| Popis: |
Summary Geophysics aims at understanding the structure of the Earth from physical measurements. Due to the complexity and dimensions of the subsurface, geophysicists are required to handle very large observational datasets and to perform intensive computer simulations to improve the knowledge of the subsurface. For this reason, solid-earth simulations, such as the one needed for forward and backward wave propagation modelling, are a very good candidate for high-performance computing (HPC). Geophysical applications have usually been taking advantage of large clusters of CPU-powered compute nodes. However, in the last few years, using graphics processing units (GPUs) is becoming increasingly more popular in the high performance computing area. The vast computational power of GPUs provides a great justification to invest resources to adapt algorithms and modify existing codes. Here, I briefly introduce two geophysical algorithms, namely the Immersive Boundary Conditions (IBC) and Marchenko focusing and discuss their GPU implementations. I emphasise how the discretised equations that underlie the two methods comfortably allows one to take full advantage of the tremendous power offered by GPUs while, at the same time, limiting the slow data transfer from host to device (and vice versa). |
