Parallel Implementation and Performance Analysis of a 3D Oil Reservoir Data Visualization Tool on the Cell Broadband Engine and CUDA GPU

Abstract

Usefulness of graphically visualizing and manipulating large data sets in oil and gas exploration and production is as important as ever. This paper describes the development and parallelization of a multi-phase 3D oil-water reservoir visualization tool on the IBM Cell computer and CUDA enabled GPU. An independent Oil reservoir simulator described in [1] was used to generate the pressure and oil / water saturation values over a certain period of time. The oil reservoir visualization tool displays data grids in a 3D environment and allows the user to interact with it. Due to large speed requirements, our aim is to parallelize the computations required to interact with and visualize the grid, mainly transformation [2], zooming, camera movement [3] and compute intensive lighting model [4][5]. This tool also allows the user to playback the simulation results over a time duration and fetches data values upon mouse click at a particular grid point on a particular day. The development environments are nVIDIA CUDA and IBM Cell SDK 3.0 along with QT and OpenGL libraries. Various experiments were run on an x86 computer with nVIDIA Quadro FX 5800 GPU, and on an IBM Cell BE computer with 1 QS20 Cell blade containing two 9-core Cell processor packages. Our results indicate that the nVIDIA GPU provides on average, speed up of 67x over serial implementation and IBM Cell BE with 16 SPE SIMD implementation 32x over the serial implementation.

A Server Model for Reliable Communication on Cell/B.E.

“…respectively and achieves significant performance according to the architecture features. [24] presents the development and parallelization of 3D oilwater reservoir data visualization tool in Cell/B.E. and the SIMD implementation is greatly superior to the serial implementation.…”

Deployment and Analysis of a Hybrid Shared/Distributed-Memory Parallel Visualization Tool for 3-D Oil Reservoir Grid on OpenStack Cloud Computing

“…The first goal of our study is evolving parallel methods for visualization of serial oil reservoir [34] to allow a real-time processing of the data, smooth and fast data processing by the analyst, and to promptly adapt to any data changes in the oil reservoir simulator (pressure, oil and water saturation). The oil reservoir visualizer allows the user to load the data VOLUME 8, 2020 of the 3-D grid and sketch each cell, and enter the data for each block of the oil reservoir.…”

“…In this section, we evaluate the performance of the virtual platform on OpenStack private Cloud, as the most popular open source Cloud platform, by providing a comparison with respect to traditional real platform which is proposed in [8]. The authors in [8] provide a good comparison with CUDA implementation in [34]. CUDA falls short in scalability for large data-sets hence, we limit our comparison to the performance of a virtual cluster of multiple VMs running MPI implementation with respect to the HPC system described in [8] to better understand the effect of network bandwidth and latency.…”

Optimizing Particle Systems through CUDA-Assisted Multithreading

“…It is found that CUDA performs better when transferring data to and from the GPU and that CUDA's kernel execution is also consistently faster than OpenCL, despite the two implementations running nearly identical code [12,13]. Historically, various applications have been studied on shared memory multiprocessors, GPUs, and message passing systems, and their performance evaluated on these systems [17,18,19,20,25,26,27]. Uberflow [23] is a GPU-based particle engine featuring particle advection, sorting, and rendering.…”