1. Field of the Invention
Embodiments of the present invention relate generally to graphics system architectures and more specifically to generating multiple simultaneous unique outputs from a single display pipeline.
2. Description of the Related Art
A typical computer system includes, without limitation, a central processing unit (CPU), a graphics processing unit (GPU), at least one display device, and one or more input devices, such as a keyboard and a mouse. The display device generates a sequence of images from sequential video frames, each composed from a structured stream of pixels. The stream of pixels is typically generated by a video pipeline within the GPU. The video pipeline performs pixel processing operations, known to persons skilled in the art, which enable proper display of video images. For example, the video pipeline may perform “gamma correction” on a pixel in order to properly account for certain non-linear characteristics of an attached display device. A display device is frequently called a “head.” Consequently, the circuitry within the GPU used to support the display device is also referred to as a “head.”
When a user attaches one or more additional display devices to the computer system to view independent images, additional video pipelines are needed to generate the independent images. Modern GPU devices typically include two independent video pipelines, allowing the user to attach one or two display devices to the GPU. Each GPU also commonly includes numerous digital and analog video output ports used to support various video standards. However, despite the availability of these video output ports included in the GPU device, the GPU is still limited to driving two display devices (or the number of display devices equal to the number of available independent video pipelines within the GPU).
One approach to increasing the overall number of display devices that may be attached to a computer system is to configure the computer system to include additional GPUs. These additional GPUs provide additional video pipelines that can drive the independent display devices. One popular computer system configuration includes four similar or identical display devices. This configuration requires that two GPUs be present within the computer system, where each GPU includes two video pipelines. While many applications that require four display devices do not require the additional graphics processing capability of an additional GPU, the second GPU is nonetheless required in order to provide the necessary video pipelines used to drive the additional display devices. The second GPU increases the cost of the computer system and increases the amount of power needed to operate the computer system.
Another approach to increasing the number of display devices that may be attached to a computer system is to include additional independent video pipelines within the GPU, thereby allowing the GPU to drive additional independent display devices. However, the circuitry needed to implement a video pipeline is expensive and would add to the overall cost of the GPU, which should be optimized for the common usage case of driving one or two display devices.
As the foregoing illustrates, what is needed in the art is a GPU architecture that accommodates additional display devices without the expense of including additional display pipelines within the GPU.
One embodiment of the present invention sets forth a horizontal scaling filter included in a video pipeline configured to process video data. The horizontal scaling filter includes a first set of storage elements, where each storage element is configured to receive pixel data in a first clock cycle and to transmit the pixel data in a second clock cycle to produce delayed pixel data, a plurality of multipliers configured to generate weighted pixel data from the delayed pixel data, and a summation unit configured to sum the weighted pixel data to produce output pixel data.
One advantage of the disclosed system is that, by interleaving pixels from different images, the interleaved pixels for multiple heads may be processed within the same video pipeline, thereby achieving greater overall efficiency within the host GPU. Further, the horizontal scaling filter within the video pipeline advantageously enables video processing operations to be performed on either non-interleaved or interleaved pixel data.
So that the manner in which the above recited features of the present invention can be understood in detail, a more particular description of the invention, briefly summarized above, may be had by reference to embodiments, some of which are illustrated in the appended drawings. It is to be noted, however, that the appended drawings illustrate only typical embodiments of this invention and are therefore not to be considered limiting of its scope, for the invention may admit to other equally effective embodiments.
The CPU 110 processes programming instructions stored within system memory 140 and coordinates the activities of the I/O bridge 112 and GPU 120. The I/O bridge 112 enables the CPU 110 to interface to peripheral devices 118 via the peripheral bus 116. Peripheral devices 118 may include, without limitation, a hard disk drive adapter and a network adapter. The peripheral bus 116 may be a Peripheral Component Interconnect (PCI) type bus or any other appropriate type of system bus. The I/O bridge 112 also receives input from the keyboard 115 and mouse 114 for processing by the CPU 110.
The CPU 110 directs the GPU 120 to generate images 152 and store the images in the local memory 150 for display on display devices 160. Each image 152-0 through 152-3 may be displayed on a corresponding display device. For example, image 152-0 may be displayed on display device 160-0 via the GPU 120.
The GPU 120 may include a graphics processing engine 122, video output encoders 130, a cross bar 128, video pipelines 126, and multiplexers 124. The graphics processing engine 122 may include a plurality of processing elements, such as fixed function and programmable processing cores.
The video output encoders 130 receive pixel data and timing data represented in digital form and generate a set of output signals 132 that conform to a selected video signal standard. For example, the video output encoder 130 may generate analog video signals consistent with the video graphics adapter (VGA) industry standard. Alternately, the video output encoder 130 may generate a set of serial high-speed digital bit streams consistent with the digital video interface (DVI) industry standard. A given video output encoder 130 is typically configured to produce only analog or only digital signals, therefore, a GPU 120 may include enough video output encoders 130 to accommodate the various industry standards for video signals. In one embodiment, the video output encoders 130 incorporate a de-multiplexer circuit (not shown) configured to select certain data presented to the video output encoder 130 for display, while discarding other data not intended for the given video output encoder 130.
