In materials sciences, the deep understanding of the microstructure dynamics relies on the high accuracy of the numerical methods. During the last two decades, the thin interface phase-field (TIPF) methods have emerged as a powerful numerical tool to tackle the time dependent free boundary problems (i. e. Stefan problems) like solidification problems.
The TIPF can solve problems at relevant experimental scales (~mm) with high accuracy by bridging the scales between the capillary (~nm) and the diffusion (~mm) lengths. To achieve such a high accuracy at those scales, the codes are developed using high performance computing (HPC) techniques.
In the present work, the homemade developed code is based on Cuda C. It will be requested from the intern student to optimize the code considering different multi-GPUs techniques (HPC) and set a guideline of the most efficient technique according to data centers architectures.
