High fidelity virtual room acoustics can be approached through direct numerical simulation of wave propagation in a defined space. 3D Finite Difference Time Domain schemes can be employed, and adept well to a parallel programming model. This paper examines the various approaches for calculating these schemes using the Nvidia CUDA architecture. We test the different possibilities for structuring computation, based on the available memory objects and thread-blocking model. A standard test simulation is computed at double precision under different arrangements. We find that a 2D extended tile blocking system, combined with shared memory usage, produces the fastest computation for our scheme. However, shared memory usage is only marginally faster than direct global memory access, using the latest FERMI GPUs.
(See document for exact affiliation information.)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
