AnyGL: A Vast Scale Cross breed Conveyed Design Framework.


82 views
Uploaded on:
Category: Fashion / Beauty
Description
AnyGL: A Substantial Scale Half and half Circulated Design Framework. Ph.D Applicant: Yang Jian( 杨建 ) Supervisor: Jiaoying Shi( 石教英 ) State Key Lab of CAD&CG, Zhejiang Univ. 2002. Substance. Objective of PC Illustrations Design Parallel Representation Engineering AnyGL in Points of interest Conclusion
Transcripts
Slide 1

AnyGL: A Large Scale Hybrid Distributed Graphics System Ph.D Candidate: Yang Jian( 杨建 ) Supervisor: Jiaoying Shi( 石教英 ) State Key Lab of CAD&CG, Zhejiang Univ. 2002

Slide 2

Contents Goal of Computer Graphics Architecture Parallel Graphics Architecture AnyGL in Details Conclusion Further Research

Slide 3

Goal of Graphics Architecture Real-time Triangle rendering limit Rasterization rate Texture size Display Resolution Realistic Local enlightenment, Gouraud, Phong Global brightening, beam follow, beam throwing, Radiosity

Slide 4

Goal of Graphics Architecture Programmable, more impacts quicker, reasonable non-sensible

Slide 5

Contents Goal of Computer Graphics Architecture Parallel Graphics Architecture AnyGL in Details Conclusion Further Research

Slide 6

Parallel Graphics Architecture Development of Graphcis Architecure Classification of Parallel Graphics Architecure Hardware System Software System Scalability of Parallel Graphics Architecure

Slide 7

Development of Graphics Architecture 1rd Generantion, 1975~ 1985, Geometry transfromation, clasp, render items in view of wire edge 2rd Generation, mid 1985~ 1990’ Video memory, profundity cradle, back face winnow, gouraud rendering third Generation, 1990‘ ~ 2000’s Texture Mapping hostile to assumed name, mutlis-test

Slide 8

Development of Graphics Architecture fourth Generation , 1997~ Local brightening Hardware lighting, complex surface mapping(bump , packed, 2000 SIMD, 2001 Programmable processor, 2001 More programmable processors, 2002 fifth Generations, 2005? Worldwide brightening

Slide 9

Parallel Graphics Architecture Development of Graphcis Architecure Sorting Classification of Parallel Graphics Architecure Hardware System Software System Scalability of Parallel Graphics Architecure Misc.

Slide 10

Sorting Classification [S. Molnar et al. 1994] sort-first sort-center sort-last [M. Eldridge 2000], Pomegranate Transform & Lighting Rasterization Texture Fragment Display

Slide 11

Typical Graphics Pipeline

Slide 12

sort-first

Slide 13

sort-center

Slide 14

sort-last

Slide 15

Parallel Graphics Architecture Development of Graphcis Architecure Sorting Classification of Parallel Graphics Architecure Hardware System Software System Scalability of Parallel Graphics Architecure

Slide 16

Hardware System of PGA SGI Reality Engine sort-center offer transport, show each 2 scanlines one processor

Slide 17

SGI Reality Engine

Slide 18

Hardware System of PGA PixelFlow Sort-last ring-net, form picture SIMD is very expolited Pomegranate 5 stages 64 hubs

Slide 19

Pomegranate

Slide 20

Hardware System of PGA Evans & Sutherland Freedom 3000 sort-last Kubota Denali sort-last Sepia, Compaq Reserch 2000 sort-last ServerNet-II

Slide 21

Parallel Graphics Architecture Development of Graphcis Architecure Sorting Classification of Parallel Graphics Architecure Hardware System Software System Scalability of Parallel Graphics Architecure Misc.

Slide 22

Sortware System GLR Client/Server GLX, X Windows Princeton Multi-projector System sort-initial one hub distriutes summons 8 hubs get and render orders Parallel Mesa sort-last twofold swapping, parallel pipeline compositing

Slide 23

Sortware System WireGL sort-first state following and state switch, messy bits, Lazy-redesign 3 sorts of OpenGL charges 16 customers + 16 servers Lightning-2 gathers pictures

Slide 24

WireGL(Chromium)

Slide 25

Parallel Graphics Architecture Development of Graphcis Architecure Sorting Classification of Parallel Graphics Architecure Hardware System Software System Scalability of Parallel Graphics Architecure Misc.

