CN102262603B - Advanced graph memory - Google Patents
Advanced graph memory Download PDFInfo
- Publication number
- CN102262603B CN102262603B CN201110185455.1A CN201110185455A CN102262603B CN 102262603 B CN102262603 B CN 102262603B CN 201110185455 A CN201110185455 A CN 201110185455A CN 102262603 B CN102262603 B CN 102262603B
- Authority
- CN
- China
- Prior art keywords
- memory
- data
- impact damper
- graphic
- module
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 claims description 21
- 230000005540 biological transmission Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 abstract description 8
- 230000003139 buffering effect Effects 0.000 abstract 4
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 230000003213 activating effect Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000004040 coloring Methods 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
Landscapes
- Image Generation (AREA)
- Controls And Circuits For Display Device (AREA)
Abstract
The invention provides an advanced graph memory. The advanced graph memory comprises an intelligent graph processing buffer, an internal memory controller and a memory chip. The intelligent graph processing buffer is connected with the internal memory controller and the memory chip. The intelligent graph processing buffer comprises a whole situation control module, a data selecting module, a buffering module and a logical operation unit; the memory chip comprises a graph memory and a pixel memory; the whole situation control module is connected with the internal memory controller, the data selecting module, the buffering module and the logical operation unit; the logical operation unit is connected with the buffering unit and the data selecting module; the data selecting module is connected with the internal memory controller, the buffering module, the logical operation unit and the pixel memory. By adding the intelligent graph processing buffer which has the graph processing function into a standard DRAM (Dynamic Random Access Memory) memory, the graph data processing can be realized in the DRAM, the load of the whole system is effectively reduced and the idle DRAM memory is completely utilized.
Description
[technical field]
The present invention relates to a kind of advanced figure memory, it can be used for computing machine, notebook computer, workstation or virtual reality system.Sort memory chip both can be used as general primary memory, also can realize drawing function; It can generate very complicated parallel computing figure or virtual reality system.
[background technology]
Now, computer system or the integrated graphic process unit of use CPU (central processing unit), or use relevant graphics calculations computing machine to adopt special Graphics Processing Unit.Graphical content or graphic element or be stored in standard DRAM (Dynamic Random Access Memory) the primary memory part of computing machine, or be stored in the dedicated graphics memory on the display card that inserts computing machine.
But, can find out from JEDEC (electron device engineering federation) relevant criterion DRAM and graphic memory specification, in the function quintessence of storer, be the same.
In essence, graphic process unit or CPU, by from graphic memory reading out data, carry out all graphic operations, complete the task such as coordinate transform, graph rendering, then result are write to graphic process unit, store.
Prior art provides dedicated graphics memory, or by computer primary memory part, store graphics data and content.These are standardization DRAM storer in essence, wherein the higher message transmission rate of graphic memory support.But graphic memory itself does not carry out any computations or other are about graphing capability.
[summary of the invention]
The object of the invention is to propose a kind of advanced figure memory, to solve the problems of the technologies described above.
To achieve these goals, the present invention adopts following technical scheme:
A kind of advanced figure memory, comprises intelligent graphic processing impact damper, Memory Controller Hub and memory chip; Described intelligent graphic is processed impact damper and is connected Memory Controller Hub and memory chip.
Described intelligent graphic is processed impact damper and is comprised global control module, data selection module, buffer module and arithmetic logic unit; Described memory chip comprises graphic memory and pixel store; Described global control module connects Memory Controller Hub, data selection module, buffer module and arithmetic logic unit; Arithmetic logic unit connects buffer cell and data selection module; Data selection module connects Memory Controller Hub, buffer module, graphic memory and pixel store.
The traffic that intelligent graphic is processed between impact damper and Memory Controller Hub is less than the traffic between intelligent graphic processing impact damper and memory chip.
Described global control module receives the order that Memory Controller Hub sends, and controls data selection module data are selected, and controls buffer module data are cushioned, and steering logic arithmetic element is carried out calculation process to data; Data selection module is selected data under the control of global control module; Buffer module cushions data; Steering logic arithmetic element is carried out calculation process to data.
