high performance computing on graphics processing units: hgpu.org

The staggered-grid finite difference (FD) method demands significantly computational capability and is inefficient for seismic wave modelling in 2-D viscoelastic media on a single PC. To improve computation speedup, a graphic processing units (GPUs) accelerated method was proposed, for modern GPUs have now become ubiquitous in desktop computers and offer an excellent cost-to-performance-ratio parallelism. The geophysical model is decomposed into subdomains for PML absorbing conditions. The vertex and fragment processing are fully used to solve FD schemes in parallel and the latest updated frames are swapped in Framebuffer Object (FBO) attachments as inputs for the next simulation step. The seismic simulation program running on modern PCs provides significant speedup over a CPU implementation, which makes it possible to simulate realtime complex seismic propagation in high resolution of 2048*2048 gridsizes on low-cost PCs.