embodiment of the present invention. The hub 110 is located in a unique position on the PC bus, between the CPU (Central Processing Unit) and a cluster of GPUs 130. The multiple graphic pipeline architecture, as described in Fig. I, consists of Software Application 121, Graphic Library 122, Software Hub Driver 123, GPU Driver 124, Hardware Hub 110, cluster of GPUs 130, and displays) 140. Usually, one of the GPUs is designated as a display unit.
It should be noted, that it is possible to have more than one display unit, or include a display unit directly inside the Hardware Hub. A display unit can drive multiple screens, as well.
The Hub mechanism consists of a Hardware Hub component I10, located on the PC bus between I/O (Input/output) chipset 160, being a I/O module, and a cluster of CrPUs 130, and a Software module comprising Software Hub Driver 123, executed by the PC.
The Hardware Hub I10 carries on at least the following action:
- distributes decomposed polygon stream and graphic commands between GPUs;
- composites graphics output for display according to different parallel modes;
- makes cache of data; and - alternates modes of parallelism.
The Software Hub Driver 123, besides controlling the Hardware Hub I10, also carries on at least the following actions:
- interacts with the OS (Operation System) and graphic library, such as OpenGL, DirectX;
- performs real-time analysis of the data stieam;
- determines the parallelization mode; and decomposes the data and command stream.
One advantage of this method is that the unique location of the Har dware Hub 110 allows it to control the graphic pipeline, while being tr ansparent to the application 121. The application I21, along with Graphic Library 122, such as OpenGL, keeps working as it was a single GPU.
Another advantage of this method is that the unique location of the Hardware Hub 110 allows it to control a graphic pipeline between the User Interface 150 and Display 140, while being transparent to the GPU. Each GPU of the cluster keeps working as if it is the only graphic processor hooked on the I/O chipset 160.
Still another advantage of this method is that the unique location of the Hardware Hub 110 allows it to control the graphic pipeline for different parallelization modes: image division mode, time division mode or object division mode.
Further advantage of this method is that the unique location of the Hardware Hub 110 allows it to sense in real-time the varying parameters of application's load, such as polygon count, texture volume, human interaction, and to intelligently decide and carry on the current optimal parallelization method.
It should be noted that according to all embodiments of the pr esent invention, the displays) 140 may be connected directly or indirectly to the Hardware Hub 11o.
Reference is now made to Fig. 2, which discloses the components of the Hardv~are Hub I10, according to an embodiment the present invention.
Control Unit 210, accepts proprietary commands from the Software Hub Driver over the bus, and accordingly controls the other units. Hub Roister 230 routs polygonal data and graphic command stream from left to right, routs pixel data from right to compositing unit, and routs compositing results to the right. Compositing Unit 240 performs various compasiting schemes according to pai allelization mode. Auxiliary Memory block 220 is used for storing internzediate processing results of one or more GPUs, data required for composition and transferring the processed data for display.
The Hardware Hub 110 utilizes its units according to parallelization modes: an Object division mode, an Image division mode, a Time division mode. These modes are adaptively handled by the Hardware Hub 110 according to application needs.
Fig. 3 discloses the operation of the Object division parallelization mode, according to an embodiment the pr went invention. CPU executes the 3-D
graphic application 310, which along with standard graphic library generate a stream of graphic open ations and polygonal data. They are typically organized in data blocks either as Display List, Ver tex Array or free polygons, which ara polygons that ar a neither organized in Display List nor in Vertex Array. The Software Hub Driver at step 320 marks the polygonal data and graphic commands far distribution between multiple GPUs, in a way that the overall Iaad is balanced. The marked data is forwarded to Hardware Hub. At step 330 it is distributed via the Hub Routes to multiple GPUs. After the rendering process in GPUs at step 340 is accomplished, the Frame Buffers are retrieved and forwarded via the Hub Routes to the compositing unit at step 350. Here the fi ame buffers are composited to a single frame buffer, which is forwarded to the designated GPU for display. The single frame buffer is displayed at step 360.
Fig. 4 discloses the open ation of the Image division parallelization mode, according to an embodiment the present invention. In this mode the Software Hub Driver assigns different viewports to GPUs in order to subdivide the screen between them. The viewports aspects are set according to Load balancii2g considerations, to keep the overall GPU Ioad evenly balanced. In step 420, the entire polygonal data and graphic commands as a moved to the Hub Routes at Hardware Hub, and they are transmitted at step 430 to multiple GPUs. All GPUs receive the same data. After r endering at step 440, the partial frame buffers are brought to compositing unit in Hardware Hub for the full image creation at step 450, and then this image is moved out to designated GPU for display. The full image is displayed at step 460.
Fig. 5 disclases the operation of the Time division parallelization mode, actor ding to an embodiment the present invention. In time division the processing of each frame takes N frame time units, while N GPUs (or N
clusters of GPUs) are participating. The entire amount of polygons is forwarded to each GPU, for rendering. At each frame time unit, the Software Hub Driver redirects the polygonal data and graphic commands at step 530 to a cluster of GPUs at a time, while alternating between them. The data is transferred to the above cluster of GPUs via the Hub Routes, rendered in the GPUs at step 540, and then the resulting frame buffer at step 550 is redirected via Hub Routes to the designated GPU for display. All GPUs are coon dinated by SoftuTare Hub Driver to create a continuous sequence of frames. The resulting frame buffer is displayed at step 560.
The Hardware Hub competence is its scaling technology: Architecture and cluster of proprietary processes devoted to scale existing GPUs performance in PC based systems, by enabling the use of multiple G-PUs in parallel on the level of chip, card or chip IP (Intellectual Proper ty) core, and handling multiple bus paths between the CPU and GPU. The technology achieves linear increase in performance. It is~ invariant to a graphics vendor and also it is transparent to an application. In the present invention, the gr aphic pipeline bottlenecks of vertex processing, fragment processing and bus transfer are completely and intelligently resolved. As bottlenecks may shift between frames, the Hardware Hub is designed with a smart real-time feedback system between the Control Unit 2I0, disclosed in Fig. 2, and Software Hub Driver 123, disclosed in Fig.
1, by means of the bus, utilizing the different parallelization modes to overcome different bottlenecks and maintain maximum performance at the frame level.
The Software Hub Driver The Software Hub Driver is a software package residing in the PC and coexisting with computer's open ating system, standar d graphic libr ary, application and Vendor's GPU Driver. Fig. 6 is a schematic block diagram of a possible integration of the Software Hub Driver 630 into the operating system environment actor ding to an embodiment the pr went invention. Next to graphic application block 6I0 there is standard graphic libr ary block 620.
The Software Hub Driver 630 is located beneath the standard graphic library 620, intercepting the graphic command and data stream on the way to the Vendor's CPU Driver 640. The Software Hub Driver 630 also controls the Hardware Hub 660.
Fig.7 is a functional block diagr am presenting the main tasks of the Software Hub Driver, according to an embodiment the present invention. OS interface and Utilities block 710 is responsible for inter ception of the graphic commands from the standard graphic library, forwarding and creating graphic commands to Vendor's GPU Driver, controlling the Har dwar a Hub, r egistr y and installation, OS services and utilities. State Monitoring block 720 is responsible for maintaining consistency of graphic machine states across the GPUs, based on the input graphic commands stream. Application and graphic resources analysis block 730 is responsible for the application observation -estimating the type of graphic load and bottleneck in the current application graphic context, graphic resources (GPUs) load estimation for load balance based on graphic commands stream and time measurements, handling the feedback from GPUs in regard to load balancing. Parallelism policy management block 740 is based on Load analysis. All parallelization modes, which are the Object division mode, Image division mode and Time division mode, are combined together in order to achieve best perfor mance and optimal load handling. Parallelization policy is based on the analysis of the load, and it must preserve the state of the graphic system at all relevant GPUs across the elects onic cir cuit or chip. ~'or example, changing of a state by adding a new light sour ce in the scene at some time point, must affect all subsequent polygons at different GPUs. Parallelism policy management block 740 is responsible for the interpretation of the policy for specific manipulation in gr aphic functions. Distributed graphic functions control block 750 is responsible for modification of gr aphic command and data stream based on the parallelization policy.
Obiect Division Decomposition Process Object division is a well known concept, but data decomposition (partition of data and gr aphic commands between GPUs), while being also a known concept, has not been applied yet effectively, as it imposes various great _~8_ difficulties. These difficulties are handled successfully by a proposed process and its implementation according to the present invention.
The decomposition, and more importantly, the composition, must be accurate and efficient. Certain open ations must be performed in the order they are submitted by the application. For example, in case of semi-transparency, the commands and polygon stream must keep a certain order for creating a correct graphic r esult.
Intelligent decomposition of data and graphic commands is needed, pr eserving the basic features of graphic libraries as state machines, and complying with the graphic standards. The proposed decomposition process, according to the present invention, is performed by the Software Hub Driver. CPU runs the 3-D
graphic application, generating flow of graphic commands and data. They are typically organized in blocks, such as Display Lists or Vertex Arrays, stored in the system memory.
According to the present invention, the Software Hub Driver, running in the CPU, decomposes the set of scene polygons (or vertices). Their physical distribution is performed by the Har dwar a Hub.
The polygons are rendered in the GPU, while maintaining the r esulting Fr ame Buffer in local memory. All FBs are transferred, via Hub Routes, to compositing unit in Hardware Hub, to be merged into single FB. Finally, the composited FB is forwar ded for display.
The Software Hub Driver carries out the following process of distribution of the polygons between the multiple GPUs. It is assumed, that the regular way the graphic application works, remains unchanged. Per frame, a typical application generates a str eam of gr aphic calls that includes blocks of gr aphic _ ~g _ data; each block consists of a list of geometric operations, such as single vertex operations or' buffer based operations (vertex array). Typically, the decomposition process splits the data between GPUs pr eserving the blocks as basic data units. Geometric open ations are attached to the blocks) of data, instructing the way the data is handled. A block is directed to designated GPUs. However, there are operations belonging to the group of Blocking Operations, such as Flush, Swap, Alpha blending, which affect the entire graphic system, setting the system to blocking mode. Blocking operations are exceptional in that they require a composed valid FB data, thus in the parallel setting of the pr went invention, they have an effect on alI GPUs. Therefor e, whenever one of the Blocking open ations is issued, all the GPUs must be synchronized. Each frame has at least 2 blocking operations: Flush and Swap, which terminate the frame.
Fig. 8 is a flow chart presenting the process for distribution of the polygons between the multiple GPUs, according to an embodiment the present invention. The fi ame activity starts with distributing blocks of data between GPUs. Each graphic operation is tested for blocking mode at step 820. In a regular (non-blocking) path, data is redirected to the designated GPU at the step 830. This loop is repeated until a blocking operation is detected.
~~hen the blocking operation is detected, all GPUs must be synchronized at step 840 by at least the following sequence:
- performing a flush open ation in order to terminate rendering and clean up the internal pipeline (flushing) in GPU;
- perfor ming a co mposition in order to mer ge the contents of FBs into a single FB; and - transmitting the contents of said single FB back to all GPUs, in order to create a common ground for continuation.
The Sv=ap operation activates the double buffering mechanism, swapping the back and front color buffers. If Swap is detected at step 850, it means that a composited complete frame must be terminated at all GPU, except GPUO. All GPUs have the final composed contents of a FB designated to store said contents, but only the one connected to the screen (GPUO) displays the image at step 860.
Another case is operations that are applied globally to the scene and need to be broadcasted to all the GPUs. If one of the other blocking operations is identified, such as Alpha blending for transparency, then all GPUs are~flushed as before at step 840, and merged into a common FB. This time the Svcrap aperation is not detected (step 850), and therefor a all GPUs have the same data, and as Iong as the blocking mode is on (step 870), aII of them keep processing the same data (step 880). If the end of the block mode is detected at step 8'70, GPUs return working on designated data (step 830).
Adaptive handlin~of graphic load by combining three division methods In addition, the present invention introduces a dynamic load-balancing technique that combines the object division method with the image division and time division methods in image and time domains, based on the load exhibits by previous processing stages. Combining all the three parallel methods into a unified framework dramatically increases the effectiveness of our invention.
Parallel processing is implemented by a pipeline, such as any common GPU
allows the data to be processed in parallel, in time, image and object domains.
The processing performed on the graphical processing system, either in par allel on multi-GPU or sequential, results in a sequence of complete raster images stor ed in a frame buffer, and sent to the display unit. These images are -referred as flames in short. A frame consists of fragments. A fragment is an extended pixel stored in memory, which consists of attributes such as color, alpha, depth, stencil, etc. When processing is performed in parallel in the time domain, typically each GPU is responsible for the production of a complete frame. In the other two domains, which are the image and object domains, alI
G~'U operate in parallel to produce a single frame. Scr een-space parallel-processing implies that each GPU r enders a subset of the fragments of each frame, and object parallel-processing implies that the input data for each frame, in particular the geometric data (e.g., the polygon set representing the scene) is distributed between the multi GPUs.
Each one of three domains (time, image and object domains) has advantages arid disadvantages. The effectiveness of each discipline is a dynamic function based on input data. Moreover, in many cases no single discipline is superior.
In these cases a combination of two or even all the three disciplines may yield tl2e most optimum results.
Tie present invention describes the specification for a parallel-processing system of three-dimensional data. It provides a novel process for object p~.rallel-processing that consists of efficient decomposition of the data between tl~e different GPU, and then the composition of the frames produced on the various GPUs into a final frame ready to be r endered.
The present invention provides a method to integrate aII the thxee parallel nodes dynamically into a unified framework to achieve maximum load balancing. At each frame, the set of available GPUs can be reconfigured based on the time it took to render the previous flames, and the bottlenecks exhibited during the processing of these frames.
_ ~7 _ Fig. 9 discloses a sample configuration of the system, employing 8 GPUs, according to ari embodiment of the pr esent invention. Accor ding to the above sample configuration, a balanced graphic application is assumed. The GPUs are divided into two groups for time division parallelism. GPUs indexed with 1, 2, 3, and 4 are configured to process even frames and GPUs indexed with 5, 6, 7, and 8 are configured to process odd frames. Within each group, two GPU
subgroups are set for image division: the GPUs with the lower indexes (1,2 and 5,6 respectively) are configured to process half of the screen, and the high-indexed GPU (3,4 and 7,8 respectively) are configured to process the other half.
