Geometryshaders

Geometry shaders are a programmable stage in the graphics pipeline that sits between vertex shaders and the rasterizer. They operate on entire primitives—points, lines, and triangles—produced by the vertex stage and can generate new geometry before rasterization.

They receive one primitive per invocation, along with per-vertex data. The shader can transform vertices, discard primitives, or emit multiple new vertices to form additional primitives. Through APIs, developers specify the input primitive type (points, lines, or triangles) and a maximum number of vertices that can be emitted per invocation. The output is then passed downstream to the rasterizer.

Common uses include expanding points to billboards or quads, thickening lines, extruding geometry, generating silhouettes, procedural geometry, and implementing effects such as geometry-based tessellation or shadow volumes. Geometry shaders can also modify or duplicate existing geometry and pass through per-vertex data to further stages.

Performance and portability are important considerations. Geometry shaders are typically more expensive than vertex or fragment shaders and can increase bandwidth and processing load. Some hardware or API configurations restrict the maximum number of emitted vertices or overall resource usage, and in performance-critical applications they may be avoided in favor of tessellation, compute-based methods, or other techniques.

In current graphics APIs, geometry shaders are widely supported in OpenGL, Direct3D, and Vulkan, though their usage varies and they are not always the preferred solution. They are an optional stage and are often employed selectively to balance flexibility against performance.