Described advanced figure memory comprises order wire and data line; Memory Controller Hub connects the global control module of intelligent graphic processing impact damper by command signal line; Memory Controller Hub is selected module by data line connection data; The data that described intelligent graphic is processed impact damper reception pixel store, graphic memory send to pixel store, graphic memory to store after treatment again; Memory Controller Hub is processed the duty of the order control intelligent graphic processing impact damper of impact damper transmission by command signal alignment intelligent graphic; Memory Controller Hub sends data to intelligent graphic by data line and processes impact damper or receive the data that intelligent graphic processing impact damper was processed.
Described advanced figure memory comprises at least two pairs of graphics process impact dampers and memory chip; Graphics process impact damper connects corresponding reservoir chip; At least two graphics process impact dampers connect same Memory Controller Hub.
Described graphic memory and pixel store are positioned at the different storage zone of described memory chip.
The data transmission that intelligent graphic is processed the data of impact damper read memory chip or data identical or that revised write to identical or different memory chip is without intelligent graphic processing impact damper and memory chip other device in addition.
Intelligent graphic is processed impact damper from a memory chip reading out data, and writes identical or different memory chip after these data are changed.
Intelligent graphic is processed impact damper reading out data from described graphic memory, and these data is carried out writing with different forms after conversion operations the different memory addresss of graphic memory.
Compared with prior art, the present invention has the following advantages: a kind of advanced figure memory of the present invention, the intelligent graphic by setting up in standard DRAM storer with graphics processing function is processed impact damper, realize graphic processing data in DRAM inside, thereby the traffic realizing between intelligent graphic processing impact damper and Memory Controller Hub is less than the traffic between intelligent graphic processing impact damper and memory chip; Effectively reduce the load of whole system, make full use of idle DRAM storer.
[brief description of the drawings]
Figure 1A-Fig. 1 C is advanced figure memory schematic diagram;
Fig. 2 is memory module and the system schematic based on advanced figure memory;
Fig. 3 is that the information of diverse location on display screen is with respect to the schematic diagram of different storage chips;
Fig. 4 is the storage schematic diagram of pixel and figure on a kind of advanced figure memory;
Fig. 5 is the structural representation that intelligent graphic is processed impact damper.
[embodiment]
Below in conjunction with accompanying drawing, the present invention is described in further detail.
Refer to shown in Figure 1A to Fig. 1 C, a kind of advanced figure memory (AGM of the present invention, Advanced Graphics Memory) intelligent graphic process impact damper (IB, IntelligentBuffer) can be arranged at Memory Controller Hub (MC, Memory controller) and memory module between, connect 4 memory chips of Memory Controller Hub (MC) and memory module, as shown in Figure 1A; As shown in Figure 1B, an intelligent graphic is processed impact damper (IB) and is arranged in memory module, connects 4 memory chips of Memory Controller Hub (MC) and memory module; As shown in Figure 1 C, 4 intelligent graphics are processed impact damper (IB) and are arranged on 4 memory chips of memory module, and 4 intelligent graphics are processed impact damper (IB) and all connected Memory Controller Hub (MC).Intelligent graphic of the present invention is processed impact damper (IB) and is had graphics processing function, owing to processing in impact damper (IB) and carry out on the spot computations at the intelligent graphic of advanced figure memory (AGM), will reduce to the overall system data traffic of microcontroller transmission.
Refer to computer system shown in Fig. 2, wherein, advanced figure memory of the present invention replaces whole primary memory.This system comprises 2 respectively memory module with 4 memory chips.Each chip storage capacity is 2G.
Refer to shown in Fig. 3, for the information of diverse location on display screen is with respect to the schematic diagram of different storage chips, this display screen is divided into 8 regions, covers 2k × 4k and always shields, be i.e. 8M pixel; 8 regions on this display screen are corresponding with 8 memory chips in Fig. 2.Each screen district 1k × 1k, i.e. 1M pixel, and according to being allocated in the memory chip shown in Fig. 2 accordingly.Corresponding cartographic data is stored and revised to each memory chip intelligently.
Refer to shown in Fig. 4, shown a kind of implementation that can express at memory chip graph data; This storer is divided into pixel store and two kinds of forms of graphic memory.Pixel store representative is presented at the pattern pixel on Fig. 3 screen, and in this implementation, each pixel represents by 64 information, 32 bit representation colors, 32 bit representation z coordinates (depth information).Thereby a 64M position that 1M pixel area consumes total memory, be about 3% of 2G position chip storage capacity.
