RAJA User Guide

If you have some familiarity with RAJA and want to get up and running quickly, check out Getting Started With RAJA. This guide contains information about accessing the RAJA code, building it, and basic RAJA usage.

If you are completely new to RAJA, please check out the RAJA Tutorial and Examples. It contains a discussion of essential C++ concepts and will walk you through a sequence of code examples that show how to use key RAJA features.

See RAJA Features for a complete, high-level description of RAJA features (like a reference guide).

Additional information about things to think about when considering whether to use RAJA in an application can be found in Application Considerations.

• Getting Started With RAJA
  • Requirements
  • Get the Code
  • Dependencies
  • Build and Install
    • Additional RAJA Back-end Build Information
    • RAJA Example Build Configuration Files
  • Learning to Use RAJA
• Using RAJA in Your Application
  • CMake Configuration File
• Build Configuration Options
  • RAJA Option Types
  • Setting Options
  • Available RAJA CMake Options and Defaults
    • Examples, tests, warnings, etc.
    • RAJA Features
    • Programming model back-end support
    • Timer Options
    • Data types, sizes, alignment, etc.
    • Other RAJA Features
  • Setting RAJA Back-End Features
• RAJA Features
  • Elements of Loop Execution
    • Simple Loops (RAJA::forall)
    • Complex Loops (RAJA::kernel)
    • Hierarchical loops (RAJA::launch)
    • Multi-dimensional loops using simple loop APIs (RAJA::CombiningAdapter)
  • Policies
    • RAJA Loop/Kernel Execution Policies
    • Parallel Region Policies
    • Reduction Policies
    • Atomic Policies
    • Local Array Memory Policies
    • RAJA Kernel Execution Policies
  • Indices, Segments, and IndexSets
    • Indices
    • Segments and IndexSets
  • View and Layout
    • RAJA Views
    • RAJA Layouts
    • RAJA Index Mapping
    • RAJA Atomic Views
    • RAJA View/Layouts Bounds Checking
  • Reduction Operations
    • Reduction Types
    • Reduction Examples
    • Reduction Policies
    • Experimental Reduction Interface
  • Atomic Operations
    • Atomic Operations
    • CUDA Atomics Architecture Dependencies
    • DESUL Atomics Support
  • Scan Operations
    • Scan Operations
    • RAJA Inclusive Scans
    • RAJA Exclusive Scans
    • RAJA Scan Operators
  • Sort Operations
    • Sort Operations
    • RAJA Unstable Sorts
    • RAJA Stable Sorts
    • RAJA Comparison Operators
  • Resources
    • Type-Erasure
    • Memory Operations
    • Kernel Execution and Resources
    • Events
    • Example
  • Local Array
    • Memory Policies
  • Loop Tiling
  • WorkGroup
    • Policies
    • Arguments
    • Allocators
    • WorkPool
    • WorkGroup
    • WorkSite
  • Vectorization (SIMD/SIMT)
    • Why Are We Doing This?
    • Register
    • Vector Register
    • Tensor Register
    • Matrix Registers
  • Plugins
    • Using RAJA Plugins
    • Creating Plugins For RAJA
• RAJA Cook Book
  • Cooking with Reductions
    • Reductions with RAJA::forall
• Application Considerations
• RAJA Tutorial and Examples
  • RAJA Tutorial
  • A Little C++ Background
    • C++ Templates
    • Elements of C++ Lambda Expressions
    • A Few Notes About Lambda Usage With RAJA
  • RAJA Examples and Exercises
  • Simple Loops and Basic RAJA Features
    • Basic Loop Execution: Vector Addition
    • Iteration Spaces: Segments and IndexSets
    • Iteration Space Coloring: Mesh Vertex Sum
    • Sum Reduction: Vector Dot Product
    • Reduction Types and Kernels with Multiple Reductions
    • Atomic Operations: Computing a Histogram
    • Parallel Scan Operations
    • Parallel Sort Operations
    • Data Views and Layouts
    • Permuted Layout: Batched Matrix-Multiplication
  • Complex Loops and Advanced RAJA Features
  • Nested Loops with RAJA::kernel
    • Basic RAJA::kernel Mechanics and Nested Loop Ordering
    • RAJA::kernel Execution Policies
    • OffsetLayout: Five-point Stencil
  • Nested Loops with RAJA::launch
    • RAJA::Launch Basics
    • RAJA::Launch Execution Policies
    • Naming kernels for NVTX/ROCTX tools
  • Comparing RAJA::kernel and RAJA::launch: Matrix-Transpose
    • Matrix Transpose
    • Tiled Matrix Transpose
    • Tiled Matrix Transpose with Local Array
  • Other RAJA Features and Usage Examples
    • Workgroup Constructs: Halo Exchange
    • Matrix Multiplication: RAJA::kernel