This page is under construction and updates are being made on a regular basis.
OpenMP (Open Multi-Processing) is a widely-used API that supports multi-platform shared memory multiprocessing programming in C, C++, and Fortran. It provides a set of compiler directives, library routines, and environment variables that allow developers to write parallel code for multi-core and multi-processor systems.
OpenMP Basics
- Introduction to OpenMP
- OpenMP Programming Model
- Hello Parallel World!
- Thread Management
- Environment Variables
- OpenMP Runtime Library
- Compilation and Execution
- Execution Flow and Control
- Basic Performance Concepts
- First Parallel Application
Parallel Regions
- Parallel Construct Fundamentals
- Conditional Parallelization
- Nested Parallelism
- Thread Binding and Affinity
- Thread Limits
- Parallel Region Best Practices
- Dynamic Thread Teams
- Orphaned Directives
Work-Sharing Constructs
- Work-Sharing Concepts
- Parallel For Loop Basics
- Loop Scheduling Policies
- Schedule Clause Details
- Loop Dependencies and Parallelizability
- Sections Construct
- Single Construct
- Master Construct
- Workshare (Fortran) / Parallel Loop Variations
- Work-Sharing Best Practices
Data Environment
- Data Scoping Fundamentals
- Shared Variables
- Private Variables
- Firstprivate Variables
- Lastprivate Variables
- Default Clause
- Threadprivate Variables
- Copyin and Copyprivate
- Reduction Concepts
- Data Environment Best Practices
Synchronization
- Synchronization Overview
- Barrier Construct
- Critical Section
- Atomic Operations
- Reduction Clause
- User-Defined Reductions
- Locks and Mutual Exclusion
- Ordered Construct
- Flush Directive
- Synchronization Best Practices
Tasking
- Task-Based Parallelism Concepts
- Task Directive Basics
- Task Synchronization
- Task Dependencies
- Taskloop Directive
- Task Priority
- Taskgroup Construct
- Untied and Mergeable Tasks
- Recursive Algorithms with Tasks
- Task-Based Best Practices
SIMD and Vectorization
- SIMD Fundamentals
- SIMD Directive
- SIMD Clauses
- Combining SIMD with Parallelism
- Declare SIMD
- SIMD Reductions
- Alignment and Memory Access
- SIMD Best Practices
Advanced Topics
- Cancellation
- Target Directives (Offloading)
- Teams Construct
- Distribute Construct
- Memory Allocators
- Thread Affinity and Places
- Loop Transformations
- Metadirectives
- User-Defined Mappers
- Interoperability
Performance and Optimization
- Performance Measurement
- Speedup and Scalability Analysis
- Load Balancing
- False Sharing
- NUMA Effects
- Reducing Synchronization Overhead
- Memory Access Optimization
- Profiling and Debugging
- Compiler Optimization
- Best Practices Summary
