§16§§This book constitutes the refereed proceedings of the 21st International Workshop on OpenMP, IWOMP 2025, in Charlotte, NC, USA, during October 1 October 3, 2025§§The 14 full papers included in this book were carefully reviewed and selected from 19 submissions. They were organized in topical sections as follows:New Directions: Communication and I/O; New Directions: Handling New Constraints and Devices; New Performance and Correctness Tooling; Advanced Capability Evaluation and New Directions: Programming Advances.§ §04§§.- New Directions: Communication and I/O.§.- Discussion of Device-Device Collective Communication in OpenMP Target Offloading.§.- OMPCCL: Extending OpenMP with Portable Collective Operations for Multiple Devices.§.- OpenMP Meets I/O: Portable and Runtime-Managed File Tasks.§.- New Directions: Handling New Constraints and Devices.§.- Parallelizing Irregular DOACROSS Loops Using ChatGPT and Transactional Memory in OpenMP.§.- OpenMP-RT: Pragma Support for Scheduling Periodic Real-Time Tasks.§.- OpenMP-Q: Quantum Task Offloading in OpenMP.§.- New Performance and Correctness Tooling.§.- Profile Generation for GPU Targets.§.- Data Race Satisfiability on Array Elements.§.- Predicting Performance for OpenMP GPU Parameter Choices.§.- Advanced Capability Evaluation.§.- Evaluating LLVM OpenMP Offload Optimizations on NVIDIA GH200 Grace Hopper Superchip and AMD Instinct MI300A Accelerator Architectures.§.- Evaluating OpenMP on Aurora s Intel GPU §04§Max Series 1550.§.- Demonstrating OpenMP Offload Performance with the STREAmS-2 Application and the AMD Next-Gen Fortran Compiler.§.- New Directions: Programming Advances.§.- ChatPORT: Fine-tuned LLM for Easy Code {PORT}ing.§.- Programming GPUs with OpenMP and Python.§