Graphic memory storage is used for representing the graph data that will show, is generally and supports the more triangle of labyrinth.Represent in method at this storer, a triangle approximately needs 384 storage 3 angular coordinates and structural information.Therefore, a 2G bit memory could be stored approximately 5,000,000 triangles! If the word line length 8k activating in DRAM, along this word line storage triangle information representing, also approximately 21 triangles of line activating of a word line.
Referring to shown in Fig. 5, is the preferred structural representation of one of intelligent graphic processing impact damper (IB) in the present invention, and its basic structure can be different because of the environment of application, but basic function is similar.
Refer to shown in Fig. 5, the intelligent graphic of advanced figure memory (AGM) is processed impact damper (IB) and is comprised global control module (GC, Globe Control), data selection module (DSData Select), buffer module (Buffer) and arithmetic logic unit (ALU, ArithmeticLogic Unit).Global control module connects Memory Controller Hub (MC), data selection module, buffer module and arithmetic logic unit; Arithmetic logic unit connects buffer cell and data selection module; Data selection module connects Memory Controller Hub (MC), buffer module, graphic memory and pixel store.
Major function and the relation of these modules are as follows:
Global control module receives the order that Memory Controller Hub (MC) sends, and controls the duty of other modules in intelligent graphic processing impact damper (IB), controls data selection module data are selected; Controlling buffer module cushions data; Steering logic arithmetic element is carried out calculation process to data;
Data selection module is selected data under the control of global control module;
Buffer module cushions data;
Steering logic arithmetic element is carried out calculation process to data.
It is as follows with the relation of pixel store, graphic memory and Memory Controller Hub that intelligent graphic is processed impact damper (IB):
The data that intelligent graphic is processed impact damper (IB) module reception pixel store, graphic memory send to pixel store, graphic memory to store after treatment again.
Memory Controller Hub (MC) can be processed to intelligent graphic the duty of the order control intelligent graphic processing impact damper (IB) of impact damper (IB) transmission by command signal line; Memory Controller Hub (MC) can send data to intelligent graphic by data line and process impact damper (IB) or receive the data that intelligent graphic processing impact damper (IB) was processed.
In the present invention, the intelligent graphic by setting up in memory module with graphics processing function is processed impact damper (IB), can the data in the inner directly invoke memory of memory module carry out graphics process, without by the machine, data being sent in central graphic process unit, and then result is write to graphic memory, this has further improved the performance of system.
Had this configuration, advanced figure memory of the present invention (AGM) just can be carried out independent operation according to the graphical information of having stored, and reduces the traffic of accumulator system to Memory Controller Hub (MC).Only lift some possible examples that are operating as:
A) automatic depth resolution and the transmission to pixel screen:
Advanced figure memory (AGM) reads the information in graphic memory and sends it in pixel store.Advanced figure memory (AGM) and the data comparison in pixel store, carry out the depth resolution of z coordinate, and remove hiding data.This needs read memory, compares, and then writes pixel store.
Suppose a polygon Zhan Ping district approximately 5%, according to current memory architecture, each storage chip need complete about 100 ten thousand polygons operation, substantially need to be between Memory Controller Hub (MC) and advanced figure memory (AGM) transmission information.
B) the automatic motion of graphics memory object:
If observer changes the shape to Drawing Object, can cause the motion of Drawing Object, as transverse shifting, rotation or amplification are dwindled.This can realize by matrix multiplication conventionally, as: each coordinate and 9 matrix elements multiply each other.Therefore, a triangular polygon needs matrix element=81, the record × 9 time multiplying of 3 coordinate × 3.Advanced figure memory (AGM) will read word line, carries out these multiplyings, and is stored in same storage unit to operate.If advanced figure memory (AGM) possesses the pipeline multiplier that can carry out 20 clock period this operation, so, adopt that the advanced figure memory (AGM) of current memory technology is per second carries out approximately 1,000 ten thousand polygon conversion.