Finally, for the obj ect division, GPUs indexed with l, 3, 5 and- 7 or a fed with half of the obj ects, and GPUs indexed with 2 , 4, 6 and 8 are fed with the other half of the objects.
If at some point the system detects that the bottlenecks exhibited in previous frames occur at the raster stage of the pipeline, it means that fragment processing dominates the time it takes to render the frames and that the configur ation is imbalanced. At that point the GPUs or a reconfigured, so that each GPU will render a quarter of the screen within the respective frame. The original partition for time division, between GPUs 1,x,3,4 and between 5,6,7,8 still holds, but GPU 2 and GPU 5 are configured to render the first quarter of screen in even and odd frames respectively. GPUs I and GPU 6 - the second quarter , GPU 4 and GPU 'l - the thin d quarter, and GPU 3 and GPU 8 - the for th quarter . No obj ect division is implied.
In addition, if at some point the system detects that the bottleneck exhibited in pr evious frames occurs at the geometr y stage of the pipe, the GPUs are reconfigured, so that each GPU will process a quay ter of the geometrical data vcTithin the respective frame. That is, GPU 3 and GPU 5 are configured to process the fir st quar ter of the polygons in even and odd frames respectively.
GPU 1 and GPU 7 - the second quarter, GPU 4 and GPU 6 - the third quarter and GPU 2 and GPU 8 - the forth quarter. No image division is implied.
It should be noted, that taking 8 GPUs is sufficient in or der to combine all three parallel modes, which are time, image and object division modes, per fr ame. Taking the nuznber of GPUs larger than 8, also enables combining all 3 modes, but in a non-symmetric fashion. The flexibility also exists in frame count in a time'division cycle. In the above example, the cluster of 8 GPUs was broken down into the two groups, each group handling a frame. However, it is possible to extend the number of frames in a time division mode to a sequence, which is longer than 2 frames, for example 3 or 4 frames.
Taking a smaller number of GPUs still allows the combination of the parallel modes, however the combination of two modes only. For example, taking only 4 GPUs enables to combine image and object division modes, without time division mode. Tt is clearly understood from Fig. 9, while taking the group of GPUI, GPU2, GPU3 and GPU4, which is the left cluster. Similarly, the group of GPU1, GPU2, GPU5 and GPU6, which is the upper cluster, employs both object and time division modes. Finally, the configuration of the group of GPU2, GPU4, GPUS and GPU6, which is the middle cluster, employs image and time division modes.
It should be noted, that similarly to the above embodiments, any combination between the parallel modes can be scheduled to evenly balance the graphic load.
It also should be noted, that according to the present invention, the parallelization process between all GPUs rnay be based on an object division mode or image division mode or time division mode or any combination thereof in or der to optimize the processing performance of each frame.
While some embodiments of the invention have been described by way of illustration, it will be apparent that the invention can be put into practice vcTith many modifications, variations and adaptations, and with the use of numerous equivalents or alternative solutions that are within the scope of persons skilled in the art, without departing from the spirit of the invention or exceeding the scope of the claims.
Q'~7EI~, A PC BUS
Field of the Invention The present invention relates to a method and system for 3-D (three-dimensional) multiple graphic processing. More particularly, the invention relates to a method and system for improving the parallelization of image processing by Graphic Processing. Units (CPUs), based on unified framework of three parallelization methods, which are time division, image division and object division methods.
Definitions, Acronyms and Abbreviations Throughout this specification, the following definitions are employed:
GPLT: GPU (Graphic Processing Unit) like the CPU (Cents al Processing Unit), a GPU is a single-chip processor which is used primarily for computing 3-D
functions. This includes tasks such as, lighting effects, object transformations, and 3-D motion. These are mathematically-intensive tasks, which other wise, would put quite a strain on the CPU, but since the specialized GPU can handle most of the 3-D calculations, it helps the computer to perform these tasks more efficiently, and, of tour se, faster .
Polygon: Polygons in 3-D graphics are two-dimensional primitives, allowing generating and displaying of 3-D complex graphical objects. The polygons are the faces of the object and are composed from N vertices. Actually, a polygon is a closed plane figu r e, bounded by tbxee or mor a line segments.
Frame Buffer: a Fran2e Buffer (FB) is a buffer that stores the contents of an image, pixel by pixel. Generally, the portion of memory is reserved for holding the complete bit-mapped image that is sent to the monitor, for display.
Typically the frame buffer is stored in the memory chips on the video adapter.
In some instances; however, the video chipset is integrated into the motherboard design, and the frame buffer is stored in the general main memory.
Obiect compositin~nit: performs re-composition of multiple three-dimensional rasters into final image. The merged data is resolved for the closest pixel to the viewer in 3-D space, based on the depth value of pixels.
The new method, based on autonomous associative decision, allows the use of multiple GPUs for any fr ame complexity.
Display List: a Display List is a description of the 3-D scene through a list of gr aphic pr imitives, such as polygons and attributes. The display list provides inter mediate image storage for quick image r etrieval.
Vertex array: a Vertex Array is an array of vertices describing the 3-D scene.
A Vertex Array provides intermediate image storage for quick image retrieval.
Alpha blending Alpha blending controls the way in which the graphic infor mation is displayed, such as levels of tr ansparency, or opacity.
Backer ound of the Invention The three dimensional graphic pipeline architecture breaks-down into segmented stages of CPU, Bus, GPU vertex processing and GPU fiagment (pixel) processing. A given pipeline is only as strong as the weakest link of one of the above stages, thus the main bottleneck determines the overall throughput. Enhancing performance is all that required for reducing or eliminating bottlenecks. The major bottleneck strongly depends on the application. Extreme cases are CA:D-Like (Computer Aided Design) applications, char acterized by an abundance of polygons (vertices), vs. video-game applications having a small polygon count but intensive fragment activity (e.g., texturing). The first class suffers from vertex processing bottlenecks, while the second class suffers from fr agment bottlenecks. Both ar a frequently jammed over the PC bus. Many applications have mixed characteristics, where bottlenecks may randomly alternate between extremes, on a single frame basis.
The only way to improve the performance of the GPU is by means of parallelizing multiple GPUs according to one of the bottleneck solving methods. There are two pr edominant methods for rendering gr aphic data with multiple GPUs. These methods include time division (time domain composition), in which each GPU renders the next successive frame, and image division (screen space composition), in which each GPU renders a subset of the pixels of each flame. The third one, much less popular, is the object division (polygon decomposition) method.
In the time division method each GPU render s the next successive frame. It has the disadvantage of having each GPU render an entire frame. Thus, the speed at which each frame is r endered is limited to the rendering rate of a single GPU. While multiple GPUs enable a higher fr ame rate, a delay can be imparted in the response time (latency) of the system to a user's input. This occur s because, while at any given time, only one GPU is engaged in displaying a rendered frame, each of the GPUs is in the process of rendering one of a series of frames in a sequence. To maintain the high frame rate, the system delays the user's input until the specific GPU, which fir st received the signal cycles through the sequence, is again engaged in displaying its rendered frame.
In practical applications, this condition serves to limit the number of GPUs that are used in a system. With large data sets, then a is another bottleneck, due to the fact that each GPU must be able to access all the data. This requires -d-either maintaining multiple copy operations of lar ge data sets or possible conflicts in accessing the single copy operation.
Image division method splits the screen between N GPUs, such that each one displays 1/N of the image. The entir a polygon set is transfers ed to each GPU
for processing, however, the pixel processing is significantly reduced to the vvindoi~~ size. Image division has no latency issues, but it has a similar bottleneck with large data sets, since each GPU must examine the entire database to determine which graphic elements fall within the portion of the scr een allocated to said GPU. Image division method suits applications with intensive pixel processing.
Object division method is based on distribution of data subsets between multiple GPUs. The data subsets are rendered in the GPU pipeline, and converted to Frame Buffer (FB) of fragments (sub-image pixels). The multiple FB's sub-images have to be merged (composited) to generate the final image to be displayed. Object division delivers parallel rendering on the level of a single frame of very complex data consisting of large amount of polygons. The input data is decomposed in the polygon level and re-composed in the pixel level. A
proprietary driver intelligently distributes data streams, which are generated ..
by the application, between aII GPUs. The rasters, generated by the GPUs, are composited into final raster, and moved to the display. The object division method well suits applications that need to render a vast amount of geometr ical data. Typically, these are CAD, Digital Content Cr eation, and comparable visual simulation applications, considered as "viewers," meaning that the data has been pre-designed such that their three-dimensional positions in space are not under the interactive control of the user. However, the user does have interactive control over the viewer's position, the direction of view, and the scale of the graphic data. The user also may have control over the selection of a subset of the data and the method by which it is rendered.
This includes manipulating the effects of image lighting, color ation, transparency and other visual characteristics of the underlying data.
In above applications, the data tends to be very complex, as it usually consists of massive amount of geometrical entities at the display List or vertex array.
Therefore, the construction time of a single frame tends to be very long (e.g., typically 0.5 sec for 20 million polygons), which in turn slows down the overall system performance.
Therefore, then a is a need to provide a system which can guarantee the best system performance, being exposed to high traffic over the PC (Personal Computer) Bus.
It is an object of the present invention to provide an amplified strength of the GPU by means of parallelizing multiple GPUs.
It is another object of the present invention to provide a system, wherein the construction time of a single frame does not slow down the overall system response.
It is still another object of the present invention to provide a system and method, wherein the graphic pipeline bottlenecks of vertex processing and fragment processing are transparently and intelligently resolved.
It is still a fur then obj ect of the present invention to provide a system and method that has high scalability and unlimited scene complexity.
Tt is still a further object of the present invention to provide a process overcoming difficulties that are imposed by the data decomposition, which is partition of data and graphic commands between GPUs.
It is still a further object of the present invention to provide a method and s~Tstem for an intelligent decomposition of data and graphic commands, preserving the basic featur es of graphic libr aries as state machines and complying v~ith graphic standards.
Other objects and advantages of the invention will become apparent as the description proceeds.
Summary of the Invention '~'he present invention is directed to a system for improving the parallelization of image pr ocessing, using one or more par allelization modes, wherein the image that is displayed on at Ieast one computer screen by one or more Graphic Processing Units, which comprises:
one or more software applications, for issuing gr aphic commands;
one or more graphic libraries, for storing data used to implement the graphic commands;
one or more Software Hub Drivers, for controlling a Hardware Hub, for interacting with the operation system of the computer and the graphic libraries, for performing r eal-tine analysis of a data stream, from which frames of the image ar a generated, for deter mining the parallelization mode of each GPU, and for forwarding the data stream or a portion thereof to each GPU;
one or more GPU Drivers, for allowing the GPUs to interact with the graphic Iibr ar ies; and at least one I/O module for interconnecting between the Software module and the Hardware Hub, uTherein, the Hardware Hub distributes between the GPUs, for each frame, graphic commands and the data stream or a portion thereof, according to their relative complexity within the image, anal defines the complexity The Softv~Tare _7_ Hub Driver also composites a gr aphics output for display, using the outputs obtained from at least one GPU, while alternating, whenever required, the parallelization mode for the each fr ame.
Parallelization is based on an object division mode or on an image division mode or on a time division mode or on any combination thereof. The hardware hub comprises a compositing unit for composing a complete frame from processed portions of the data str cam. The hardware hub comprises a hub router for r outing polygonal data, for r outing gr aphic command stream, for routing pixel data and for routing the results of composition, while operating in the object division mode or in the image division mode or in the time division mode or in any combination thereof. The hardware hub comprises a control unit for receiving commands from the Software Hub Driver within the I/0 module. The hardware hub comprises a memory unit for storing intermediate processing results of one or more GPUs and data required for composition and transferring the processed data for display.
Preferably, the Software Hub Driver is capable of performing the following operations:
interception of the graphic commands from the standar d graphic library by means of the OS interface and utilities;
forwarding and creating graphic commands to the GPU Driver by means of the OS interface and utilities;
controlling the Hardware Hub, registry and installation operations by means of the OS interface and utilities:
maintaining the consistency of graphic machine states across the GPUs, based on the input graphic commands stream, while using state monitoring;
estimating the type of graphic load and overload in the executed application graphic context, while using application and gr aphic resources analysis;
_g_ load estimation of the GPUs load balance based on graphic commands stream and time measurements, while using application and gr aphic resources analysis;
adjusting the load distribution between GPUs according to feedback received from each GPU regarding the Ioad balance, while using application and graphic resources analysis;
performing manipulation in graphic functions accor ding to the current parallelization mode; and controlling the distributed graphic functions, while modifying the graphic commands and the data stream according to the current parallelization mode.
The present invention is directed to a method for improving the parallelization of image processing, using one or mare parallelization modes, wherein the image that is displayed on at least one computer screen by one or more Gr aphic Processing Units. Softwar a applications are provided for issuing gr aphic command and graphic libraries are provided for storing data used to implement the graphic commands. A Software Hub Drivers is provided for controlling a Hardware Hub, for interacting with the operation system of the computer and the graphic libraries, for performing real-time analysis of a data stream, from which frames of the image are generated, for determining the parallelization mode of each GPU, and for for war ding the data str eam or a portion thereof to each GPU. CPU Drivers ara provided for allowing the GPUs to interact with the 7.
graphic libraries and an I/0 module is provided for interconnecting between the Software module and the Har dware Hub. graphic commands and the data stream ar a portion thereof are distributed between the GPUs for each frame by the Hardware Hub, according to their relative complexity within the image, wherein the complexity is defined by the Software Hub Driver. The Software Hub Driver also composites a graphics output for display, using the outputs _g_ obtained from at least one GPU, while alternating, whenever required, the par~Ilelization made for the each fr ame.
Whenever the parallelization mode is an Object di~nsion parallelization mode, the following steps are performed;
for each frame, generating a stream of graphic operations and polygonal data;
marking the polygonal data and gr aphic commands by means of the Software Hub Driver for distribution between multiple GPUs;
sending the marked data to the Hardware Hub;
distributing the marked data via the Hub Routes to the multiple GPUs;
rendering the data by means of GPUs;
retrieving the data fr om the Frame Buffer s and forwarding the r etrieved data to the compositing unit via the Hub Routes;
compositing the content of the Frame Buffers into a single Frame Buffer; and forwarding the content of the single Frame Buffer to at least one designated GPU for display.
