Post on 20-Dec-2015
Introduction to Geometry Shaders
Patrick Cozzi
Analytical Graphics, Inc.
Overview
Geometry Shaders in the Pipeline Primitive Types Applications Performance
Birds Eye View
Create or destroy primitives on the GPU Requires DirectX 10 or GL_ARB_geometry_shader4
GeometryShader
Geometry Shaders in the Pipeline
VertexShader
FragmentShader
Vertices in world coordinates
VertexShader
FragmentShader
Vertices in world coordinates
Perspective Divide and
Viewport Transformation
clip coordinates window coordinates
Geometry Shaders in the Pipeline
GeometryShader
VertexShader
FragmentShader
PrimitiveAssembly
PDand
VT
GeometryShader
VertexShader
FragmentShader
PrimitiveAssembly
ClippingPDand
VT
GeometryShader
VertexShader
FragmentShader
PDand
VT
clip coordinates clip coordinates
window coordinates
Primitive Types
GeometryShader
Output primitives can be disconnected
Primitive Types
Input Primitives GL_POINTS GL_LINES GL_TRIANGLES Adjacency
Output Primitives GL_POINTS GL_LINE_STRIP GL_TRIANGLE_STRIP
Primitive Types
Input primitive type doesn’t have to equal output primitive type
blogs.agi.com/insight3d/index.php/2008/10/23/geometry-shader-for-debugging-normals/
Applications
Implement glPolygonModeTriangles Points or Line Strips
Emulate GL_ARB_point_spritePoints Triangle Strips
Applications
Displacement Mapping Single pass generate a cube map Extrusions
Shadow volumesFins along silhouettes for fur rendering
Applications: Fur in Lost Planet
Render surface, write buffers forFur ColorAngleLength
GS turns each pixel into a translucent polylineAutomatic LOD
Images from meshula.net/wordpress/?p=124
color angle length
Applications: Fur in Lost Planet
Performance
Duplicates per-vertex operations for vertices shared by primitives
GeometryShader
VertexShader
5 vertices processed 9 vertices processed
Performance
GeometryShader
GeometryShader
GeometryShader
GeometryShader
Must guarantee order in == order out
Performance
Order guarantee affects parallelism
GeometryShader
GeometryShader
GeometryShader
Reorder Buffer
Clipping
Performance
Buffer size needs to support a number of threads running in parallel
Performance
Maximum number of vertices a GS will output:
GEOMETRY_VERTICES_OUT_ARB
Determines the speed of GS execution Make this and vertex sizes as small as
possible
Performance
GeForce 8, 9, and GTX2xxOutput size = vertex size *
GEOMETRY_VERTICES_OUT_ARB
Maximum output size: 1,024 scalarsPerformance is inversely proportional
to output sizeNot a continuous function:
• 1-20 scalars: Peak Performance• 27-40 scalars: 50% Performance• On GeForce 8800 GTX
Performance
BenefitsReduces vertex buffer memory usage
• Compute in GS, e.g. normals• Create more geometry• No need to duplicate (e.g. compared to
equivalent VS implementation)
Less memory == less bus trafficReduces vertex attribute setup cost
Summary
Modify incoming primitive or make a limited number of copies
Not for Large scale amplificationInstancing
Resources
www.microsoft.com/downloads/details.aspx?FamilyId=96CD28D5-4C15-475E-A2DC-1D37F67FA6CD&displaylang=en
Introduction to Direct3D 10SIGGRAPH 2007 Course Notes
Resources
www.opengl.org/registry/specs/ARB/geometry_shader4.txt
GL_ARB_geometry_shader4
Resources
www.realtimerendering.com
Section 3.5
Resources
developer.nvidia.com/object/gpu_programming_guide.html
Section 4.6