Jakub Poła

Application of GPU technology to OpenFOAM simuilations

Modern computational problems require a lot of computing resources. Serial processing are things in the past due to its inefficiency. Physical constraints put the limit on the CPU frequencies therefore the technology vendors switched into parallel architectures by adding more cores into processor. One way to accelerate calculations is to take advantage of shared memory and mu HPC systems based on CPU. The workload is divided between available cores thanks to which the solution can be obtained faster. The alternative way is to utilize the computational potential of modern GPU which performance is much higher compared to CPU. During the talk I will describe the two ideas showing how to involve the GPU into OpenFOAM suite, by presenting SpeedIT and SpeedIT Flow applications.