Wl~.enever the par allelization mode ~is an Image di~nsion parallelization mode, the following steps are performed:
sub dividing the screen to portions and assigning differ ent viewpo.r is to GPUs by means of the Software Hub Driver;
moving the entire polygonal data and graphic commands to the Hub Routes;
tr ansmitting the entir a polygonal data and graphic commands to GPUs, wherein each GPU receives the same data;
rendering the data by means of GPUs;
forwarding a portion of the content stored in the Frame Buffers to compositing unit in Hardwar a Hub for the complete image creation; and -IO-forwarding the image to at least one designated GPU for display.
Whenever the parallelization mode is a Time division parallelization mode, the following steps are performed:
forwarding to each one of the multiple GPUs the entire amount of polygons for r endering;
redirecting the entire polygonal data and graphic commands by means of Software Hub Driver to all GPUs, while alternating between them;
rendering the data by means of GPUs;
transferring rendered data from at least one GPU via the Hub Router; and r edirecting the resulting content of the Frame Buffer via Hub Router to at least one designated GPU for display.
The distribution of polygons between multiple GPUs is performed by distributing blocks of data between multiple GPUs and by testing each graphic operation for blocking mode, in which one or more parallelization modes are carried out, thereafter. The data is redirected in regular non-blocking path to at least one designated GPU, This process is repeated until a blocking operation is detected. Then GPUs are synchronized by performing a flush operation in order to terminate rendering and clean up the internal pipeline in each GPU;
performing a composition open ation for merging the contents of the Frame Buffers into a single Frame Buffer and byte ansmitting the single Frame Buffer back to all GPUs. Then the composited complete frame is terminated at all .
GPUs, except one or mor a designated GPUs, whenever a Swap operation is detected and displaying the image by means of the one or more designated GPUs. the same data is processed by all GPUs, as long as the blocking mode is active and the Swap operation is not detected. whenever the blocking mode is inactive, the designated data is further processed by multiple GPUs,.
Brief Description of the Drawings In the dr awings:
Fig. 1 is a block diagram of a multiple GPU architecture system, according to an embodiment of the pr esent invention;
Fig. 2 is a block diagram of Hardwar a Hub components, actor ding to an embodiment of the pr went invention;
Fig. 3 is a block diagram of Object division parallelization mode, according to an embodiment of the present invention;
Fig. 4 is a block diagram of Image division parallelization mode, actor ding to an embodiment of the present invention;
Fig. 5 is a block diagram of Time division parallelization mode, according to an embodiment of the pr esent invention;
- . Fig. 6 is a schematic block diagram of a possible integration of the Software Hub Driver into the operating system environment, according to an embodiment of the pr esent invention;
- Fig. 'l is a functional block diagr am presenting the main tasks of the Software Hub Driver, according to an embodiment of the present invention;
- Fig. 8 is a flow chart presenting an process for distribution of the polygons between the multiple GPUs, according to an embodiment of the pr esent invention; and Fig. 9 discloses a sample configuration of the system, employing 8 GPUs, according to an embodiment of the present invention.
Tt will be appreciated that for simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn to scale. For example, the dimensions of some of the elements may be exaggerated relative to other elements for clarity. Further, where considered appropriate, reference _ 12_ numerals may be repeated between the figures to indicate corresponding or analogous elements.
Detailed Description of the Preferred Embodiments The Multiple 3-D Graphic Pipeline The current invention calls for the introduction of an extended PC graphic architecture including novel operational component, the 3-D pipeline Hub.
Fig. 1 presents multiple GPU architecture system I00, according to an embodiment of the present invention. The hub 110 is located in a unique position on the PC bus, between the CPU (Central Processing Unit) and a cluster of GPUs 130. The multiple graphic pipeline architecture, as described in Fig. I, consists of Software Application 121, Graphic Library 122, Software Hub Driver 123, GPU Driver 124, Hardware Hub 110, cluster of GPUs 130, and displays) 140. Usually, one of the GPUs is designated as a display unit.
It should be noted, that it is possible to have more than one display unit, or include a display unit directly inside the Hardware Hub. A display unit can drive multiple screens, as well.
The Hub mechanism consists of a Hardware Hub component I10, located on the PC bus between I/O (Input/output) chipset 160, being a I/O module, and a cluster of CrPUs 130, and a Software module comprising Software Hub Driver 123, executed by the PC.
The Hardware Hub I10 carries on at least the following action:
- distributes decomposed polygon stream and graphic commands between GPUs;
- composites graphics output for display according to different parallel modes;
- makes cache of data; and - alternates modes of parallelism.
The Software Hub Driver 123, besides controlling the Hardware Hub I10, also carries on at least the following actions:
- interacts with the OS (Operation System) and graphic library, such as OpenGL, DirectX;
- performs real-time analysis of the data stieam;
- determines the parallelization mode; and decomposes the data and command stream.
One advantage of this method is that the unique location of the Har dware Hub 110 allows it to control the graphic pipeline, while being tr ansparent to the application 121. The application I21, along with Graphic Library 122, such as OpenGL, keeps working as it was a single GPU.
Another advantage of this method is that the unique location of the Hardware Hub 110 allows it to control a graphic pipeline between the User Interface 150 and Display 140, while being transparent to the GPU. Each GPU of the cluster keeps working as if it is the only graphic processor hooked on the I/O chipset 160.
Still another advantage of this method is that the unique location of the Hardware Hub 110 allows it to control the graphic pipeline for different parallelization modes: image division mode, time division mode or object division mode.
Further advantage of this method is that the unique location of the Hardware Hub 110 allows it to sense in real-time the varying parameters of application's load, such as polygon count, texture volume, human interaction, and to intelligently decide and carry on the current optimal parallelization method.
It should be noted that according to all embodiments of the pr esent invention, the displays) 140 may be connected directly or indirectly to the Hardware Hub 11o.
Reference is now made to Fig. 2, which discloses the components of the Hardv~are Hub I10, according to an embodiment the present invention.
Control Unit 210, accepts proprietary commands from the Software Hub Driver over the bus, and accordingly controls the other units. Hub Roister 230 routs polygonal data and graphic command stream from left to right, routs pixel data from right to compositing unit, and routs compositing results to the right. Compositing Unit 240 performs various compasiting schemes according to pai allelization mode. Auxiliary Memory block 220 is used for storing internzediate processing results of one or more GPUs, data required for composition and transferring the processed data for display.
The Hardware Hub 110 utilizes its units according to parallelization modes: an Object division mode, an Image division mode, a Time division mode. These modes are adaptively handled by the Hardware Hub 110 according to application needs.
Fig. 3 discloses the operation of the Object division parallelization mode, according to an embodiment the pr went invention. CPU executes the 3-D
graphic application 310, which along with standard graphic library generate a stream of graphic open ations and polygonal data. They are typically organized in data blocks either as Display List, Ver tex Array or free polygons, which ara polygons that ar a neither organized in Display List nor in Vertex Array. The Software Hub Driver at step 320 marks the polygonal data and graphic commands far distribution between multiple GPUs, in a way that the overall Iaad is balanced. The marked data is forwarded to Hardware Hub. At step 330 it is distributed via the Hub Routes to multiple GPUs. After the rendering process in GPUs at step 340 is accomplished, the Frame Buffers are retrieved and forwarded via the Hub Routes to the compositing unit at step 350. Here the fi ame buffers are composited to a single frame buffer, which is forwarded to the designated GPU for display. The single frame buffer is displayed at step 360.
Fig. 4 discloses the open ation of the Image division parallelization mode, according to an embodiment the present invention. In this mode the Software Hub Driver assigns different viewports to GPUs in order to subdivide the screen between them. The viewports aspects are set according to Load balancii2g considerations, to keep the overall GPU Ioad evenly balanced. In step 420, the entire polygonal data and graphic commands as a moved to the Hub Routes at Hardware Hub, and they are transmitted at step 430 to multiple GPUs. All GPUs receive the same data. After r endering at step 440, the partial frame buffers are brought to compositing unit in Hardware Hub for the full image creation at step 450, and then this image is moved out to designated GPU for display. The full image is displayed at step 460.
Fig. 5 disclases the operation of the Time division parallelization mode, actor ding to an embodiment the present invention. In time division the processing of each frame takes N frame time units, while N GPUs (or N
clusters of GPUs) are participating. The entire amount of polygons is forwarded to each GPU, for rendering. At each frame time unit, the Software Hub Driver redirects the polygonal data and graphic commands at step 530 to a cluster of GPUs at a time, while alternating between them. The data is transferred to the above cluster of GPUs via the Hub Routes, rendered in the GPUs at step 540, and then the resulting frame buffer at step 550 is redirected via Hub Routes to the designated GPU for display. All GPUs are coon dinated by SoftuTare Hub Driver to create a continuous sequence of frames. The resulting frame buffer is displayed at step 560.
The Hardware Hub competence is its scaling technology: Architecture and cluster of proprietary processes devoted to scale existing GPUs performance in PC based systems, by enabling the use of multiple G-PUs in parallel on the level of chip, card or chip IP (Intellectual Proper ty) core, and handling multiple bus paths between the CPU and GPU. The technology achieves linear increase in performance. It is~ invariant to a graphics vendor and also it is transparent to an application. In the present invention, the gr aphic pipeline bottlenecks of vertex processing, fragment processing and bus transfer are completely and intelligently resolved. As bottlenecks may shift between frames, the Hardware Hub is designed with a smart real-time feedback system between the Control Unit 2I0, disclosed in Fig. 2, and Software Hub Driver 123, disclosed in Fig.
1, by means of the bus, utilizing the different parallelization modes to overcome different bottlenecks and maintain maximum performance at the frame level.
The Software Hub Driver The Software Hub Driver is a software package residing in the PC and coexisting with computer's open ating system, standar d graphic libr ary, application and Vendor's GPU Driver. Fig. 6 is a schematic block diagram of a possible integration of the Software Hub Driver 630 into the operating system environment actor ding to an embodiment the pr went invention. Next to graphic application block 6I0 there is standard graphic libr ary block 620.
The Software Hub Driver 630 is located beneath the standard graphic library 620, intercepting the graphic command and data stream on the way to the Vendor's CPU Driver 640. The Software Hub Driver 630 also controls the Hardware Hub 660.
Fig.7 is a functional block diagr am presenting the main tasks of the Software Hub Driver, according to an embodiment the present invention. OS interface and Utilities block 710 is responsible for inter ception of the graphic commands from the standard graphic library, forwarding and creating graphic commands to Vendor's GPU Driver, controlling the Har dwar a Hub, r egistr y and installation, OS services and utilities. State Monitoring block 720 is responsible for maintaining consistency of graphic machine states across the GPUs, based on the input graphic commands stream. Application and graphic resources analysis block 730 is responsible for the application observation -estimating the type of graphic load and bottleneck in the current application graphic context, graphic resources (GPUs) load estimation for load balance based on graphic commands stream and time measurements, handling the feedback from GPUs in regard to load balancing. Parallelism policy management block 740 is based on Load analysis. All parallelization modes, which are the Object division mode, Image division mode and Time division mode, are combined together in order to achieve best perfor mance and optimal load handling. Parallelization policy is based on the analysis of the load, and it must preserve the state of the graphic system at all relevant GPUs across the elects onic cir cuit or chip. ~'or example, changing of a state by adding a new light sour ce in the scene at some time point, must affect all subsequent polygons at different GPUs. Parallelism policy management block 740 is responsible for the interpretation of the policy for specific manipulation in gr aphic functions. Distributed graphic functions control block 750 is responsible for modification of gr aphic command and data stream based on the parallelization policy.
Obiect Division Decomposition Process Object division is a well known concept, but data decomposition (partition of data and gr aphic commands between GPUs), while being also a known concept, has not been applied yet effectively, as it imposes various great _~8_ difficulties. These difficulties are handled successfully by a proposed process and its implementation according to the present invention.
The decomposition, and more importantly, the composition, must be accurate and efficient. Certain open ations must be performed in the order they are submitted by the application. For example, in case of semi-transparency, the commands and polygon stream must keep a certain order for creating a correct graphic r esult.
Intelligent decomposition of data and graphic commands is needed, pr eserving the basic features of graphic libraries as state machines, and complying with the graphic standards. The proposed decomposition process, according to the present invention, is performed by the Software Hub Driver. CPU runs the 3-D
graphic application, generating flow of graphic commands and data. They are typically organized in blocks, such as Display Lists or Vertex Arrays, stored in the system memory.
According to the present invention, the Software Hub Driver, running in the CPU, decomposes the set of scene polygons (or vertices). Their physical distribution is performed by the Har dwar a Hub.
The polygons are rendered in the GPU, while maintaining the r esulting Fr ame Buffer in local memory. All FBs are transferred, via Hub Routes, to compositing unit in Hardware Hub, to be merged into single FB. Finally, the composited FB is forwar ded for display.
The Software Hub Driver carries out the following process of distribution of the polygons between the multiple GPUs. It is assumed, that the regular way the graphic application works, remains unchanged. Per frame, a typical application generates a str eam of gr aphic calls that includes blocks of gr aphic _ ~g _ data; each block consists of a list of geometric operations, such as single vertex operations or' buffer based operations (vertex array). Typically, the decomposition process splits the data between GPUs pr eserving the blocks as basic data units. Geometric open ations are attached to the blocks) of data, instructing the way the data is handled. A block is directed to designated GPUs. However, there are operations belonging to the group of Blocking Operations, such as Flush, Swap, Alpha blending, which affect the entire graphic system, setting the system to blocking mode. Blocking operations are exceptional in that they require a composed valid FB data, thus in the parallel setting of the pr went invention, they have an effect on alI GPUs. Therefor e, whenever one of the Blocking open ations is issued, all the GPUs must be synchronized. Each frame has at least 2 blocking operations: Flush and Swap, which terminate the frame.
