Vertex Assembly Programming Guide

• 1. Introduction
  • 1.1. Notation
  • 1.2. Instruction set overview
    • 1.2.1. Supervisors and workers
    • 1.2.2. Execution pipelines
  • 1.3. Memory architecture
    • 1.3.1. Load and store instructions
• 2. Writing vertices in assembly
  • 2.1. The Vertex class
  • 2.2. Vertex state
    • 2.2.1. Vector and VectorList types
    • 2.2.2. Specifying memory constraints
  • 2.3. Calling conventions
    • 2.3.1. External codelets
  • 2.4. Worker stack and scratch space
    • 2.4.1. Specifying stack size
    • 2.4.2. Examples
    • 2.4.3. Recursion and function pointers
  • 2.5. Vertex name mangling
• 3. Vector types
  • 3.1. Parameters
    • 3.1.1. Layout
    • 3.1.2. Minimum alignment
    • 3.1.3. Interleaved memory
  • 3.2. Memory layout
    • 3.2.1. Pointer compression
      • ScaledPtr32
      • ScaledPtr64
    • 3.2.2. Vector<T> layout
    • 3.2.3. Vector<Input<Vector<T>>> layout
    • 3.2.4. VectorList layout
      • DELTANELEMENTS layout
      • DELTAN layout
• 4. Vertex pipelines
  • 4.1. Memory conflicts
  • 4.2. Modified pipeline
  • 4.3. Fill and drain
• 5. Assembly hints & tips
  • 5.1. Using the assembler
    • 5.1.1. Assembler macros
    • 5.1.2. Labels
    • 5.1.3. Recording the code size of the vertex
    • 5.1.4. Place each vertex in a unique section
  • 5.2. Architectural tips
    • 5.2.1. Aligning repeat bodies
    • 5.2.2. Over-reading and over-processing
    • 5.2.3. Scratch space
    • 5.2.4. Loading constants
    • 5.2.5. Pre-division by 6
  • 5.3. General
    • 5.3.1. Focus on optimising the vectorised case
    • 5.3.2. Testing
    • 5.3.3. Bit twiddling
• 6. Application binary interface
  • 6.1. Types
    • 6.1.1. Floating point types
    • 6.1.2. Structure types
    • 6.1.3. Bit fields
  • 6.2. Vertex calling convention
  • 6.3. Function calling convention
    • 6.3.1. Parameters
    • 6.3.2. Returns
    • 6.3.3. Entry and exit
    • 6.3.4. Register assignments
  • 6.4. Stack frame
• 7. Glossary
• 8. Trademarks & copyright