Grafikpipelines

Grafikpipelines, often referred to as graphics pipelines, describe the sequence of processing steps that convert 3D scene data into a 2D image on the screen. They are a central concept in modern GPU architectures and can work alongside compute pipelines for non-graphics tasks. A graphics pipeline combines programmable shader stages with fixed-function operations and is usually configured through a pipeline state object in many graphics APIs.

The typical stages include: input assembler, which gathers vertex data from buffers according to an input layout; vertex shader, which processes each vertex (transformations, per-vertex lighting, etc.); optional tessellation stages (hull and domain shaders) that subdivide geometry for higher detail; optional geometry shader, which can generate or modify primitives; rasterizer, which converts primitives into fragments and handles interpolation; fragment (pixel) shader, which computes final color and per-fragment effects (textures, lighting, etc.); and the output-merger or blending stage, which applies depth, stencil tests and writes to the framebuffer. Some pipelines also perform post-processing via compute shaders or full-screen passes.

APIs expose pipelines as configurations that combine shader programs, fixed-function state, and resource bindings. The graphics pipeline contrasts with the compute pipeline, which relies on compute shaders for general-purpose processing. Performance hinges on parallel execution of many vertices and fragments, memory bandwidth, and data locality, with pipeline objects facilitating compilation, validation, and efficient state management.