Fig. 8 is a flow chart presenting the process for distribution of the polygons between the multiple GPUs, according to an embodiment the present invention. The fi ame activity starts with distributing blocks of data between GPUs. Each graphic operation is tested for blocking mode at step 820. In a regular (non-blocking) path, data is redirected to the designated GPU at the step 830. This loop is repeated until a blocking operation is detected.
~~hen the blocking operation is detected, all GPUs must be synchronized at step 840 by at least the following sequence:
- performing a flush open ation in order to terminate rendering and clean up the internal pipeline (flushing) in GPU;
- perfor ming a co mposition in order to mer ge the contents of FBs into a single FB; and - transmitting the contents of said single FB back to all GPUs, in order to create a common ground for continuation.
The Sv=ap operation activates the double buffering mechanism, swapping the back and front color buffers. If Swap is detected at step 850, it means that a composited complete frame must be terminated at all GPU, except GPUO. All GPUs have the final composed contents of a FB designated to store said contents, but only the one connected to the screen (GPUO) displays the image at step 860.
Another case is operations that are applied globally to the scene and need to be broadcasted to all the GPUs. If one of the other blocking operations is identified, such as Alpha blending for transparency, then all GPUs are~flushed as before at step 840, and merged into a common FB. This time the Svcrap aperation is not detected (step 850), and therefor a all GPUs have the same data, and as Iong as the blocking mode is on (step 870), aII of them keep processing the same data (step 880). If the end of the block mode is detected at step 8'70, GPUs return working on designated data (step 830).
Adaptive handlin~of graphic load by combining three division methods In addition, the present invention introduces a dynamic load-balancing technique that combines the object division method with the image division and time division methods in image and time domains, based on the load exhibits by previous processing stages. Combining all the three parallel methods into a unified framework dramatically increases the effectiveness of our invention.
Parallel processing is implemented by a pipeline, such as any common GPU
allows the data to be processed in parallel, in time, image and object domains.
The processing performed on the graphical processing system, either in par allel on multi-GPU or sequential, results in a sequence of complete raster images stor ed in a frame buffer, and sent to the display unit. These images are -referred as flames in short. A frame consists of fragments. A fragment is an extended pixel stored in memory, which consists of attributes such as color, alpha, depth, stencil, etc. When processing is performed in parallel in the time domain, typically each GPU is responsible for the production of a complete frame. In the other two domains, which are the image and object domains, alI
G~'U operate in parallel to produce a single frame. Scr een-space parallel-processing implies that each GPU r enders a subset of the fragments of each frame, and object parallel-processing implies that the input data for each frame, in particular the geometric data (e.g., the polygon set representing the scene) is distributed between the multi GPUs.
Each one of three domains (time, image and object domains) has advantages arid disadvantages. The effectiveness of each discipline is a dynamic function based on input data. Moreover, in many cases no single discipline is superior.
In these cases a combination of two or even all the three disciplines may yield tl2e most optimum results.
Tie present invention describes the specification for a parallel-processing system of three-dimensional data. It provides a novel process for object p~.rallel-processing that consists of efficient decomposition of the data between tl~e different GPU, and then the composition of the frames produced on the various GPUs into a final frame ready to be r endered.
The present invention provides a method to integrate aII the thxee parallel nodes dynamically into a unified framework to achieve maximum load balancing. At each frame, the set of available GPUs can be reconfigured based on the time it took to render the previous flames, and the bottlenecks exhibited during the processing of these frames.
_ ~7 _ Fig. 9 discloses a sample configuration of the system, employing 8 GPUs, according to ari embodiment of the pr esent invention. Accor ding to the above sample configuration, a balanced graphic application is assumed. The GPUs are divided into two groups for time division parallelism. GPUs indexed with 1, 2, 3, and 4 are configured to process even frames and GPUs indexed with 5, 6, 7, and 8 are configured to process odd frames. Within each group, two GPU
subgroups are set for image division: the GPUs with the lower indexes (1,2 and 5,6 respectively) are configured to process half of the screen, and the high-indexed GPU (3,4 and 7,8 respectively) are configured to process the other half.
Finally, for the obj ect division, GPUs indexed with l, 3, 5 and- 7 or a fed with half of the obj ects, and GPUs indexed with 2 , 4, 6 and 8 are fed with the other half of the objects.
If at some point the system detects that the bottlenecks exhibited in previous frames occur at the raster stage of the pipeline, it means that fragment processing dominates the time it takes to render the frames and that the configur ation is imbalanced. At that point the GPUs or a reconfigured, so that each GPU will render a quarter of the screen within the respective frame. The original partition for time division, between GPUs 1,x,3,4 and between 5,6,7,8 still holds, but GPU 2 and GPU 5 are configured to render the first quarter of screen in even and odd frames respectively. GPUs I and GPU 6 - the second quarter , GPU 4 and GPU 'l - the thin d quarter, and GPU 3 and GPU 8 - the for th quarter . No obj ect division is implied.
In addition, if at some point the system detects that the bottleneck exhibited in pr evious frames occurs at the geometr y stage of the pipe, the GPUs are reconfigured, so that each GPU will process a quay ter of the geometrical data vcTithin the respective frame. That is, GPU 3 and GPU 5 are configured to process the fir st quar ter of the polygons in even and odd frames respectively.
GPU 1 and GPU 7 - the second quarter, GPU 4 and GPU 6 - the third quarter and GPU 2 and GPU 8 - the forth quarter. No image division is implied.
It should be noted, that taking 8 GPUs is sufficient in or der to combine all three parallel modes, which are time, image and object division modes, per fr ame. Taking the nuznber of GPUs larger than 8, also enables combining all 3 modes, but in a non-symmetric fashion. The flexibility also exists in frame count in a time'division cycle. In the above example, the cluster of 8 GPUs was broken down into the two groups, each group handling a frame. However, it is possible to extend the number of frames in a time division mode to a sequence, which is longer than 2 frames, for example 3 or 4 frames.
Taking a smaller number of GPUs still allows the combination of the parallel modes, however the combination of two modes only. For example, taking only 4 GPUs enables to combine image and object division modes, without time division mode. Tt is clearly understood from Fig. 9, while taking the group of GPUI, GPU2, GPU3 and GPU4, which is the left cluster. Similarly, the group of GPU1, GPU2, GPU5 and GPU6, which is the upper cluster, employs both object and time division modes. Finally, the configuration of the group of GPU2, GPU4, GPUS and GPU6, which is the middle cluster, employs image and time division modes.
It should be noted, that similarly to the above embodiments, any combination between the parallel modes can be scheduled to evenly balance the graphic load.
It also should be noted, that according to the present invention, the parallelization process between all GPUs rnay be based on an object division mode or image division mode or time division mode or any combination thereof in or der to optimize the processing performance of each frame.
While some embodiments of the invention have been described by way of illustration, it will be apparent that the invention can be put into practice vcTith many modifications, variations and adaptations, and with the use of numerous equivalents or alternative solutions that are within the scope of persons skilled in the art, without departing from the spirit of the invention or exceeding the scope of the claims.
Claims (13)
1. A system for improving the parallelization of image processing, using one or more parallelization modes, wherein said image that is displayed on at least one computer screen by one or more Graphic Processing Units, comprising at least:
a. one or mor a software applications, for issuing graphic commands;
b. one or mor a graphic libraries, for storing data used to implement said graphic commands;
c. one or more Software Hub Drivers, for controlling a Hardware Hub, for interacting with the operation system of said computer and said graphic libraries, for performing real-time analysis of a data stream, from which frames of said image are generated, for determining the parallelization mode of each GPU, and for forwarding said data stream or a portion thereof to each GPU;
d. one or more GPU Drivers, for allowing said GPUs to interact with said graphic libraries; and e. at least one I/O module for interconnecting between said Software module and said Hardware Hub, wherein, said Hardware Hub distributes, for each frame, between said GPUs, graphic commands and said data stream or a portion thereof, according to their relative complexity within said image, said complexity is defined by said Software Hub Driver; and composites a graphics output for display, using the outputs obtained fr om at least one GPU, while alternating, whenever required, said parallelization mode for said each frame.
a. one or mor a software applications, for issuing graphic commands;
b. one or mor a graphic libraries, for storing data used to implement said graphic commands;
c. one or more Software Hub Drivers, for controlling a Hardware Hub, for interacting with the operation system of said computer and said graphic libraries, for performing real-time analysis of a data stream, from which frames of said image are generated, for determining the parallelization mode of each GPU, and for forwarding said data stream or a portion thereof to each GPU;
d. one or more GPU Drivers, for allowing said GPUs to interact with said graphic libraries; and e. at least one I/O module for interconnecting between said Software module and said Hardware Hub, wherein, said Hardware Hub distributes, for each frame, between said GPUs, graphic commands and said data stream or a portion thereof, according to their relative complexity within said image, said complexity is defined by said Software Hub Driver; and composites a graphics output for display, using the outputs obtained fr om at least one GPU, while alternating, whenever required, said parallelization mode for said each frame.
2. System according to claim 1, wherein the parallelization is based on an object division mode or on an image division mode or on a time division mode of on any combination thereof.
3. System according to claim 1, wherein the hardware hub comprises at least one compositing unit at least for composing a complete frame from processed portions of the data stream.
4. System according to claim 2, wherein the hardware hub comprises at least one hub router for routing polygonal data, for routing graphic command stream, for routing pixel data and for routing the results of composition, while operating in the object division mode or in the image division mode or in the time division mode or in any combination thereof.
5. System according to claim l, wherein the hardware hub comprises at least one control unit for receiving commands from the Software Hub Driver within the I/O module.
6. System according to claim 1, wherein the hardware hub comprises a memory unit for storing intermediate processing results of one or more GPUs and data required for composition and transferring the processed data for display.
7. System according to claim 1, wherein the Software Hub Driver is capable of performing the following operations:
a. interception of the graphic commands from the standard graphic library by means of the OS interface and utilities;
b. forwarding and creating graphic commands to the GPU Driver by means of the OS interface and utilities;
c. controlling the Hardware Hub, registry and installation operations by means of the OS interface and utilities:
d. maintaining the consistency of graphic machine states across the GPUs, based on the input graphic commands stream, while using state monitoring;
e. estimating the type of graphic load and overload in the executed application graphic context, while using application and graphic resources analysis;
f. load estimation of the GPUs load balance based on graphic commands stream and time measurements, while using application and graphic resources analysis;
g. adjusting the load distribution between GPUs according to feedback received from each GPU regarding said load balance, while using application and graphic resources analysis;
h. performing manipulation in graphic functions according to the current parallelization mode; and i. controlling the distributed graphic functions, while modifying said graphic commands and said data stream according to said current parallelization mode.
a. interception of the graphic commands from the standard graphic library by means of the OS interface and utilities;
b. forwarding and creating graphic commands to the GPU Driver by means of the OS interface and utilities;
c. controlling the Hardware Hub, registry and installation operations by means of the OS interface and utilities:
d. maintaining the consistency of graphic machine states across the GPUs, based on the input graphic commands stream, while using state monitoring;
e. estimating the type of graphic load and overload in the executed application graphic context, while using application and graphic resources analysis;
f. load estimation of the GPUs load balance based on graphic commands stream and time measurements, while using application and graphic resources analysis;
g. adjusting the load distribution between GPUs according to feedback received from each GPU regarding said load balance, while using application and graphic resources analysis;
h. performing manipulation in graphic functions according to the current parallelization mode; and i. controlling the distributed graphic functions, while modifying said graphic commands and said data stream according to said current parallelization mode.
8. A method for improving the parallelization of image processing, using one or more parallelization modes, wherein said image that is displayed on at least one computer screen by one or more Graphic Processing Units, comprising at least:
a. providing one or more software applications, for issuing graphic commands;
b. providing one or more graphic libraries, for storing data used to implement said graphic commands;
c. providing one or more Software Hub Drivers, for controlling a Hardware Hub, for interacting with the operation system of said computer and said graphic libraries, for performing real-time analysis of a data stream, from which frames of said image are generated, for determining the parallelization mode of each GPU, and for forwarding said data stream or a portion thereof to each GPU;
d. providing one or mor a GPU Drivers, for allowing said GPUs to interact with said graphic libraries.
e. providing at least one I/O module for interconnecting between said Software module and said Hardware Hub; and f. for each frame, distributing between said GPUs and by means of said Hardware Hub, graphic commands and said data stream or a portion thereof, according to their relative complexity within said image, said complexity is defined by said Software Hub Driver; and compositing a graphics output for display, using the outputs obtained from at least one GPU, while alternating, whenever required, said parallelization mode for said each frame.
a. providing one or more software applications, for issuing graphic commands;
b. providing one or more graphic libraries, for storing data used to implement said graphic commands;
c. providing one or more Software Hub Drivers, for controlling a Hardware Hub, for interacting with the operation system of said computer and said graphic libraries, for performing real-time analysis of a data stream, from which frames of said image are generated, for determining the parallelization mode of each GPU, and for forwarding said data stream or a portion thereof to each GPU;
d. providing one or mor a GPU Drivers, for allowing said GPUs to interact with said graphic libraries.
e. providing at least one I/O module for interconnecting between said Software module and said Hardware Hub; and f. for each frame, distributing between said GPUs and by means of said Hardware Hub, graphic commands and said data stream or a portion thereof, according to their relative complexity within said image, said complexity is defined by said Software Hub Driver; and compositing a graphics output for display, using the outputs obtained from at least one GPU, while alternating, whenever required, said parallelization mode for said each frame.
9. A method actor ding to claim 8, wherein the parallelization mode is an Object division parallelization mode and the following steps are performed:
a. for each frame, generating a stream of graphic operations and polygonal data;
b. marking the polygonal data and graphic commands by means of the Software Hub Driver for distribution between multiple GPUs;
c. sending the marked data to the Hardware Hub;
d. distributing said marked data via the Hub Routes to said multiple GPUs;
e. rendering the data by means of GPUs;
f. retrieving the data from the Frame Buffers and forwarding the retrieved data to the compositing unit via the Hub Routes;
g. compositing the content of said Frame Buffer s into a single Frame Buffer; and h. forwarding the content of said single Frame Buffer to at least one designated GPU for display.
a. for each frame, generating a stream of graphic operations and polygonal data;
b. marking the polygonal data and graphic commands by means of the Software Hub Driver for distribution between multiple GPUs;
c. sending the marked data to the Hardware Hub;
d. distributing said marked data via the Hub Routes to said multiple GPUs;
e. rendering the data by means of GPUs;
f. retrieving the data from the Frame Buffers and forwarding the retrieved data to the compositing unit via the Hub Routes;
g. compositing the content of said Frame Buffer s into a single Frame Buffer; and h. forwarding the content of said single Frame Buffer to at least one designated GPU for display.
