OpenACC is a directive-based parallel programming model designed to simplify the process of writing high-performance computing (HPC) applications. OpenACC codes can be run on a wide variety of platforms (CPUs and GPUs different vendors etc).

Beginner Level – Fundamentals

• Introduction to OpenACC with Fortran
• First OpenACC Fortran Program – Parallel Loop on GPU
• Understanding Fortran Array Indexing in OpenACC
• Basic Parallel Loop Directive with DO Loops
• Loop Directive with Num_gangs, Num_workers, Vector_length
• Data Clauses – Copyin, Copyout, Copy with Fortran Arrays
• Data Clauses – Create, Present, Delete
• Array Sections and Partial Array Transfers

Beginner Level – Data Management

• Data Regions with Fortran Arrays
• Working with Allocatable Arrays
• Multi-dimensional Arrays in OpenACC
• Update Directive for Array Synchronization
• Managed Memory with Fortran
• Column-Major vs Row-Major Performance Considerations

Intermediate Level – Parallel Constructs

• Kernels vs Parallel Directives in Fortran
• Gang, Worker, Vector with Nested DO Loops
• Reduction Operations on Fortran Arrays
• Atomic Operations in Fortran
• Loop Collapse for Nested DO Loops
• Tile Clause with Multi-dimensional Arrays
• Cache Directive for Fortran Arrays

Intermediate Level – Advanced Directives

• Async and Wait with Fortran Arrays
• Multiple Arrays and Data Dependencies
• Routine Directive for Subroutines
• Routine Directive for Functions
• Sequential and Auto Clauses
• Host_data Directive in Fortran

Intermediate Level – Dynamic Memory

• Allocatable Arrays with OpenACC
• Dynamic Memory Management
• Resizing Arrays in Parallel Regions
• Memory Allocation on Device

Advanced Level – Optimization

• Optimizing Fortran Array Transfers
• Data Reuse with Fortran Arrays
• Minimizing Memory Movement
• Efficient Multi-dimensional Array Access

Advanced Level – Complex Patterns

• Matrix Operations with OpenACC
• Stencil Patterns in Fortran
• Parallel Reductions on 2D Arrays
• Transpose Operations

Performance Analysis

• Compiler Feedback for Fortran Programs
• Profiling Fortran OpenACC Applications
• Performance Bottleneck Identification

CPU Fallback

• Running Fortran OpenACC on Multicore CPUs
• Performance Portability in Fortran
• Target-specific Optimizations