C) automatically play up and aberration:
Equally, read after polygon information, advanced figure memory (AGM) can use to be played up and colouring algorithm Show Color and the surperficial image with ad hoc structure automatically, and is stored in pixel store.If can reach 100 ten thousand polygonal processing speeds per second for the polygon of 5% pixel screen size, can adopt the method.
Based on above realization, under the help of advanced figure memory (AGM), storer backup system can greatly reduce the traffic between Memory Controller and the DRAM storer of reality.In above system, Memory Controller Hub (MC) only sends the instruction such as mobile or spin data to advanced figure memory (AGM), and advanced figure memory (AGM) can automatically perform operation.In order to show on computer screen, Memory Controller Hub (MC) only needs to provide instruction, upgrades graphics shape data, transmits shape information, and read pixel store information between each independent storage chip.
Said system also can be used as standard computer storer, and just advanced figure memory (AGM) needs some data processing expenses, so cause chip price higher (than the high 5-30% of generic storage chip price).
But said method can produce extremely powerful figure and virtual reality system, and hardware cost is very low.Suppose 4 modules of a system based on 16x2GB memory chip, this system can provide the parallel computation of 64 times of one single chips, allows the nearly content of 300,000,000 triangular polygons, and speed is the operation of 100,000,000 triangular polygons per second.
Claims (5)
1. an advanced figure memory, is characterized in that, comprises intelligent graphic processing impact damper, Memory Controller Hub and memory chip; Described intelligent graphic is processed impact damper and is connected Memory Controller Hub and memory chip; Described intelligent graphic is processed impact damper and is arranged at memory chip outside;
Described intelligent graphic is processed impact damper and is comprised global control module, data selection module, buffer module and arithmetic logic unit;
Described memory chip comprises graphic memory and pixel store;
Described global control module connects Memory Controller Hub, data selection module, buffer module and arithmetic logic unit; Arithmetic logic unit connects buffer cell and data selection module; Data selection module connects Memory Controller Hub, buffer module, graphic memory and pixel store;
The traffic that intelligent graphic is processed between impact damper and Memory Controller Hub is less than the traffic between intelligent graphic processing impact damper and memory chip;
Described global control module receives the order that Memory Controller Hub sends, and controls data selection module data are selected, and controls buffer module data are cushioned, and steering logic arithmetic element is carried out calculation process to data;
Data selection module is selected data under the control of global control module;
Buffer module cushions data;
Steering logic arithmetic element is carried out calculation process to data;
Described advanced figure memory comprises order wire and data line;
Memory Controller Hub connects the global control module of intelligent graphic processing impact damper by command signal line; Memory Controller Hub is selected module by data line connection data;
The data that described intelligent graphic is processed impact damper reception pixel store, graphic memory send to pixel store, graphic memory to store after treatment again;
Memory Controller Hub is processed the duty of the order control intelligent graphic processing impact damper of impact damper transmission by command signal alignment intelligent graphic; Memory Controller Hub sends data to intelligent graphic by data line and processes impact damper or receive the data that intelligent graphic processing impact damper was processed.
2. a kind of advanced figure memory as claimed in claim 1, is characterized in that, described graphic memory and pixel store are positioned at the different storage zone of described memory chip.
3. a kind of advanced figure memory as claimed in claim 1, it is characterized in that, the data transmission that intelligent graphic is processed the data of impact damper read memory chip or data identical or that revised write to identical or different memory chip is without intelligent graphic processing impact damper and memory chip other device in addition.
4. a kind of advanced figure memory as claimed in claim 1, is characterized in that, intelligent graphic is processed impact damper from a memory chip reading out data, and these data are carried out writing identical or different memory chip after conversion operations.