10. A method according to claim 8, wherein the parallelization mode is an Image division parallelization mode and the following steps are performed:
a. subdividing the screen to portions and assigning different viewports to GPUs by means of the Software Hub Driver;
b. moving the entire polygonal data and graphic commands to the Hub Routes;
c. transmitting said entire polygonal data and graphic commands to GPUs, wherein each GPU receives the same data;
d. rendering the data by means of GPUs;
e. forwarding a portion of the content stored in the Frame Buffers to compositing unit in Hardware Hub for the complete image creation;
and f. forwarding said image to at least one designated GPU for display.
a. subdividing the screen to portions and assigning different viewports to GPUs by means of the Software Hub Driver;
b. moving the entire polygonal data and graphic commands to the Hub Routes;
c. transmitting said entire polygonal data and graphic commands to GPUs, wherein each GPU receives the same data;
d. rendering the data by means of GPUs;
e. forwarding a portion of the content stored in the Frame Buffers to compositing unit in Hardware Hub for the complete image creation;
and f. forwarding said image to at least one designated GPU for display.
11. A method according to claim 8, wherein the par allelization mode is a Time division par allelization mode and the following steps are performed:
a. forwarding to each one of the multiple GPUs the entire amount of polygons for rendering;
b. redirecting the entire polygonal data and graphic commands by means of Software Hub Driver to all GPUs, while alternating between them;
c. rendering the data by means of GPUs;
d. transferring rendered data from at least one GPU via the Hub Routes; and e. redirecting the resulting content of the Frame Buffer via Hub Routes to at least one designated GPU for display.
a. forwarding to each one of the multiple GPUs the entire amount of polygons for rendering;
b. redirecting the entire polygonal data and graphic commands by means of Software Hub Driver to all GPUs, while alternating between them;
c. rendering the data by means of GPUs;
d. transferring rendered data from at least one GPU via the Hub Routes; and e. redirecting the resulting content of the Frame Buffer via Hub Routes to at least one designated GPU for display.
12. A method according to claim 8, wherein the distribution of polygons between multiple GPUs is performed by:
a. distributing blocks of data between multiple GPUs;
b. testing each graphic operation for blocking mode, in which one or more parallelization modes are carried out;
c. redirecting the data in regular non-blocking path to at least one designated GPU;
d. repeating step (b) and (c) until a blocking operation is detected;
e. synchronizing GPUs by the following sequence:
e.1. performing a flush operation in order to terminate rendering and clean up the internal pipeline in each GPU;
e.2. performing a composition operation for merging the contents of the Frame Buffers into a single Frame Buffer;
and e.3. transmitting said single Frame Buffer back to all GPUs;
f. terminating the composited complete frame at all GPUs, except one or more designated GPUs, whenever a Swap operation is detected and displaying the image by means of said one or more designated GPUs;
g. processing the same data by means of all GPUs, as long as the blocking mode is active and the Swap operation is not detected; and h. continuing to process the designated data by means of multiple GPUs, whenever the blocking mode is inactive.
a. distributing blocks of data between multiple GPUs;
b. testing each graphic operation for blocking mode, in which one or more parallelization modes are carried out;
c. redirecting the data in regular non-blocking path to at least one designated GPU;
d. repeating step (b) and (c) until a blocking operation is detected;
e. synchronizing GPUs by the following sequence:
e.1. performing a flush operation in order to terminate rendering and clean up the internal pipeline in each GPU;
e.2. performing a composition operation for merging the contents of the Frame Buffers into a single Frame Buffer;
and e.3. transmitting said single Frame Buffer back to all GPUs;
f. terminating the composited complete frame at all GPUs, except one or more designated GPUs, whenever a Swap operation is detected and displaying the image by means of said one or more designated GPUs;
g. processing the same data by means of all GPUs, as long as the blocking mode is active and the Swap operation is not detected; and h. continuing to process the designated data by means of multiple GPUs, whenever the blocking mode is inactive.
13. Method according to claim 8, wherein the parallelization is based on an object division mode or on an image division mode or on a time division mode or on any combination thereof.
Applications Claiming Priority (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US52310203P | 2003-11-19 | 2003-11-19 | |
US52308403P | 2003-11-19 | 2003-11-19 | |
US60/523,084 | 2003-11-19 | ||
US60/523,102 | 2003-11-19 | ||
PCT/IL2004/001069 WO2005050557A2 (en) | 2003-11-19 | 2004-11-19 | Method and system for multiple 3-d graphic pipeline over a pc bus |
Publications (1)
Publication Number | Publication Date |
---|---|
CA2546427A1 true CA2546427A1 (en) | 2005-06-02 |
Family
ID=34623170
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CA002546427A Abandoned CA2546427A1 (en) | 2003-11-19 | 2004-11-19 | Method and system for multiple 3-d graphic pipeline over a pc bus |
Country Status (6)
Country | Link |
---|---|
US (9) | US7808499B2 (en) |
EP (1) | EP1687732A4 (en) |
JP (1) | JP2007512613A (en) |
CN (1) | CN1890660A (en) |
CA (1) | CA2546427A1 (en) |
WO (1) | WO2005050557A2 (en) |
Families Citing this family (86)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8745173B1 (en) * | 2011-12-30 | 2014-06-03 | hopTo Inc. | Client computing system for and method of receiving cross-platform remote access to 3D graphics applications |
US7222305B2 (en) * | 2003-03-13 | 2007-05-22 | Oracle International Corp. | Method of sharing a desktop with attendees of a real-time collaboration |
US7808499B2 (en) | 2003-11-19 | 2010-10-05 | Lucid Information Technology, Ltd. | PC-based computing system employing parallelized graphics processing units (GPUS) interfaced with the central processing unit (CPU) using a PC bus and a hardware graphics hub having a router |
US7457921B2 (en) * | 2005-02-23 | 2008-11-25 | Microsoft Corporation | Write barrier for data storage integrity |
US8217950B2 (en) * | 2005-05-27 | 2012-07-10 | Advanced Micro Devices, Inc. | Video processing across multiple graphics processing units |
US10026140B2 (en) * | 2005-06-10 | 2018-07-17 | Nvidia Corporation | Using a scalable graphics system to enable a general-purpose multi-user computer system |
US7629978B1 (en) * | 2005-10-31 | 2009-12-08 | Nvidia Corporation | Multichip rendering with state control |
US7830387B2 (en) | 2006-11-07 | 2010-11-09 | Microsoft Corporation | Parallel engine support in display driver model |
KR100803220B1 (en) * | 2006-11-20 | 2008-02-14 | 삼성전자주식회사 | Method and apparatus for rendering of 3d graphics of multi-pipeline |
US7953912B2 (en) * | 2008-02-22 | 2011-05-31 | International Business Machines Corporation | Guided attachment of accelerators to computer systems |
US8250578B2 (en) * | 2008-02-22 | 2012-08-21 | International Business Machines Corporation | Pipelining hardware accelerators to computer systems |
US8726289B2 (en) * | 2008-02-22 | 2014-05-13 | International Business Machines Corporation | Streaming attachment of hardware accelerators to computer systems |
CN102016798A (en) * | 2008-03-11 | 2011-04-13 | 韩国科亚电子股份有限公司 | Processing 3D graphics supporting fixed pipeline |
US9117007B2 (en) * | 2008-05-14 | 2015-08-25 | Microsoft Technology Licensing, Llc | Visualization of streaming real-time data |
US8319782B2 (en) * | 2008-07-08 | 2012-11-27 | Dell Products, Lp | Systems and methods for providing scalable parallel graphics rendering capability for information handling systems |
CN101639930B (en) * | 2008-08-01 | 2012-07-04 | 辉达公司 | Method and system for processing graphical data by a series of graphical processors |
US8228337B1 (en) * | 2008-10-03 | 2012-07-24 | Nvidia Corporation | System and method for temporal load balancing across GPUs |
US8373709B2 (en) * | 2008-10-03 | 2013-02-12 | Ati Technologies Ulc | Multi-processor architecture and method |
US8427474B1 (en) | 2008-10-03 | 2013-04-23 | Nvidia Corporation | System and method for temporal load balancing across GPUs |
US8892804B2 (en) | 2008-10-03 | 2014-11-18 | Advanced Micro Devices, Inc. | Internal BUS bridge architecture and method in multi-processor systems |
US8368701B2 (en) * | 2008-11-06 | 2013-02-05 | Via Technologies, Inc. | Metaprocessor for GPU control and synchronization in a multiprocessor environment |
GB2465812A (en) * | 2008-12-01 | 2010-06-02 | Imagination Tech Ltd | Distributed processing for rendering 3D images |
KR101519825B1 (en) * | 2008-12-05 | 2015-05-13 | 삼성전자주식회사 | Data processing device and control method of the same |
US9479358B2 (en) * | 2009-05-13 | 2016-10-25 | International Business Machines Corporation | Managing graphics load balancing strategies |
US20110063305A1 (en) * | 2009-09-16 | 2011-03-17 | Nvidia Corporation | Co-processing techniques on heterogeneous graphics processing units |
JP5590849B2 (en) * | 2009-10-08 | 2014-09-17 | キヤノン株式会社 | Data processing apparatus including parallel processing circuit having a plurality of processing modules, its control apparatus, its control method, and program |
FR2952731B1 (en) | 2009-11-13 | 2011-11-04 | Bull Sas | METHOD AND DEVICE FOR OPTIMIZING THE EXECUTION OF SOFTWARE APPLICATIONS IN A MULTIPROCESSOR ARCHITECTURE COMPRISING SEVERAL INPUT / OUTPUT CONTROLLERS AND SECONDARY CALCULATION UNITS |
US9830889B2 (en) | 2009-12-31 | 2017-11-28 | Nvidia Corporation | Methods and system for artifically and dynamically limiting the display resolution of an application |
US20110242115A1 (en) * | 2010-03-30 | 2011-10-06 | You-Ming Tsao | Method for performing image signal processing with aid of a graphics processing unit, and associated apparatus |
US20110298816A1 (en) * | 2010-06-03 | 2011-12-08 | Microsoft Corporation | Updating graphical display content |
US8675002B1 (en) | 2010-06-09 | 2014-03-18 | Ati Technologies, Ulc | Efficient approach for a unified command buffer |
JP2012003619A (en) * | 2010-06-18 | 2012-01-05 | Sony Corp | Information processor, control method thereof and program |
JP5835942B2 (en) * | 2010-06-25 | 2015-12-24 | キヤノン株式会社 | Image processing apparatus, control method thereof, and program |
US20120001925A1 (en) | 2010-06-30 | 2012-01-05 | Ati Technologies, Ulc | Dynamic Feedback Load Balancing |
US20120001905A1 (en) * | 2010-06-30 | 2012-01-05 | Ati Technologies, Ulc | Seamless Integration of Multi-GPU Rendering |
WO2012054089A2 (en) * | 2010-10-21 | 2012-04-26 | Net Power And Light Inc. | Distributed processing pipeline and distributed layered application processing |
US20120147015A1 (en) * | 2010-12-13 | 2012-06-14 | Advanced Micro Devices, Inc. | Graphics Processing in a Multi-Processor Computing System |
US8866825B2 (en) * | 2010-12-15 | 2014-10-21 | Ati Technologies Ulc | Multiple display frame rendering method and apparatus |
US8786619B2 (en) | 2011-02-25 | 2014-07-22 | Adobe Systems Incorporated | Parallelized definition and display of content in a scripting environment |
JP5699755B2 (en) | 2011-03-31 | 2015-04-15 | 富士通株式会社 | Allocation method, allocation device, and allocation program |
CN102270095A (en) * | 2011-06-30 | 2011-12-07 | 威盛电子股份有限公司 | Multiple display control method and system |
US9727385B2 (en) * | 2011-07-18 | 2017-08-08 | Apple Inc. | Graphical processing unit (GPU) implementing a plurality of virtual GPUs |
US9019292B2 (en) | 2011-09-12 | 2015-04-28 | Microsoft Technology Licensing Llc | Reordering graph execution for processing optimization |
US20130148947A1 (en) * | 2011-12-13 | 2013-06-13 | Ati Technologies Ulc | Video player with multiple grpahics processors |
US8842122B2 (en) * | 2011-12-15 | 2014-09-23 | Qualcomm Incorporated | Graphics processing unit with command processor |
US8769052B1 (en) | 2011-12-30 | 2014-07-01 | hopTo Inc. | Cloud-based server computing system for and method of providing cross-platform remote access to 3D graphics applications |
US8766990B1 (en) | 2011-12-30 | 2014-07-01 | hopTo Inc. | Server computing system for and method of providing cross-platform remote access to 3D graphics applications |
US8838749B1 (en) | 2011-12-30 | 2014-09-16 | hopTo Inc. | Cloud based client computing system for and method of receiving cross-platform remote access to 3D graphics applications |
WO2013102798A1 (en) * | 2012-01-06 | 2013-07-11 | Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirketi | Distributed image generation system |
US8692832B2 (en) | 2012-01-23 | 2014-04-08 | Microsoft Corporation | Para-virtualized asymmetric GPU processors |
US20130328884A1 (en) * | 2012-06-08 | 2013-12-12 | Advanced Micro Devices, Inc. | Direct opencl graphics rendering |