Slide 26

Scalability of Parallel Graphics Architecure 5 parameters, [Eldridge et al.2000] triangle information rate triangle rendering rate surface size pixel rate show determination

Slide 27

Contents Goal of Computer Graphics Architecture Parallel Graphics Architecture AnyGL in Details Conclusion Further Research

Slide 28

AnyGL Goal Architecture Classification of OpenGL Commands State Tracking Compression Scalability Parallel Graphics API

Slide 29

Goal of AnyGL Hibrid Distributed Graphics System sort-first sort-last Large scale sort-to begin with, new state following sort-last, picture pressure Good Scalability

Slide 30

AnyGL Goal Architecture Command characterization Analysis of Dependency Relation State Tracking Compression Scalability

Slide 31

Architecture of AnyGL 4 sorts of Nodes Geometry Distributing Node ( G-hub ) Goemetry Rendering None ( R-hub ) Image Composite Node ( C-hub ) Display Node ( D-hub ) Sort-first G-hub, R-hub Sort-Last C-hub, D-hub

Slide 32

Architecture of AnyGL

Slide 34

Geometry Rendering Node Capture opengl orders of use Packet OpenGL orders Computer State changes Send OpenGL summon parcels

Slide 35

Geometry Rendering Node Receive Packet Check if connection switch is required Context switch Unpack the charge bundle Send corlor & profundity support to C-hub

Slide 36

Image Composite Node Receive shading and profundity information from R-hubs If they are compacted, decompress them. Form the got information Send created shading and profundity cradle to D-hubs

Slide 37

Display Node Receive shading and profundity information from C-hub Assemble picture for last show Display the last picture by CRT, Projector, LCD...

Slide 38

Logical Division of Command Buffer Command cradle is separated into 2 sections Command Code support Parameter Buffer Command Code cushion ptr- - ; Parameter Buffer ptr++

Slide 40

Command Packet Structure

Slide 41

Subdivision of glBegin-glEnd Condition bunches of orders show up between a couple of glBegin-glEnd The summon cradle or parameter is full, while glBegin does not meet the comparing glEnd Requirement Make beyond any doubt that glBegin and glEnd are coordinated in every bundle Maitain fulfillment of primitive

Slide 42

Subdivision of glBegin-glEnd 算法 2.1 glBegin 命令打包 设 glBegin 和 glEnd 匹配标示为布尔变量 glBeginEnd ,初始值为 1 ; 设 glVertex 指令计数器为 vtxCount ,初始值等于 0 ; 设 nPrimtiveType 位图元类型变量; glBegin 编码压入当前命令缓冲区å¤\'部; glBegin 参数(图元类型)添加到当前参数缓冲区; glBeginEnd = 0 ; nPrimitiveType = glBegin 命令参数;

Slide 43

Subdivision of glBegin-glEnd 算法 2.2 glVertex* 命令打包 将 glVertex* 编码压入当前命令缓冲区; 参数添加到当前参数缓冲区; if(glBeginEnd==1)vtxCount++ ,否则程序报告非法调用; 算法 2.3 glEnd 函数命令打包 将 glEnd 编码压入当前命令缓冲区; glBeginEnd=1 ; vtxVount=0 ;

Slide 44

Subdivision of glBegin-glEnd 算法 2.4 大几何图元指令组 OpenGL 指令的分割算法 if (glBeginEnd) then 返回,否则执行下面的步骤; if(vtxCount ==0 ) 返回; 计算需要回溯的图元数目为 vtxBackdate ; 如果图元类型为 GL_TRIANGLES_FAN 或者 GL_POLYGON ,需要回溯第一个顶点数据, bRecoverFirst=1 ; 将 glEnd 命令打包当前命令缓冲区; 生成一个新的缓冲区; 调用 glBegin 指令; if(bRecoverFirst == 1) ; { 回溯并执行最近的一次 glBegin 指令调用; 将其后的指令到第一个 glVertex* 命令加入新的缓冲区; } 搜ç\'¢ vBackdate 次 glVertex* 指令调用; 将其后顺序调用的所有 OpenGL 指令及参数分别增加到新的命令缓冲区和参数缓冲区。

Slide 45

AnyGL Goal Architecture Command grouping Analysis of Dependency Relation State Tracking Compression Scalability

Slide 46

Command Classification OpenGL 1.3 17 sorts Four sorts in AnyGL primitive orders state charges remote remapping orders speical orders

Slide 47

Re

Recommended
View more...