The cross bar 128 routes data from video pipelines 126 to video encoders 130. For example, the cross bar 128 may be configured to route data from video pipeline 126-0 to video encoder 130-1 and data from video pipeline 126-1 to video encoder 130-0. Or the cross bar 128 may be configured to route the same data from video pipeline 126-0 to video output encoder 130-0 and also video output encoder 130-1. In this second scenario, pixel data for both video output encoder 130-0 and 130-1 are interleaved in the same data stream. The video output encoder 130-0 selects only data, referred to as “selected display data,” that is intended for display on display device 160-0. The video output encoder 130-0 discards data intended for display device 160-1. Similarly, video output encoder 130-1 extracts selected display data intended for display on display device 160-1, while discarding data intended for display device 160-0.
The video pipelines 126 receive interleaved pixel data from multiplexers 124 and generate processed interleaved pixel data. For example, the video pipelines 126 may perform a gamma correction on the interleaved pixel data to account for non-linearity exhibited by the display devices 160. The processed interleaved pixel data is de-multiplexed by video output encoders 130 and used for display on the display devices 160. The multiplexers 124 receive pixel data from one or more images 152 stored within local memory 150 to generate the interleaved pixel data. The interleaved pixel data may include pixels for a single head, or interleaved pixels for multiple heads.
The crossbar 128 replicates and routes the processed interleaved pixel data 224 to two or more video output encoders 130. For example, the processed interleaved pixel data 224 may be routed to video output encoder 130-0 (via data stream 240), as well as to video output encoder 130-1 (via data stream 242). In this scenario, pixels 202 are processed by video pipeline 126-0 and routed, along with pixels 203, to video output encoder 130-0. Video output encoder 130-0 selects only data related to pixels 202 to generate a video output signal 132-0. Similarly, pixels 203 are processed by video pipeline 126-0 and routed, along with pixels 202, to video output encoder 130-1. Video output encoder 130-1 selects only data related to pixels 203 to generate a video output signal 132-1.
In a similar way, pixels 204 from image 152-2 are processed by video pipeline 126-1 in order to generate video output signal 132-4 and pixels 205 from image 152-3 are processed by video pipeline 126-1 in order to generate video output signal 132-3.
By interleaving pixel data processed by the video pipelines 126, logical heads associated with a specific video pipeline 126-0 or 126-1 are available to drive multiple display devices. These logical heads are also referred to as “sub-heads.” While the video pipelines 126 may perform a substantial portion of pixel processing without needing to know which pixel is associated with which sub-head, cursor logic 270 within the video pipelines 126 should be aware of which sub-head is associated with a given pixel being processed in order to properly generate a cursor for display on the appropriate sub-head. The cursor logic 270 may examine a tag associated with a pixel being processed by the video pipeline 126 to determine if the cursor should be visible on the target sub-head of the pixel. If the cursor should be visible on the target sub-head and the cursor intersects the pixel in screen space, then the cursor logic 270 may perform an overlay operation on the pixel being processed. If the cursor should not be visible on the target sub-head, then the cursor logic 270 does not perform an overlay operation on the pixel being processed.
Persons skilled in the art will recognize that there are many known techniques for tagging pixel data with sub-head information to properly inform the cursor logic 270 of an association between pixel data and a target sub-head.
Image 350 includes pixels 302, which are transmitted to an interleaving function 330 as pixel data 310. Image 352 includes pixels 303, which are transmitted to an interleaving function 330 as pixel data 312. The interleaving function 330 generates the single source image 354 by alternating pixels from pixel data 310 and pixel data 312. The interleaving function 330 may be implemented using software, dedicated hardware, or any combination thereof.
Data within the single source image 354 is transmitted to the video pipeline 126-0 as interleaved pixel data 320, which includes pixels from each source image 350 and 352. The interleaved pixel data 320 is substantially identical in form to interleaved pixel data 220 of
Each storage element 450-456 may store one or more pixel attributes used for scaling. Such pixel attributes may include, for example, red, green and blue intensity values associated with a pixel. During normal operation, input pixel data 410 is captured by storage element 450 and stored as delayed data 412. Similarly, delayed data 412 is captured by storage element 452 and stored as delayed data 414, and so forth. Multiplier 440 receives delayed data 412 and an associated weight 432 and generates weighted pixel data, which is transmitted to the summation unit 448. Similarly, multipliers 442 through 446 generate weighted pixel data from delayed data 414 through 418 and weights 434 through 438. The weighted pixel data is transmitted to the summation unit 448, which adds the weighted pixel data to produce a weighted sum value over the weighted pixel data. The weighted sum value is transmitted from the horizontal scaling filter 400 as output pixel data 460. Pixel data at each location in the horizontal scaling filter 400 is marked “a,” indicating the pixel data is associated with the same sub-head. Persons skilled in the art will recognize that there are many possible techniques for implementing the horizontal scaling filter 400 in the context of the video pipelines 126.