CN102868601B (en) * | 2012-09-24 | 2015-05-13 | 上海天玑科技股份有限公司 | Routing system related to network topology based on graphic configuration database businesses |
TWI482118B (en) * | 2012-10-18 | 2015-04-21 | Wistron Corp | System and method for processing load balance of graphic streams |
TW201421420A (en) * | 2012-11-21 | 2014-06-01 | Inst Information Industry | Graphic processing unit virtual apparatus, graphic processing unit host apparatus, and graphic processing unit program processing methods thereof |
KR102124395B1 (en) | 2013-08-12 | 2020-06-18 | 삼성전자주식회사 | Graphics processing apparatus and method thereof |
US9348602B1 (en) | 2013-09-03 | 2016-05-24 | Amazon Technologies, Inc. | Resource allocation for staged execution pipelining |
US20150286456A1 (en) | 2014-01-11 | 2015-10-08 | Userful Corporation | Method and System of Video Wall Setup and Adjustment Using GUI and Display Images |
KR102277353B1 (en) | 2014-02-21 | 2021-07-15 | 삼성전자주식회사 | Video processing method and electronic device |
US9471955B2 (en) * | 2014-06-19 | 2016-10-18 | Apple Inc. | Multiple display pipelines driving a divided display |
CN104915320A (en) * | 2015-05-28 | 2015-09-16 | 浪潮电子信息产业股份有限公司 | GPU (Graphic Processing Unit) server |
CN106326186B (en) * | 2015-06-29 | 2019-04-30 | 深圳市中兴微电子技术有限公司 | A kind of system on chip, pattern drawing method, middle layer and embedded device |
US10499072B2 (en) * | 2016-02-17 | 2019-12-03 | Mimax, Inc. | Macro cell display compression multi-head raster GPU |
EP3301572B1 (en) * | 2016-09-30 | 2023-05-24 | Dassault Systèmes | Method, program and system for simulating a 3d scene with a set of computing resources running in parallel |
KR20180038793A (en) * | 2016-10-07 | 2018-04-17 | 삼성전자주식회사 | Method and apparatus for processing image data |
US10504259B2 (en) | 2017-04-24 | 2019-12-10 | Intel Corporation | Separately processing regions or objects or interest from a render engine to a display engine or a display panel |
US11055807B2 (en) * | 2017-06-12 | 2021-07-06 | Apple Inc. | Method and system for a transactional based display pipeline to interface with graphics processing units |
CN108153593A (en) * | 2017-12-22 | 2018-06-12 | 合肥寰景信息技术有限公司 | The GPU cluster framework and its load-balancing method of a kind of high concurrent |
CN108416836A (en) * | 2018-02-06 | 2018-08-17 | 上海滚滚文化传播有限公司 | It is a kind of to render the caricature production method for turning two dimensional surface by three-dimensional modeling |
US10573034B2 (en) | 2018-04-13 | 2020-02-25 | Honeywell International Inc. | System and method for translation of graphics to newer format using pattern matching |
KR102455820B1 (en) * | 2018-05-31 | 2022-10-18 | 후아웨이 테크놀러지 컴퍼니 리미티드 | Apparatus and method for command stream optimization and enhancement |
US10747207B2 (en) | 2018-06-15 | 2020-08-18 | Honeywell International Inc. | System and method for accurate automatic determination of “alarm-operator action” linkage for operator assessment and alarm guidance using custom graphics and control charts |
US10699159B2 (en) | 2018-08-17 | 2020-06-30 | Honeywell International Inc. | Objects aggregation and standardization for legacy graphics conversion |
CN109408456B (en) * | 2018-12-07 | 2023-08-29 | 中国地质大学(武汉) | S905D chip and STM32 chip based cooperative hardware circuit |
US11397624B2 (en) * | 2019-01-22 | 2022-07-26 | Arm Limited | Execution of cross-lane operations in data processing systems |
CN109962691A (en) * | 2019-04-04 | 2019-07-02 | 同济大学 | A kind of distributed self-adaption filter design method for sensor network |
CN111045623B (en) * | 2019-11-21 | 2023-06-13 | 中国航空工业集团公司西安航空计算技术研究所 | Method for processing graphics commands in multi-GPU splicing environment |
US11042957B1 (en) | 2019-12-12 | 2021-06-22 | Honeywell International Inc. | Identification of process and control linkages from custom graphics for the purpose of backbuilding standard operating procedures |
US11080814B1 (en) * | 2020-02-03 | 2021-08-03 | Sony Interactive Entertainment Inc. | System and method for efficient multi-GPU rendering of geometry by pretesting against screen regions using prior frame information |
CN111491059B (en) * | 2020-04-09 | 2021-08-10 | 上海众链科技有限公司 | Image rendering enhancement system and method |
CN111679877B (en) * | 2020-05-27 | 2023-08-29 | 浙江大华技术股份有限公司 | Method and device for changing background of terminal equipment and electronic equipment |
CN112057851A (en) * | 2020-09-02 | 2020-12-11 | 北京蔚领时代科技有限公司 | Multi-display-card-based real-time rendering method for single-frame picture |
DE112020007672T5 (en) * | 2020-11-12 | 2023-08-24 | Nvidia Corporation | INTRA LAYER ADAPTER FOR FIFTH GENERATION NEW RADIO COMMUNICATIONS (5G-NR) |
CN114625329A (en) * | 2020-12-11 | 2022-06-14 | 慧荣科技股份有限公司 | Multi-screen display control device |
CN114625330A (en) | 2020-12-11 | 2022-06-14 | 慧荣科技股份有限公司 | Multi-screen display control device |
US11574579B2 (en) * | 2021-06-23 | 2023-02-07 | Dell Products L.P. | Enabling display filters in collaborative environments |
CN113436232B (en) * | 2021-06-29 | 2023-03-24 | 上海律信信息科技有限公司 | Hardware acceleration method based on tracking algorithm |
Family Cites Families (219)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH07104923B2 (en) * | 1988-12-28 | 1995-11-13 | 工業技術院長 | Parallel image display processing method |
JP2770598B2 (en) * | 1990-06-13 | 1998-07-02 | 株式会社日立製作所 | Graphic display method and apparatus |
CA2073516A1 (en) | 1991-11-27 | 1993-05-28 | Peter Michael Kogge | Dynamic multi-mode parallel processor array architecture computer system |
JP3199205B2 (en) * | 1993-11-19 | 2001-08-13 | 株式会社日立製作所 | Parallel processing unit |
EP0693737A3 (en) * | 1994-07-21 | 1997-01-08 | Ibm | Method and apparatus for managing multiprocessor graphical workload distribution |
US5745762A (en) | 1994-12-15 | 1998-04-28 | International Business Machines Corporation | Advanced graphics driver architecture supporting multiple system emulations |
US5687357A (en) | 1995-04-14 | 1997-11-11 | Nvidia Corporation | Register array for utilizing burst mode transfer on local bus |
US5754866A (en) | 1995-05-08 | 1998-05-19 | Nvidia Corporation | Delayed interrupts with a FIFO in an improved input/output architecture |
US5758182A (en) | 1995-05-15 | 1998-05-26 | Nvidia Corporation | DMA controller translates virtual I/O device address received directly from application program command to physical i/o device address of I/O device on device bus |
US5623692A (en) | 1995-05-15 | 1997-04-22 | Nvidia Corporation | Architecture for providing input/output operations in a computer system |
US5909595A (en) | 1995-05-15 | 1999-06-01 | Nvidia Corporation | Method of controlling I/O routing by setting connecting context for utilizing I/O processing elements within a computer system to produce multimedia effects |
US5794016A (en) | 1995-12-11 | 1998-08-11 | Dynamic Pictures, Inc. | Parallel-processor graphics architecture |
KR100269106B1 (en) | 1996-03-21 | 2000-11-01 | 윤종용 | Multiprocessor graphics system |
CA2263264C (en) | 1996-08-14 | 2006-11-28 | Northern Telecom Limited | Internet-based telephone call manager |
US6118462A (en) | 1997-07-01 | 2000-09-12 | Memtrax Llc | Computer system controller having internal memory and external memory control |
US6169553B1 (en) | 1997-07-02 | 2001-01-02 | Ati Technologies, Inc. | Method and apparatus for rendering a three-dimensional scene having shadowing |
US6201545B1 (en) | 1997-09-23 | 2001-03-13 | Ati Technologies, Inc. | Method and apparatus for generating sub pixel masks in a three dimensional graphic processing system |
US6856320B1 (en) | 1997-11-25 | 2005-02-15 | Nvidia U.S. Investment Company | Demand-based memory system for graphics applications |
US6337686B2 (en) | 1998-01-07 | 2002-01-08 | Ati Technologies Inc. | Method and apparatus for line anti-aliasing |
US6496187B1 (en) | 1998-02-17 | 2002-12-17 | Sun Microsystems, Inc. | Graphics system configured to perform parallel sample to pixel calculation |
US6182196B1 (en) * | 1998-02-20 | 2001-01-30 | Ati International Srl | Method and apparatus for arbitrating access requests to a memory |
US6473089B1 (en) | 1998-03-02 | 2002-10-29 | Ati Technologies, Inc. | Method and apparatus for a video graphics circuit having parallel pixel processing |
US6259460B1 (en) | 1998-03-26 | 2001-07-10 | Silicon Graphics, Inc. | Method for efficient handling of texture cache misses by recirculation |
US7038692B1 (en) | 1998-04-07 | 2006-05-02 | Nvidia Corporation | Method and apparatus for providing a vertex cache |
US6092124A (en) | 1998-04-17 | 2000-07-18 | Nvidia Corporation | Method and apparatus for accelerating the rendering of images |
US6184908B1 (en) | 1998-04-27 | 2001-02-06 | Ati Technologies, Inc. | Method and apparatus for co-processing video graphics data |
US6212617B1 (en) | 1998-05-13 | 2001-04-03 | Microsoft Corporation | Parallel processing method and system using a lazy parallel data type to reduce inter-processor communication |
US6477687B1 (en) | 1998-06-01 | 2002-11-05 | Nvidia U.S. Investment Company | Method of embedding RAMS and other macrocells in the core of an integrated circuit chip |
US7068272B1 (en) | 2000-05-31 | 2006-06-27 | Nvidia Corporation | System, method and article of manufacture for Z-value and stencil culling prior to rendering in a computer graphics processing pipeline |
US6636215B1 (en) | 1998-07-22 | 2003-10-21 | Nvidia Corporation | Hardware-assisted z-pyramid creation for host-based occlusion culling |
US6646639B1 (en) | 1998-07-22 | 2003-11-11 | Nvidia Corporation | Modified method and apparatus for improved occlusion culling in graphics systems |
US7023437B1 (en) | 1998-07-22 | 2006-04-04 | Nvidia Corporation | System and method for accelerating graphics processing using a post-geometry data stream during multiple-pass rendering |
US6415345B1 (en) | 1998-08-03 | 2002-07-02 | Ati Technologies | Bus mastering interface control system for transferring multistream data over a host bus |
US6191800B1 (en) * | 1998-08-11 | 2001-02-20 | International Business Machines Corporation | Dynamic balancing of graphics workloads using a tiling strategy |
US6492987B1 (en) | 1998-08-27 | 2002-12-10 | Ati Technologies, Inc. | Method and apparatus for processing object elements that are being rendered |
US6188412B1 (en) | 1998-08-28 | 2001-02-13 | Ati Technologies, Inc. | Method and apparatus for performing setup operations in a video graphics system |
US6292200B1 (en) | 1998-10-23 | 2001-09-18 | Silicon Graphics, Inc. | Apparatus and method for utilizing multiple rendering pipes for a single 3-D display |
US6362825B1 (en) | 1999-01-19 | 2002-03-26 | Hewlett-Packard Company | Real-time combination of adjacent identical primitive data sets in a graphics call sequence |
US6753878B1 (en) | 1999-03-08 | 2004-06-22 | Hewlett-Packard Development Company, L.P. | Parallel pipelined merge engines |
JP3169933B2 (en) | 1999-03-16 | 2001-05-28 | 四国日本電気ソフトウェア株式会社 | Parallel drawing device |
US6535209B1 (en) | 1999-03-17 | 2003-03-18 | Nvidia Us Investments Co. | Data stream splitting and storage in graphics data processing |
US6288418B1 (en) | 1999-03-19 | 2001-09-11 | Nvidia Corporation | Multiuse input/output connector arrangement for graphics accelerator integrated circuit |
US6577320B1 (en) | 1999-03-22 | 2003-06-10 | Nvidia Corporation | Method and apparatus for processing multiple types of pixel component representations including processes of premultiplication, postmultiplication, and colorkeying/chromakeying |
US6181352B1 (en) | 1999-03-22 | 2001-01-30 | Nvidia Corporation | Graphics pipeline selectively providing multiple pixels or multiple textures |
US6333744B1 (en) | 1999-03-22 | 2001-12-25 | Nvidia Corporation | Graphics pipeline including combiner stages |
DE19917092A1 (en) | 1999-04-15 | 2000-10-26 | Sp3D Chip Design Gmbh | Accelerated method for grid forming of graphic basic element in order beginning with graphic base element instruction data to produce pixel data for graphic base element |
US6990076B1 (en) * | 1999-05-18 | 2006-01-24 | Mcnamara Timothy G | Synchronous bi-directional data transfer having increased bandwidth and scan test features |
US6424320B1 (en) * | 1999-06-15 | 2002-07-23 | Ati International Srl | Method and apparatus for rendering video |
US6469703B1 (en) | 1999-07-02 | 2002-10-22 | Ati International Srl | System of accessing data in a graphics system and method thereof |
US6442656B1 (en) | 1999-08-18 | 2002-08-27 | Ati Technologies Srl | Method and apparatus for interfacing memory with a bus |
US6578068B1 (en) | 1999-08-31 | 2003-06-10 | Accenture Llp | Load balancer in environment services patterns |
US6352479B1 (en) | 1999-08-31 | 2002-03-05 | Nvidia U.S. Investment Company | Interactive gaming server and online community forum |
US6657635B1 (en) | 1999-09-03 | 2003-12-02 | Nvidia Corporation | Binning flush in graphics data processing |
US6353439B1 (en) | 1999-12-06 | 2002-03-05 | Nvidia Corporation | System, method and computer program product for a blending operation in a transform module of a computer graphics pipeline |
US6452595B1 (en) | 1999-12-06 | 2002-09-17 | Nvidia Corporation | Integrated graphics processing unit with antialiasing |
