Implementation of Alternate Direction Implicit Method with CUDA
Nov 26, 2014
|Simmakers specialists have analyzed the efficiency of applying CUDA technology to the implementation of the Douglas-Rachford scheme (alternating direction implicit method, ADI method). The ADI method is used to solve partial differential equations of the parabolic type, such as heat equations in three dimensions. It was found that the Douglas-Rachford scheme presented natural parallelism allowing for effective ADI mathematical solver parallelization with CUDA (Nvidia).
In the article, we provide calculation results demonstrating the performance acceleration for software-implemented Douglas-Rachford scheme using CUDA technology when compared with the implementation of the same scheme on a CPU.
On average, the calculations on the GPU version of ADI solver using CUDA technology on the Nvidia GTX Titan video card ran 30 times faster than the calculations on a one core of Intel Core i7 CPU.
Note: Simmakers experts conducted a comparative performance analysis of CPU and GPU mathematical solver versions of Frost 3D Universal using the production well cluster model. Speed comparison showed that on average the GPU version of the program performed calculations 20 times faster than a single core CPU version.
More about the implementation of the alternating direction implicit method with CUDA.