This work addresses the HPC challenges of fluid structure iteration (FSI), focusing on the computational efficiency of mesh resampling. A major computational challenge arises from the fact that the method requires gather and scatter memory access, which introduces a significant memory barrier, and the fraction of time to solution can be high. We have developed a collection of computational methods capable of achieving significant throughput on both GPU and CPU, and have had the opportunity to test them on the new Alps supercomputer al CSCS, based on the GH200 Grace Hooper. We have also begun to explore the possibility of using 3-dimensional 4th order WENO interpolation as a transfer function, which allows the adaptive removal of the oscillatory behavior that occurs in classical interpolation methods, thus improving the quality of the results.