US6573900B1 (en) | 1999-12-06 | 2003-06-03 | Nvidia Corporation | Method, apparatus and article of manufacture for a sequencer in a transform/lighting module capable of processing multiple independent execution threads |
US6198488B1 (en) | 1999-12-06 | 2001-03-06 | Nvidia | Transform, lighting and rasterization system embodied on a single semiconductor platform |
US6844880B1 (en) | 1999-12-06 | 2005-01-18 | Nvidia Corporation | System, method and computer program product for an improved programmable vertex processing model with instruction set |
US6870540B1 (en) | 1999-12-06 | 2005-03-22 | Nvidia Corporation | System, method and computer program product for a programmable pixel processing model with instruction set |
US6417851B1 (en) | 1999-12-06 | 2002-07-09 | Nvidia Corporation | Method and apparatus for lighting module in a graphics processor |
US7002577B2 (en) | 1999-12-06 | 2006-02-21 | Nvidia Corporation | Clipping system and method for a single graphics semiconductor platform |
US6473086B1 (en) | 1999-12-09 | 2002-10-29 | Ati International Srl | Method and apparatus for graphics processing using parallel graphics processors |
US6557065B1 (en) | 1999-12-20 | 2003-04-29 | Intel Corporation | CPU expandability bus |
US6760031B1 (en) | 1999-12-31 | 2004-07-06 | Intel Corporation | Upgrading an integrated graphics subsystem |
WO2001069207A1 (en) | 2000-03-16 | 2001-09-20 | Fuji Photo Film Co., Ltd. | Measuring method and instrument utilizing total reflection attenuation |
US6975319B1 (en) | 2000-03-24 | 2005-12-13 | Nvidia Corporation | System, method and article of manufacture for calculating a level of detail (LOD) during computer graphics processing |
US6831652B1 (en) | 2000-03-24 | 2004-12-14 | Ati International, Srl | Method and system for storing graphics data |
US6747654B1 (en) * | 2000-04-20 | 2004-06-08 | Ati International Srl | Multiple device frame synchronization method and apparatus |
US6741243B2 (en) | 2000-05-01 | 2004-05-25 | Broadcom Corporation | Method and system for reducing overflows in a computer graphics system |
US6725457B1 (en) | 2000-05-17 | 2004-04-20 | Nvidia Corporation | Semaphore enhancement to improve system performance |
US6728820B1 (en) | 2000-05-26 | 2004-04-27 | Ati International Srl | Method of configuring, controlling, and accessing a bridge and apparatus therefor |
US6789154B1 (en) | 2000-05-26 | 2004-09-07 | Ati International, Srl | Apparatus and method for transmitting data |
US6633296B1 (en) | 2000-05-26 | 2003-10-14 | Ati International Srl | Apparatus for providing data to a plurality of graphics processors and method thereof |
US6662257B1 (en) | 2000-05-26 | 2003-12-09 | Ati International Srl | Multiple device bridge apparatus and method thereof |
US6670958B1 (en) | 2000-05-26 | 2003-12-30 | Ati International, Srl | Method and apparatus for routing data to multiple graphics devices |
US6664963B1 (en) | 2000-05-31 | 2003-12-16 | Nvidia Corporation | System, method and computer program product for programmable shading using pixel shaders |
US6724394B1 (en) | 2000-05-31 | 2004-04-20 | Nvidia Corporation | Programmable pixel shading architecture |
US6690372B2 (en) | 2000-05-31 | 2004-02-10 | Nvidia Corporation | System, method and article of manufacture for shadow mapping |
US6593923B1 (en) | 2000-05-31 | 2003-07-15 | Nvidia Corporation | System, method and article of manufacture for shadow mapping |
US6532013B1 (en) | 2000-05-31 | 2003-03-11 | Nvidia Corporation | System, method and article of manufacture for pixel shaders for programmable shading |
US6801202B2 (en) | 2000-06-29 | 2004-10-05 | Sun Microsystems, Inc. | Graphics system configured to parallel-process graphics data using multiple pipelines |
US7405734B2 (en) | 2000-07-18 | 2008-07-29 | Silicon Graphics, Inc. | Method and system for presenting three-dimensional computer graphics images using multiple graphics processing units |
US6959110B1 (en) | 2000-08-17 | 2005-10-25 | Nvidia Corporation | Multi-mode texture compression algorithm |
US7116331B1 (en) | 2000-08-23 | 2006-10-03 | Intel Corporation | Memory controller hub interface |
US6842180B1 (en) | 2000-09-20 | 2005-01-11 | Intel Corporation | Opportunistic sharing of graphics resources to enhance CPU performance in an integrated microprocessor |
US6502173B1 (en) | 2000-09-29 | 2002-12-31 | Ati Technologies, Inc. | System for accessing memory and method therefore |
US6532525B1 (en) | 2000-09-29 | 2003-03-11 | Ati Technologies, Inc. | Method and apparatus for accessing memory |
US6828980B1 (en) | 2000-10-02 | 2004-12-07 | Nvidia Corporation | System, method and computer program product for z-texture mapping |
US6731298B1 (en) | 2000-10-02 | 2004-05-04 | Nvidia Corporation | System, method and article of manufacture for z-texture mapping |
JP3580789B2 (en) | 2000-10-10 | 2004-10-27 | 株式会社ソニー・コンピュータエンタテインメント | Data communication system and method, computer program, recording medium |
US6961057B1 (en) | 2000-10-12 | 2005-11-01 | Nvidia Corporation | Method and apparatus for managing and accessing depth data in a computer graphics system |
US6362997B1 (en) | 2000-10-16 | 2002-03-26 | Nvidia | Memory system for use on a circuit board in which the number of loads are minimized |
US6636212B1 (en) | 2000-11-14 | 2003-10-21 | Nvidia Corporation | Method and apparatus for determining visibility of groups of pixels |
US6778181B1 (en) | 2000-12-07 | 2004-08-17 | Nvidia Corporation | Graphics processing system having a virtual texturing array |
US7027972B1 (en) | 2001-01-24 | 2006-04-11 | Ati Technologies, Inc. | System for collecting and analyzing graphics data and method thereof |
US7358974B2 (en) | 2001-01-29 | 2008-04-15 | Silicon Graphics, Inc. | Method and system for minimizing an amount of data needed to test data against subarea boundaries in spatially composited digital video |
US6888580B2 (en) | 2001-02-27 | 2005-05-03 | Ati Technologies Inc. | Integrated single and dual television tuner having improved fine tuning |
US7130316B2 (en) | 2001-04-11 | 2006-10-31 | Ati Technologies, Inc. | System for frame based audio synchronization and method thereof |
US6542971B1 (en) | 2001-04-23 | 2003-04-01 | Nvidia Corporation | Memory access system and method employing an auxiliary buffer |
US6664960B2 (en) | 2001-05-10 | 2003-12-16 | Ati Technologies Inc. | Apparatus for processing non-planar video graphics primitives and associated method of operation |
US6700583B2 (en) | 2001-05-14 | 2004-03-02 | Ati Technologies, Inc. | Configurable buffer for multipass applications |
WO2002101497A2 (en) | 2001-06-08 | 2002-12-19 | Nvidia Corporation | System, method and computer program product for programmable fragment processing in a graphics pipeline |
US6894687B1 (en) | 2001-06-08 | 2005-05-17 | Nvidia Corporation | System, method and computer program product for vertex attribute aliasing in a graphics pipeline |
US6697064B1 (en) | 2001-06-08 | 2004-02-24 | Nvidia Corporation | System, method and computer program product for matrix tracking during vertex processing in a graphics pipeline |
US6731282B2 (en) * | 2001-06-19 | 2004-05-04 | Intel Corporation | System and method for parallel rendering of images |
JP2003030641A (en) | 2001-07-19 | 2003-01-31 | Nec System Technologies Ltd | Plotting device, parallel plotting method therefor and parallel plotting program |
US7253818B2 (en) | 2001-08-07 | 2007-08-07 | Ati Technologies, Inc. | System for testing multiple devices on a single system and method thereof |
US6828987B2 (en) | 2001-08-07 | 2004-12-07 | Ati Technologies, Inc. | Method and apparatus for processing video and graphics data |
US6778189B1 (en) | 2001-08-24 | 2004-08-17 | Nvidia Corporation | Two-sided stencil testing system and method |
US6744433B1 (en) | 2001-08-31 | 2004-06-01 | Nvidia Corporation | System and method for using and collecting information from a plurality of depth layers |
US6704025B1 (en) | 2001-08-31 | 2004-03-09 | Nvidia Corporation | System and method for dual-depth shadow-mapping |
US6989840B1 (en) | 2001-08-31 | 2006-01-24 | Nvidia Corporation | Order-independent transparency rendering system and method |
US6947047B1 (en) | 2001-09-20 | 2005-09-20 | Nvidia Corporation | Method and system for programmable pipelined graphics processing with branching instructions |
US6938176B1 (en) | 2001-10-05 | 2005-08-30 | Nvidia Corporation | Method and apparatus for power management of graphics processors and subsystems that allow the subsystems to respond to accesses when subsystems are idle |
US7091971B2 (en) | 2001-10-29 | 2006-08-15 | Ati Technologies, Inc. | System, method, and apparatus for multi-level hierarchical Z buffering |
US6999076B2 (en) | 2001-10-29 | 2006-02-14 | Ati Technologies, Inc. | System, method, and apparatus for early culling |
US6677953B1 (en) | 2001-11-08 | 2004-01-13 | Nvidia Corporation | Hardware viewport system and method for use in a graphics pipeline |
US6683614B2 (en) | 2001-12-21 | 2004-01-27 | Hewlett-Packard Development Company, L.P. | System and method for automatically configuring graphics pipelines by tracking a region of interest in a computer graphical display system |
US20030117971A1 (en) | 2001-12-21 | 2003-06-26 | Celoxica Ltd. | System, method, and article of manufacture for profiling an executable hardware model using calls to profiling functions |
US7269125B2 (en) | 2001-12-26 | 2007-09-11 | Xm Satellite Radio, Inc. | Method and apparatus for timing recovery in an OFDM system |
US7012610B2 (en) | 2002-01-04 | 2006-03-14 | Ati Technologies, Inc. | Portable device for providing dual display and method thereof |
US6774895B1 (en) | 2002-02-01 | 2004-08-10 | Nvidia Corporation | System and method for depth clamping in a hardware graphics pipeline |
US6829689B1 (en) | 2002-02-12 | 2004-12-07 | Nvidia Corporation | Method and system for memory access arbitration for minimizing read/write turnaround penalties |
US6947865B1 (en) | 2002-02-15 | 2005-09-20 | Nvidia Corporation | Method and system for dynamic power supply voltage adjustment for a semiconductor integrated circuit device |
US6933943B2 (en) | 2002-02-27 | 2005-08-23 | Hewlett-Packard Development Company, L.P. | Distributed resource architecture and system |
US6700580B2 (en) | 2002-03-01 | 2004-03-02 | Hewlett-Packard Development Company, L.P. | System and method utilizing multiple pipelines to render graphical data |
US6853380B2 (en) | 2002-03-04 | 2005-02-08 | Hewlett-Packard Development Company, L.P. | Graphical display system and method |
US20030171907A1 (en) | 2002-03-06 | 2003-09-11 | Shay Gal-On | Methods and Apparatus for Optimizing Applications on Configurable Processors |
US6919896B2 (en) * | 2002-03-11 | 2005-07-19 | Sony Computer Entertainment Inc. | System and method of optimizing graphics processing |
US7009605B2 (en) | 2002-03-20 | 2006-03-07 | Nvidia Corporation | System, method and computer program product for generating a shader program |
CA2480081C (en) | 2002-03-22 | 2007-06-19 | Michael F. Deering | Scalable high performance 3d graphics |
US20030212735A1 (en) | 2002-05-13 | 2003-11-13 | Nvidia Corporation | Method and apparatus for providing an integrated network of processors |
US20040153778A1 (en) | 2002-06-12 | 2004-08-05 | Ati Technologies, Inc. | Method, system and software for configuring a graphics processing communication mode |
US6980209B1 (en) | 2002-06-14 | 2005-12-27 | Nvidia Corporation | Method and system for scalable, dataflow-based, programmable processing of graphics data |
US6812927B1 (en) | 2002-06-18 | 2004-11-02 | Nvidia Corporation | System and method for avoiding depth clears using a stencil buffer |
US6876362B1 (en) | 2002-07-10 | 2005-04-05 | Nvidia Corporation | Omnidirectional shadow texture mapping |
US6797998B2 (en) | 2002-07-16 | 2004-09-28 | Nvidia Corporation | Multi-configuration GPU interface device |
US6825843B2 (en) | 2002-07-18 | 2004-11-30 | Nvidia Corporation | Method and apparatus for loop and branch instructions in a programmable graphics pipeline |
US6954204B2 (en) | 2002-07-18 | 2005-10-11 | Nvidia Corporation | Programmable graphics system and method using flexible, high-precision data formats |
US6864893B2 (en) | 2002-07-19 | 2005-03-08 | Nvidia Corporation | Method and apparatus for modifying depth values using pixel programs |
US6952206B1 (en) | 2002-08-12 | 2005-10-04 | Nvidia Corporation | Graphics application program interface system and method for accelerating graphics processing |
US7112884B2 (en) | 2002-08-23 | 2006-09-26 | Ati Technologies, Inc. | Integrated circuit having memory disposed thereon and method of making thereof |
US6779069B1 (en) | 2002-09-04 | 2004-08-17 | Nvidia Corporation | Computer system with source-synchronous digital link |
US7324547B1 (en) | 2002-12-13 | 2008-01-29 | Nvidia Corporation | Internet protocol (IP) router residing in a processor chipset |
US6885376B2 (en) | 2002-12-30 | 2005-04-26 | Silicon Graphics, Inc. | System, method, and computer program product for near-real time load balancing across multiple rendering pipelines |
US7233964B2 (en) | 2003-01-28 | 2007-06-19 | Lucid Information Technology Ltd. | Method and system for compositing three-dimensional graphics images using associative decision mechanism |
