Graphics Processing Unit GPU Graphics Processing Unit

Scene Transformations Graphics Pipeline Scene Transformations Lighting & Shading Viewing Transformations Rasterization

Early Graphics Originally, no specialized graphics hardware All processing in software on CPU, Results transmitted to frame buffer first, external frame buffers later, internal frame buffers. CPU Frame buffer Display

GPU fixed function graphics pipeline Geometry data Transform & lighting Culling, perspective divide, viewport mapping Rasterization Texturing Alpha test, depth test, stencil test Frame buffer blending GPU fixed function graphics pipeline

Fixed function pipeline Pipeline fixed in hardware Each stage does fixed task Tasks are parameterized Inflexible

Programmable Pipeline GPU is programmable Shader programs Vertex shaders & Fragment (or pixel) shaders

Programmable GPU graphics pipeline Geometry data Vertex Shader Culling, perspective divide, viewport mapping Rasterization Fragment Shader Alpha test, depth test, stencil test Frame buffer blending

Shader Language Low level (like assembler) but high-level language compilers 4 component floating point data type SIMD Direct3D: Microsoft NV_Vertex_Program: OpenGL extension

Programmable Pipeline 60 frames per second hardware pipeline of specialized stage general-purpose parallel computational engine

Power GPUs have moved away from the traditional fixed-function 3D graphics pipeline toward a flexible general-purpose computational engine. the raw computational power of a GPU dwarfs that of the most powerful CPU, and the gap is steadily widening. GPUs have moved away from the traditional fixed-function 3D graphics pipeline toward a flexible general-purpose computational engine

Pipeline input Graphics systems: convert to triangles Pass vertices to GPU processor GPU: assemble into triangles

Rasterization

Model transformations

Lighting

Camera simulation

Texturing

hidden surfaces

Graphics pipeline evolved

Pipeline evolved

Evolved pipeline

Unified Shaders

GPGPU Single GeForce 8800 chip achieves a sustained 330 billion floating- point operations per second (Gflops) on simple benchmarks