As each element of input pixel data 510 is received, storage elements 550, 552, 554, and 556 alternately store delayed data associated with the first sub-head and then the second sub-head. Similarly, storage elements 551, 553, 555, and 557 alternately store delayed data associated with the second sub-head and then the first sub-head. After a given clock cycle, the storage elements 550, 552, 554, and 556 either store data associated with the first sub-head or the second sub-head. Additionally, storage elements 550, 552, 554, and 556 transmit delayed data to storage elements 551, 553, 555, and 557. After a given clock cycle, the storage elements 551, 553, 555, and 557 either store data associated with the second sub-head or the first sub-head. Additionally, storage elements 551, 553, 555, and 557 transmit delayed data to multipliers 540 through 546 for processing.
In the clock cycle shown, storage elements 550, 552, 554, and 556 store delayed data 570, 572, 574, and 576, which is associated with the second sub-head (again, indicated with a marking of “b”). At the same time, storage elements 551, 553, 555, and 557 store delayed data 571, 573, 575, and 577, which is associated with the first sub-head (again, indicated with a marking of “a”). During the clock cycle shown, multipliers 540 through 546 receive delayed data 571, 573, 575 and 577, along with weights 532, 534, 536, and 538 and compute weighted pixel data associated with the first sub-head. In one embodiment data associated with the first sub-head is processed by multipliers 540 through 546 and summation unit 548 in the same clock cycle. In alternate embodiments, different pipelining schemes may be used with multipliers 540 through 546 and summation unit 548.
A mode signal 612 determines if the horizontal scaling filter 600 should operate in a non-interleaved mode or an interleaved mode by driving the select input of multiplexers 661, 663, 665, and 667. For example, to operate in non-interleaved mode, multiplexer 661 passes delayed data 680 from storage element 650 to storage element 652. Similarly, multiplexer 663 passes delayed data 682 from storage element 652 to storage element 654, and so forth. In interleaved mode, multiplexer 661 passes delayed data 681 from storage element 651 to storage element 652. Similarly, multiplexer 663 passes delayed data 683 from storage element 653 to storage element 654, and so forth. In a given clock cycle, multiplexers 661 through 667 pass delayed data associated with one sub-head. In non-interleaved mode, the delayed data passed by multiplexers 661 through 667 is associated with one sub-head over successive clock cycles. In interleaved mode, the delayed data passed by multiplexers 661 through 667 is associated with alternating sub-heads over successive clock cycles.
Multipliers 640 through 646 receive the outputs of multiplexers 661 through 667 along with weights 632 through 638. The multipliers 640 through 646 generate weighted pixel data, which is transmitted to summation unit 648. The summation unit adds the weighted pixel data to produce a weighted sum value over the weighted pixel data. The weighted sum value is transmitted from the horizontal scaling filter 600 as output pixel data 690. When operating in non-interleaved mode, the output pixel data 690 is associated with one sub-head. When operating in interleaved mode, the output pixel data 690 includes data associated with alternating sub-heads. For example, the first data element transmitted as output pixel data 690 may be associated with the first sub-head, and the second data element transmitted as output pixel data 690 may be associated with the second sub-head, and so forth.
In sum, a system is presented for generating multiple video output signals, which may be horizontally scaled, from a single video pipeline within a graphics processing unit. Pixels are interleaved from multiple video sub-heads before being transmitted to the video pipeline for processing. Horizontal scaling is performed by a horizontal scaling filter within the video pipeline. The horizontal scaling filter includes delay elements configured to preserve the sub-head association of the interleaved pixels entering and exiting the horizontal scaling filter.
A cross bar transmits the processed pixel data to video output encoders, which may receive a combination of pixels generated by the video pipeline. Each video output encoder selects which pixels should be transmitted to the respective display device, while discarding pixels destined to different display devices. By multiplexing the use of video pipeline resources to process pixels for multiple heads, greater overall efficiency is achieved within the host GPU. In particular, a single GPU with two independent video pipelines may be configured to support four display devices with negligible additional cost.
While the forgoing is directed to embodiments of the present invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof. For example, aspects of the present invention may be implemented in hardware or software or in a combination of hardware and software. Therefore, the scope of the present invention is determined by the claims that follow.
The current application is a continuation-in-part of the U.S. patent application titled, “Multiple Simultaneous Unique Outputs From A Single Display Pipeline,” filed on Nov. 6, 2007 and having patent application Ser. No. 11/936,073 (NVDA/P002042). This related application is incorporated herein by reference in its entirety.
Number | Name | Date | Kind |
---|---|---|---|
5241372 | Ohba | Aug 1993 | A |
6931660 | Kalluri et al. | Aug 2005 | B1 |
20060117360 | Cooper et al. | Jun 2006 | A1 |
20060290641 | Ku et al. | Dec 2006 | A1 |
Number | Date | Country | |
---|---|---|---|
Parent | 11936073 | Nov 2007 | US |
Child | 11955307 | US |