AMD has released a new guide to help developers profile and optimize Fortran OpenMP offload applications running on AMD GPUs. The guide, part of the profiling guide series, focuses on identifying performance bottlenecks and improving GPU utilization through systematic analysis. It builds on previous articles in the series and introduces profiling techniques tailored for Fortran applications using OpenMP target offloading. The guide is designed to assist developers in analyzing and improving the performance of their applications by leveraging tools from ROCm 7.2. Experiments were conducted on an MI300A APU, which supports a unified memory model, but the methodology is applicable to discrete GPUs as well. The guide outlines a workflow that includes establishing baseline performance, identifying bottlenecks, analyzing hardware resource usage, and performing targeted optimizations. It also highlights the use of ROCm profiling tools to collect and analyze data, enabling multiple post-processing analyses on the same dataset. The guide features examples from the GenASiS_Basics repository, a production astrophysics simulation code that combines MPI with OpenMP target offloading for GPU acceleration. Developers can use the guide to build and profile GenASiS with AMD’s amdflang compiler and OpenMPI 5.0.3. *Source: [amd](https://rocm.blogs.amd.com/software-tools-optimization/profiling-guide/fortran_openmp/README.html)*