Computer graphics processing and selective visual display system – Computer graphics display memory system – Cache
Reexamination Certificate
1999-12-30
2004-01-06
Tung, Kee M. (Department: 2676)
Computer graphics processing and selective visual display system
Computer graphics display memory system
Cache
C345S543000, C345S531000, C711S129000
Reexamination Certificate
active
06674443
ABSTRACT:
BACKGROUND OF THE PRESENT INVENTION
1. Field of the Invention
The present invention relates generally to a method and apparatus for enhancing performance of graphics operations. More particularly, the invention relates to two-dimensional graphics operations.
2. Description of the Related Art
For many graphics operations such as line drawing, polygon rendering, etc., one of the main bottlenecks for speed performance is the bandwidth of the frame buffer. The bandwidth of the frame buffer is even more critical for system on a chip (SOC) devices that use unified memory architecture (UMA). In UMA systems, the same memory is shared between many agents, such as a central processing unit (CPU), a graphics accelerator, an MPEG decoder, and a CRT controller. Because there are several agents sharing the same memory, a large access latency for the memory is created. The typical access latency for the memory is fifteen clock cycles or more.
Memory systems implemented with synchronous dynamic random access memory (SDRAM) have advantage that if an agent requests a large chunk of sequential pixel data, then the first pixel data latency may be the same as normal DRAM memory, but subsequent pixel data can be transferred every clock cycle (provided the pixel data is located in the same page of memory). Unfortunately, graphics accelerators cannot use this feature directly for most drawing operations, including line drawing where the access of pixel data is not very sequential.
FIG. 1
shows a conventional UMA System
100
where an SDRAM controller
110
is electrically coupled between an SDRAM memory
120
A and several agents, such as a CPU
130
, an MPEG decoder
140
, a graphics engine
150
, and a CRT controller
160
. A frame buffer
125
comprises the combination of the SDRAM controller
110
and part of the SDRAM memory
120
A. As suggested by the UMA system
100
, each agent operating in a graphics system communicates to the SDRAM
120
A via the SDRAM controller
110
. Because each agent communicates directly to the SDRAM controller
110
, each agent must operate with a memory configuration as defined by the frame buffer
125
, whether or not each agent is optimized for the memory configuration within the frame buffer
125
.
It should be understood that the frame buffer
125
is a memory for storing pixel data for a visual display. To manipulate an image on the visual display, pixel data may be written to the frame buffer
125
by, for instance, the CPU
130
, and read by the CRT controller
160
to display the pixel data on the visual display. For instance, to make the visual display completely white, each memory element within the frame buffer
125
may be written pixel data corresponding to the color white. Also, if graphical elements, such as a triangle or rectangle, is to be drawn, the graphics engine
150
may write pixel data to the frame buffer
125
to create or manipulate different graphical elements.
Most operations in windows type programs are rectangles. Because rectangles are highly sequential elements (e.g., pixels
10
,
11
, . . . ,
20
) for each scan line, standard scan line configured memories are suitable for standard windows operations. However, in systems that utilize non-horizontal or non-vertical lines, such as computer aided design (CAD) systems and map display systems, access to pixel data is not sequential in the frame buffer
125
, thereby causing a problem or degrading system performance.
The system performance issues are related to the frame buffer
125
generally being implemented using SDRAM memory. SDRAM memory typically has a certain latency or access time to access a first memory location (e.g., pixel
10
), but not much time to access the next memory location (e.g., pixel
11
). While this is adequate for forming straight lines (i.e., horizontal or vertical), non-straight lines (i.e., non-horizontal or non-vertical) are much slower to create or manipulate due to the memory configuration within the frame buffer
125
being arranged in a scan line format.
FIG. 2
represents a memory organization for an SDRAM
120
B having a scan line configuration for the pixel data displayed on a visual display, as is conventionally known in the art. For instance, SDRAM
120
B represents a visual display having a 768 row by 1024 column pixel image, whereby a first scan line has pixel P
1,1
200
in the upper left corner and pixel P
1,1024
210
in the upper right corner and a last scan line having pixel P
763,1
220
in the lower left corner and pixel P
768,1024
230
in the lower right corner.
FIG. 2
represents a conventional scan line configured frame buffer
125
that agents, such as the CRT controller
160
, are optimized for displaying pixel data from the frame buffer
125
to a visual display. However, the graphics engine
150
performance is significantly reduced (i.e., clock cycles increased) by the conventional scan line configuration in the frame buffer
125
.
Access patterns of many graphics operations that include line drawing show a high degree of two-dimensional locality of reference. One approach to improve memory bandwidth for such access patterns has been to use a tiled memory configuration within the frame buffer
125
. In this tiled memory configuration, pixel data in the frame buffer
125
is stored in a tiled configuration rather than the conventional configuration (i.e., sequentially one scan line after another (assuming one tile of pixels fits into one DRAM page)). Because of the two-dimensional locality of reference, chances of consecutive accesses falling inside a single tile is very high. As long as the access from the graphics operation falls within a single tile, there will be no page miss for the tiled memory configuration within the frame buffer
125
memory, compared with the conventional scan line or non-tiled memory configuration where a tile is scattered over multiple pages of DRAM.
FIG. 3
represents a frame buffer
120
C configured into tiles of fixed size. A first tile T
0
300
is indicated to be in the upper left corner and a last tile T
383
310
is indicated to be in the lower right corner of the frame buffer
120
C. While the frame buffer
125
having fixed tile sizes greatly improves frame buffer access time for graphics operations for the graphics engine
150
, fixed tiling the frame buffer
125
is not suitable for UMA systems where a number of other agents expecting the frame buffer
125
to be in the conventional scan line configuration access the frame buffer
125
. It is important to understand that performance of certain agents, such as the CRT controller
160
, is critical to a real-time system because a user sees visual effects on the visual display if the performance of the CRT controller
160
is degraded. By having the frame buffer
125
configured in a tiled fashion a degradation in the performance of the CRT controller
160
occurs as the CRT controller
160
is optimized for the conventional scan line configuration.
Additionally, there are many existing software products, including many popular video games, which access the frame buffer
125
directly. Since these software programs are not aware of tiled configurations within the frame buffer
125
, their access to the frame buffer
125
will not work properly and, consequently, these software programs cannot be run on a system with a tiled frame buffer
125
unless a separate address translator interface is added to translate scan line addresses to tiled addresses. Tiling the frame buffer
125
also has additional overhead for the CPU
130
because of the need to convert many of the windows pixel data structures, such as BitMap images, into a tiled format so that the BitMaps images can be used for graphics operations with the frame buffer
125
having a tiled configuration.
SUMMARY OF THE INVENTION
The present invention relates to a system and method for accelerating graphics operations. The system includes a memory device for accelerating graphics operations within an electronic device. A memory controller is used for controlling pixel data transmitted to and f
Banga Kanwal Preet Singh
Chowdhuri Bhaskar
Palazzolo, Jr. Frank
Zampieri Ugo
Jorgenson Lisa K.
STMicroelectronics Inc.
Szuwalski Andre
Tung Kee M.
LandOfFree
Memory system for accelerating graphics operations within an... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Memory system for accelerating graphics operations within an..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Memory system for accelerating graphics operations within an... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3192470