RenderPipelines

RenderPipelines refers to the sequences of processing steps used by a graphics API to convert scene data into a final image. They encompass graphics pipelines, which handle vertex processing, shading, rasterization, and final pixel output; compute pipelines, which execute general-purpose GPU compute shaders; and specialized pipelines for techniques such as ray tracing. The term appears in game engines and graphics frameworks as both a high-level concept and, in some ecosystems, a concrete API object such as a pipeline state object.

A typical graphics pipeline divides work into programmable shader stages—vertex, geometry, tessellation, and fragment (pixel) shaders—and fixed-function stages such as input assembly, rasterization, and the output merger. In modern APIs, a pipeline is defined by a pipeline state object that includes shader programs, input layouts, primitive topology, culling, depth/stencil and blending settings, viewport configurations, and resource bindings. Compute pipelines focus on one or more compute shaders and do not include rasterization.

Creation and use: pipelines are created ahead of time (sometimes with just-in-time compilation) and are bound to command streams where resources (buffers, textures, samplers) are bound before draw or dispatch calls. Pipeline objects enable the GPU to optimize execution, including shader specialization, layout matching, and cache reuse. Some APIs also provide pipeline caches to reduce load times.

In practice, RenderPipelines underpin real-time rendering in engines such as Unity’s Scriptable Render Pipeline family (including URP and HDRP) and Unreal Engine, as well as platform APIs like Vulkan, Direct3D 12, Metal, and ray-tracing pipelines (DXR and Vulkan Ray Tracing).