5. a kind of advanced figure memory as claimed in claim 1, it is characterized in that, intelligent graphic is processed impact damper reading out data from described graphic memory, and these data is carried out writing with different forms after conversion operations the different memory addresss of graphic memory.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110185455.1A CN102262603B (en) | 2011-07-04 | 2011-07-04 | Advanced graph memory |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110185455.1A CN102262603B (en) | 2011-07-04 | 2011-07-04 | Advanced graph memory |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102262603A CN102262603A (en) | 2011-11-30 |
| CN102262603B true CN102262603B (en) | 2014-07-09 |
Family
ID=45009237
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201110185455.1A Active CN102262603B (en) | 2011-07-04 | 2011-07-04 | Advanced graph memory |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102262603B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104504647A (en) * | 2014-12-09 | 2015-04-08 | 中国航空工业集团公司第六三一研究所 | On-chip storage hierarchical structure of image processor |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1189902A (en) * | 1995-05-11 | 1998-08-05 | 株式会社日立制作所 | Semiconductor memory with computing function and processor using same |
| CN101404177A (en) * | 2008-11-13 | 2009-04-08 | 戴葵 | Computation type memory with data processing capability |
| CN102110065A (en) * | 2011-02-21 | 2011-06-29 | 山东华芯半导体有限公司 | Cache system for reducing data transmission |
| CN202167057U (en) * | 2011-07-04 | 2012-03-14 | 山东华芯半导体有限公司 | Advanced figure memory |
-
2011
- 2011-07-04 CN CN201110185455.1A patent/CN102262603B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1189902A (en) * | 1995-05-11 | 1998-08-05 | 株式会社日立制作所 | Semiconductor memory with computing function and processor using same |
| CN101404177A (en) * | 2008-11-13 | 2009-04-08 | 戴葵 | Computation type memory with data processing capability |
| CN102110065A (en) * | 2011-02-21 | 2011-06-29 | 山东华芯半导体有限公司 | Cache system for reducing data transmission |
| CN202167057U (en) * | 2011-07-04 | 2012-03-14 | 山东华芯半导体有限公司 | Advanced figure memory |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102262603A (en) | 2011-11-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101667508B1 (en) | Multimode memory access techniques for performing graphics processing unit-based memory transfer operations | |
| EP2423821A2 (en) | Processor, apparatus, and method for fetching instructions and configurations from a shared cache | |
| CN101387988B (en) | Computer having flash memory and method of operating flash memory | |
| US10163180B2 (en) | Adaptive memory address scanning based on surface format for graphics processing | |
| WO2009023637A2 (en) | Memory device and method having on-board processing logic for facilitating interface with multiple processors, and computer system using same | |
| US11657119B2 (en) | Hardware accelerated convolution | |
| JP2013504130A (en) | Processing unit internal memory for general use | |
| CN116010299B (en) | A data processing method, device, equipment and readable storage medium | |
| CN104160420A (en) | Execution of graphics and non-graphics applications on a graphics processing unit | |
| EP0777233A1 (en) | A memory architecture using conserved adressing and systems and methods using the same | |
| CN112967666B (en) | LED display screen control device and control method capable of realizing random pixel arrangement | |
| US9196014B2 (en) | Buffer clearing apparatus and method for computer graphics | |
| CN111783933A (en) | Hardware circuit design and method for data loading device combining main memory and accelerating deep convolution neural network calculation | |
| CN102262603B (en) | Advanced graph memory | |
| CN101930357B (en) | System and method for realizing accessing operation by adopting configurable on-chip storage device | |
| WO2022095010A1 (en) | Methods and apparatus for rasterization of compute workloads | |
| US7490208B1 (en) | Architecture for compact multi-ported register file | |
| CN202167057U (en) | Advanced figure memory | |
| US20230335049A1 (en) | Display panel fps switching | |
| JP2000011190A (en) | Image processor | |
| US20240193844A1 (en) | Configurable multiple-die graphics processing unit | |
| CN109634583A (en) | A kind of the hardware TLM model and realization structure of the GPU color storage Cache module based on SystemC | |
| US11574380B2 (en) | Methods and apparatus for optimizing GPU kernel with SIMO approach for downscaling utilizing GPU cache | |
| US12117933B2 (en) | Techniques for supporting large frame buffer apertures with better system compatibility | |
| US20220406371A1 (en) | Area efficient memory cell read disturb mitigation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C56 | Change in the name or address of the patentee | ||
| CP03 | Change of name, title or address |
Address after: 710075 Shaanxi City, Xi'an province high tech Road No. 38, innovation center, A, block, floor 4 Patentee after: XI'AN UNIIC SEMICONDUCTORS Co.,Ltd. Address before: 710055 Shaanxi City, Xi'an province high tech Road No. 38, innovation center, A, block, floor 4 Patentee before: Xi'an Sinochip Semiconductors Co., Ltd. |