Leveraging the accelerated processing units for seismic imaging: A performance and power efficiency comparison against CPUs and GPUs

Abstract

Oil and gas companies rely on high performance computing to process seismic imaging algorithms such as reverse time migration. Graphics processing units are used to accelerate reverse time migration, but these deployments suffer from limitations such as the lack of high graphics processing unit memory capacity, frequent CPU-GPU communications that may be bottlenecked by the PCI bus transfer rate, and high power consumptions. Recently, AMD has launched the Accelerated Processing Unit (APU): a processor that merges a CPU and a graphics processing unit on the same die featuring a unified CPU-GPU memory. In this paper, we explore how efficiently may the APU be applicable to reverse time migration. Using OpenCL (along with MPI and OpenMP), a CPU/APU/GPU comparative study is conducted on a single node for the 3D acoustic reverse time migration, and then extended on up to 16 nodes. We show the relevance of overlapping the I/O and MPI communications with the computations for the APU and graphics processing unit clusters, that performance results of APUs range between those of CPUs and those of graphics processing units, and that the APU power efficiency is greater than or equal to the graphics processing unit one.

Ultrasonic computed tomography based on full-waveform inversion for bone quantitative imaging

“…However, thanks to the exponential growth of the power of computers this is less true each year. For instance graphics processing units (GPU) cards are an affordable and very efficient way of drastically speeding up such techniques (Michéa and Komatitsch 2010, Fabien-Ouellet et al 2017, Said et al 2017.…”

Applying Non-Local Means Filter on Seismic Exploration

“…Seismic data volumes are large, exceeding one terabyte, especially the 3D seismic data, which occupy massive (out-of-core) storage space that may be attached to a large scale SMP (Symmetric Multiprocessor) supercomputers or use a smaller scale cluster equipped with multicore CPU and OpenMP that provides the needed computational power at a fraction of the cost [10].…”

Dual tree traversal on integrated GPUs for astrophysical N-body simulations

“…On AMD GPU, each compute unit contains 64 processing elements (PEs) and work-items are SIMD processed by wavefronts of 64 work-items. A more detailed presentation of this AMD APU can be found in Said et al (2018).…”

“…This enables to avoid explicit copies between the main memory and the GPU memory, to alleviate the possible performance bottleneck on discrete GPUs due to the PCI bus, and to allocate more memory than within a discrete GPU. Moreover, these iGPUs (along with their CPU) are usually more compute powerful than standard CPUs but offer lower compute power and memory bandwidth than discrete GPUs (Said et al, 2018). The possible performance gains over discrete GPUs depend thus on the application and algorithm features (frequency and volume of PCI transfers, proportion of work deported on GPU, etc.…”