Unformatted text preview:

1Computer GraphicsHardware Acceleration forEmbedded Level SystemsBrian Murray 02-15-2002Topics• Why?• Introduction to Computer Graphics– Object Representation– Object Rasterization• Graphics Hardware– Goals– Problems– Architecture• Embedded System Optimizations2Object Representation• Goals– Display images to user– Be able to represent images in abstract form• Solution– Use specialized algorithms that take primitivesand convert them into “pixels”.Primitive Transformations• Geometric Transformation– Move Primitives/Polygons (Triangles) to correctpositions in 3D space.– Project 3D points onto 2D points• Lighting– Use lights to correctly color and shade the triangles• Rasterization– Convert the resulting objects (triangles) into pixels3Rasterization - Interpolation• Interpolation of Points– Use the coordinates ateach point and find thepoints between.– Find the pointsbetween the points onthe lines.Rasterization – Lighting/Shading• Phong Illumination– Position of Lights– Orientation of polygon– Position of viewer• Gouraud Shading– Interpolate colors of atthe points of thetriangle. lightsI=Iaka+Σ[Ip(kdcosθ+ksconnα)] n=04Rasterization - Texturing• Map the points of aTexture on the points of aTriangle.– Each triangle point has ascreen coordinate (x,y)– Each triangle point has atexture coordinate (u,v)– Use interpolation to find thepoints between.Rasterization – Depth Checking• Z-Buffer– Stores Z values ofpoints as it israsterizing– Check against the Z-Buffer to find out ifyou are near or fartherfrom the screen.– If closer, write to Z-Buffer and Frame-Buffer (screen)323322333222333332222333311222233111122111111111111111111111111111111111111Frame BufferZ Buffer5Graphics Hardware• Goals– Implement the functionality of graphicsprocessing in hardware.– Speed is usually the main factor– Can be very expensiveGraphics Hardware - Problems• Problems– Many Many Memory Accesses– Large amounts of Floating point math– Must be efficiently pipelined and fast– Requires large amounts of memory, especiallyfor multiple frame-buffers.6Graphics Hardware – GeneralArchitectureEmbedded System Concerns• Space– Cores must take upminimal area• Power– Power constraints aretight• Speed– Must be fast enough tobe useful• Memory– Memory is limited• Heat– Speed and powerdemands generate heat,to which many systemsare not tolerant7Embedded Solutions• Off load Geometric Transformations to CPU• Disregard Post-Rasterizer Functions• Use an asynchronous pipeline• Shut off portions not in use or already done• Try to make maximum usage of memory available• Do as many functions as possible in fixed point• Simplify architecture and make system scalable• Eliminate extra precision to minimize hardwareRasterization Solution - DDS• Sort all vertices from top tobottom• Determine the x/y “slope” ofeach edge• Determine the x starting point(x=slope+ceil(y)-y)• Add to x the slope at each ywhile moving down the screento get beginning and endingpoints.• Interpolate between thosepointsKugler, A.. “The Setup for Triangle Rasterization”,Proceedings of the 11th Eurographics Workshop onGraphics Hardware, Poitiers, France, August 1996, pp.49--58.8Fixed Point Lighting• Convert all vectormath to Log(2)domain• Use 16bit vectors• Less than 1 percentcolor error opposed tofloating pointChen, C-H; Lees, C-Y, “A cost effective lighting processor for 3D graphicsapplication”, Image Processing, 1999. ICIP 99. Proceedings. 1999International Conference on, Volume: 2 , 1999 ,pp2 792 –796.Embedded Architecture19Embedded Architecture2Verification Procedures• Verification– Simulate memory accesses for cacheoptimization– Simulate Architecture for flow and controlverification– Simulate Rasterization with fixed and floatingto show possible error rates– Create a hardware model and verify with FPGA10Future Work• Enhance an Embedded processor for geometrictransformations• Scale the system with a parallel architecture (i.e.mini-frame buffers)• Add modern features such as mip-mapping, per-pixel operations, and z-buffer compression (orpyramidal)• Cache implementations for the system : newlocalities and enhanced


View Full Document

TAMU CSCE 689 - slides

Documents in this Course
riccardo2

riccardo2

33 pages

ffd

ffd

33 pages

intro

intro

23 pages

slides

slides

19 pages

p888-ju

p888-ju

8 pages

w1

w1

23 pages

vfsd

vfsd

8 pages

subspace

subspace

48 pages

chapter2

chapter2

20 pages

MC

MC

41 pages

w3

w3

8 pages

Tandem

Tandem

11 pages

meanvalue

meanvalue

46 pages

w2

w2

10 pages

CS689-MD

CS689-MD

17 pages

VGL

VGL

8 pages

ssq

ssq

10 pages

Load more
Download slides
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view slides and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view slides 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?