US7062272B2 (en) * | 2003-02-18 | 2006-06-13 | Qualcomm Incorporated | Method and apparatus to track count of broadcast content recipients in a wireless telephone network |
US7145565B2 (en) | 2003-02-27 | 2006-12-05 | Nvidia Corporation | Depth bounds testing |
US6911983B2 (en) | 2003-03-12 | 2005-06-28 | Nvidia Corporation | Double-buffering of pixel data using copy-on-write semantics |
US7129909B1 (en) | 2003-04-09 | 2006-10-31 | Nvidia Corporation | Method and system using compressed display mode list |
US6900810B1 (en) | 2003-04-10 | 2005-05-31 | Nvidia Corporation | User programmable geometry engine |
US6940515B1 (en) | 2003-04-10 | 2005-09-06 | Nvidia Corporation | User programmable primitive engine |
US7068278B1 (en) | 2003-04-17 | 2006-06-27 | Nvidia Corporation | Synchronized graphics processing units |
US7483031B2 (en) | 2003-04-17 | 2009-01-27 | Nvidia Corporation | Method for synchronizing graphics processing units |
US7120816B2 (en) | 2003-04-17 | 2006-10-10 | Nvidia Corporation | Method for testing synchronization and connection status of a graphics processing unit module |
US7038678B2 (en) | 2003-05-21 | 2006-05-02 | Nvidia Corporation | Dependent texture shadow antialiasing |
US7038685B1 (en) | 2003-06-30 | 2006-05-02 | Nvidia Corporation | Programmable graphics processor for multithreaded execution of programs |
US7119808B2 (en) | 2003-07-15 | 2006-10-10 | Alienware Labs Corp. | Multiple parallel processor computer graphics system |
US6995767B1 (en) | 2003-07-31 | 2006-02-07 | Nvidia Corporation | Trilinear optimization for texture filtering |
US7525547B1 (en) | 2003-08-12 | 2009-04-28 | Nvidia Corporation | Programming multiple chips from a command buffer to process multiple images |
US7015915B1 (en) | 2003-08-12 | 2006-03-21 | Nvidia Corporation | Programming multiple chips from a command buffer |
US7075541B2 (en) | 2003-08-18 | 2006-07-11 | Nvidia Corporation | Adaptive load balancing in a multi-processor graphics processing system |
US6956579B1 (en) | 2003-08-18 | 2005-10-18 | Nvidia Corporation | Private addressing in a multi-processor graphics processing system |
US7388581B1 (en) | 2003-08-28 | 2008-06-17 | Nvidia Corporation | Asynchronous conditional graphics rendering |
US8250412B2 (en) | 2003-09-26 | 2012-08-21 | Ati Technologies Ulc | Method and apparatus for monitoring and resetting a co-processor |
US7782325B2 (en) * | 2003-10-22 | 2010-08-24 | Alienware Labs Corporation | Motherboard for supporting multiple graphics cards |
US8274517B2 (en) | 2003-11-14 | 2012-09-25 | Microsoft Corporation | Systems and methods for downloading algorithmic elements to a coprocessor and corresponding techniques |
US8085273B2 (en) | 2003-11-19 | 2011-12-27 | Lucid Information Technology, Ltd | Multi-mode parallel graphics rendering system employing real-time automatic scene profiling and mode control |
US7961194B2 (en) * | 2003-11-19 | 2011-06-14 | Lucid Information Technology, Ltd. | Method of controlling in real time the switching of modes of parallel operation of a multi-mode parallel graphics processing subsystem embodied within a host computing system |
US7808499B2 (en) | 2003-11-19 | 2010-10-05 | Lucid Information Technology, Ltd. | PC-based computing system employing parallelized graphics processing units (GPUS) interfaced with the central processing unit (CPU) using a PC bus and a hardware graphics hub having a router |
US20080079737A1 (en) | 2003-11-19 | 2008-04-03 | Reuven Bakalash | Multi-mode parallel graphics rendering and display system supporting real-time detection of mode control commands (MCCS) programmed within pre-profiled scenes of the graphics-based application |
WO2006117683A2 (en) | 2005-01-25 | 2006-11-09 | Lucid Information Technology, Ltd. | Graphics processing and display system employing multiple graphics cores on a silicon chip of monolithic construction |
US7015914B1 (en) | 2003-12-10 | 2006-03-21 | Nvidia Corporation | Multiple data buffers for processing graphics data |
US7053901B2 (en) | 2003-12-11 | 2006-05-30 | Nvidia Corporation | System and method for accelerating a special purpose processor |
US7248261B1 (en) | 2003-12-15 | 2007-07-24 | Nvidia Corporation | Method and apparatus to accelerate rendering of shadow effects for computer-generated images |
JP3879002B2 (en) | 2003-12-26 | 2007-02-07 | 国立大学法人宇都宮大学 | Self-optimizing arithmetic unit |
US6975325B2 (en) | 2004-01-23 | 2005-12-13 | Ati Technologies Inc. | Method and apparatus for graphics processing using state and shader management |
US7259606B2 (en) | 2004-01-27 | 2007-08-21 | Nvidia Corporation | Data sampling clock edge placement training for high speed GPU-memory interface |
US7483034B2 (en) | 2004-02-25 | 2009-01-27 | Siemens Medical Solutions Usa, Inc. | System and method for GPU-based 3D nonrigid registration |
US7289125B2 (en) | 2004-02-27 | 2007-10-30 | Nvidia Corporation | Graphics device clustering with PCI-express |
US7027062B2 (en) | 2004-02-27 | 2006-04-11 | Nvidia Corporation | Register based queuing for texture requests |
US20050195186A1 (en) | 2004-03-02 | 2005-09-08 | Ati Technologies Inc. | Method and apparatus for object based visibility culling |
US7978194B2 (en) | 2004-03-02 | 2011-07-12 | Ati Technologies Ulc | Method and apparatus for hierarchical Z buffering and stenciling |
US20050275760A1 (en) | 2004-03-02 | 2005-12-15 | Nvidia Corporation | Modifying a rasterized surface, such as by trimming |
US7315912B2 (en) | 2004-04-01 | 2008-01-01 | Nvidia Corporation | Deadlock avoidance in a bus fabric |
US7336284B2 (en) | 2004-04-08 | 2008-02-26 | Ati Technologies Inc. | Two level cache memory architecture |
US7265759B2 (en) | 2004-04-09 | 2007-09-04 | Nvidia Corporation | Field changeable rendering system for a computing device |
US6985152B2 (en) | 2004-04-23 | 2006-01-10 | Nvidia Corporation | Point-to-point bus bridging without a bridge controller |
US20050237329A1 (en) | 2004-04-27 | 2005-10-27 | Nvidia Corporation | GPU rendering to system memory |
US7738045B2 (en) | 2004-05-03 | 2010-06-15 | Broadcom Corporation | Film-mode (3:2/2:2 Pulldown) detector, method and video device |
US7079156B1 (en) | 2004-05-14 | 2006-07-18 | Nvidia Corporation | Method and system for implementing multiple high precision and low precision interpolators for a graphics pipeline |
US7426724B2 (en) | 2004-07-02 | 2008-09-16 | Nvidia Corporation | Optimized chaining of vertex and fragment programs |
US7218291B2 (en) | 2004-09-13 | 2007-05-15 | Nvidia Corporation | Increased scalability in the fragment shading pipeline |
US7868891B2 (en) | 2004-09-16 | 2011-01-11 | Nvidia Corporation | Load balancing |
US7721118B1 (en) * | 2004-09-27 | 2010-05-18 | Nvidia Corporation | Optimizing power and performance for multi-processor graphics processing |
US7571296B2 (en) | 2004-11-11 | 2009-08-04 | Nvidia Corporation | Memory controller-adaptive 1T/2T timing control |
US7598958B1 (en) | 2004-11-17 | 2009-10-06 | Nvidia Corporation | Multi-chip graphics processing unit apparatus, system, and method |
US8066515B2 (en) | 2004-11-17 | 2011-11-29 | Nvidia Corporation | Multiple graphics adapter connection systems |
US7477256B1 (en) | 2004-11-17 | 2009-01-13 | Nvidia Corporation | Connecting graphics adapters for scalable performance |
US7451259B2 (en) | 2004-12-06 | 2008-11-11 | Nvidia Corporation | Method and apparatus for providing peer-to-peer data transfer within a computing environment |
US7275123B2 (en) | 2004-12-06 | 2007-09-25 | Nvidia Corporation | Method and apparatus for providing peer-to-peer data transfer within a computing environment |
US7545380B1 (en) * | 2004-12-16 | 2009-06-09 | Nvidia Corporation | Sequencing of displayed images for alternate frame rendering in a multi-processor graphics system |
US7372465B1 (en) | 2004-12-17 | 2008-05-13 | Nvidia Corporation | Scalable graphics processing for remote display |
US20060156399A1 (en) | 2004-12-30 | 2006-07-13 | Parmar Pankaj N | System and method for implementing network security using a sequestered partition |
US7924281B2 (en) | 2005-03-09 | 2011-04-12 | Ati Technologies Ulc | System and method for determining illumination of a pixel by shadow planes |
US7796095B2 (en) | 2005-03-18 | 2010-09-14 | Ati Technologies Ulc | Display specific image processing in an integrated circuit |
US7568056B2 (en) | 2005-03-28 | 2009-07-28 | Nvidia Corporation | Host bus adapter that interfaces with host computer bus to multiple types of storage devices |
US7681187B2 (en) | 2005-03-31 | 2010-03-16 | Nvidia Corporation | Method and apparatus for register allocation in presence of hardware constraints |
US7616207B1 (en) | 2005-04-25 | 2009-11-10 | Nvidia Corporation | Graphics processing system including at least three bus devices |
US7545101B2 (en) * | 2005-05-05 | 2009-06-09 | Leviton Manufacturing Co., Inc. | Multi-zone closed loop daylight harvesting having at least one light sensor |
US20080143731A1 (en) | 2005-05-24 | 2008-06-19 | Jeffrey Cheng | Video rendering across a high speed peripheral interconnect bus |
US7817155B2 (en) | 2005-05-24 | 2010-10-19 | Ati Technologies Inc. | Master/slave graphics adapter arrangement |
US7613346B2 (en) | 2005-05-27 | 2009-11-03 | Ati Technologies, Inc. | Compositing in multiple video processing unit (VPU) systems |
US7539801B2 (en) | 2005-05-27 | 2009-05-26 | Ati Technologies Ulc | Computing device with flexibly configurable expansion slots, and method of operation |
US20060282604A1 (en) | 2005-05-27 | 2006-12-14 | Ati Technologies, Inc. | Methods and apparatus for processing graphics data using multiple processing circuits |
US7325086B2 (en) | 2005-12-15 | 2008-01-29 | Via Technologies, Inc. | Method and system for multiple GPU support |
US7728841B1 (en) | 2005-12-19 | 2010-06-01 | Nvidia Corporation | Coherent shader output for multiple targets |
US7768517B2 (en) | 2006-02-21 | 2010-08-03 | Nvidia Corporation | Asymmetric multi-GPU processing |
US8284204B2 (en) | 2006-06-30 | 2012-10-09 | Nokia Corporation | Apparatus, method and a computer program product for providing a unified graphics pipeline for stereoscopic rendering |
-
2004
- 2004-11-19 US US10/579,682 patent/US7808499B2/en active Active
- 2004-11-19 CA CA002546427A patent/CA2546427A1/en not_active Abandoned
- 2004-11-19 EP EP04799376A patent/EP1687732A4/en not_active Withdrawn
- 2004-11-19 JP JP2006540774A patent/JP2007512613A/en active Pending
- 2004-11-19 CN CNA2004800343078A patent/CN1890660A/en active Pending
- 2004-11-19 WO PCT/IL2004/001069 patent/WO2005050557A2/en active Application Filing
-
2007
- 2007-10-23 US US11/977,164 patent/US7800619B2/en active Active
- 2007-10-23 US US11/977,161 patent/US7796129B2/en active Active
- 2007-10-23 US US11/977,172 patent/US7800610B2/en active Active
- 2007-10-23 US US11/977,155 patent/US8629877B2/en active Active
- 2007-10-23 US US11/977,178 patent/US7800611B2/en active Active
- 2007-10-23 US US11/977,177 patent/US7796130B2/en active Active
-
2011
- 2011-05-05 US US13/101,438 patent/US20110279462A1/en not_active Abandoned
-
2014
- 2014-01-13 US US14/153,105 patent/US9405586B2/en active Active
Also Published As
Publication number | Publication date |
---|---|
US7796130B2 (en) | 2010-09-14 |
US7808499B2 (en) | 2010-10-05 |
US20080136825A1 (en) | 2008-06-12 |
US20080238917A1 (en) | 2008-10-02 |
US20110279462A1 (en) | 2011-11-17 |
US7800610B2 (en) | 2010-09-21 |
WO2005050557A3 (en) | 2005-09-01 |
US9405586B2 (en) | 2016-08-02 |
US20080165184A1 (en) | 2008-07-10 |
US7800619B2 (en) | 2010-09-21 |
US20070279411A1 (en) | 2007-12-06 |
US8629877B2 (en) | 2014-01-14 |
CN1890660A (en) | 2007-01-03 |
US20080165196A1 (en) | 2008-07-10 |
US20080165197A1 (en) | 2008-07-10 |
US7796129B2 (en) | 2010-09-14 |
EP1687732A4 (en) | 2008-11-19 |
US20140125682A1 (en) | 2014-05-08 |
WO2005050557A2 (en) | 2005-06-02 |
EP1687732A2 (en) | 2006-08-09 |
US20080165198A1 (en) | 2008-07-10 |
JP2007512613A (en) | 2007-05-17 |
US7800611B2 (en) | 2010-09-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7808499B2 (en) | PC-based computing system employing parallelized graphics processing units (GPUS) interfaced with the central processing unit (CPU) using a PC bus and a hardware graphics hub having a router | |
US7812845B2 (en) | PC-based computing system employing a silicon chip implementing parallelized GPU-driven pipelines cores supporting multiple modes of parallelization dynamically controlled while running a graphics application | |
US9584592B2 (en) | Internet-based graphics application profile management system for updating graphic application profiles stored within the multi-GPU graphics rendering subsystems of client machines running graphics-based applications | |
US20080088631A1 (en) | Multi-mode parallel graphics rendering and display system supporting real-time detection of scene profile indices programmed within pre-profiled scenes of the graphics-based application | |
US20090096798A1 (en) | Graphics Processing and Display System Employing Multiple Graphics Cores on a Silicon Chip of Monolithic Construction |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
EEER | Examination request | ||
FZDE | Dead |
Effective date: 20130722 |