Reservoir Simulation on NVIDIA Tesla GPUs
نویسندگان
چکیده
In this paper, we introduce our work on accelerating a black oil simulator using GPU-based parallel iterative linear solvers. We develop iterative linear solvers and several commonly used preconditioners on NVIDIA Tesla GPUs. These solvers and preconditioners are coupled with our in-house reservoir simulator. Numerical experiments show that our GPU-based black oil simulator is sped up around six times faster than a pure CPU-based simulator.
منابع مشابه
The advantage of the GPU-based supercomputer simulation of plasma phenomena∗
A 3D kinetic study of the plasma relaxation processes caused by the propagation of an electron beam in high-temperature plasma was carried out. The mathematical model is built on the basis of the Particle-in-Cell (PIC) method. The performance for supercomputers powered by both Intel Xeon processors and Nvidia Tesla GPUs (Graphic Processor Units) is given. It is shown that the particle motion st...
متن کاملPARSEC Benchmark Suite: A Parallel Implementation on GPU using CUDA
Graphics Processing Units (GPUs) are a class of specialized parallel architectures with tremendous computational power. The Compute Unified Device Architecture (CUDA) programming model from NVIDIA facilitates programming of general purpose applications on their GPUs. In this project, we targets Parsec benchmarks to provide orders of performance speed up and reducing overall execution time on mu...
متن کاملNew Basic Linear Algebra Methods for Simulation on GPUs
We have used Graphics Processing Units (GPUs) to accelerate the solution of the types of equations typically encountered in dynamic system simulators. Compared to commercial matrix solvers that run on a CPU, we realized speedups ranging from 5 (for system size ≈ 700) to 460 (for system size ≈ 5800). While calculation time for the commercial matrix solver increased with matrix size ≈ O(N)2.3, ou...
متن کاملEvaluation of Vectorized Monte Carlo Algorithms on Gpus for a Neutron Eigenvalue Problem
Conventional Monte Carlo (MC) methods for radiation transport computations are “history-based”, which means that one particle history at a time is tracked. Simulations based on such methods suffer from thread divergence on the graphics processing unit (GPU), which severely affects the performance of GPUs. To circumvent this limitation, event-based vectorized MC algorithms can be utilized. A ver...
متن کاملMultiple-GPU-Based Frequency-Dependent Finite-Difference Time Domain Formulation Using MATLAB Parallel Computing Toolbox
A parallel frequency-dependent, finite-difference time domain method is used to simulate electromagnetic waves propagating in dispersive media. The method is accomplished by using a singleprogram-multiple-data mode and tested on up to eight NVidia Tesla GPUs. The speedup using different numbers of GPUs is compared and presented in tables and graphics. The results provide recommendations for par...
متن کامل