Applied Acoustics, 2017 (First author)
We present a parallel time-domain wave solver designed for large and high frequency acoustic domains. Our approach is based on a novel scalable method for dividing acoustic field computations specifically for large-scale distributed memory clusters using parallel Adaptive Rectangular Decomposition (ARD).
In order to efficiently compute the acoustic field for large or high frequency domains, we need to take full advantage of the compute resources of large clusters. This is done with new algorithmic contributions, including a hypergraph partitioning scheme to reduce the communication cost between the cores on the cluster, a novel domain decomposition scheme that reduces the amount of numerical dispersion error introduced by the load balancing algorithm, and a revamped pipeline for parallel ARD computation that increases memory efficiency and reduces redundant computations.
Our resulting parallel algorithm makes it possible to compute the sound pressure field for high frequencies in large environments that are thousands of cubic meters in volume. We highlight the performance of our system on large clusters with 16,000 cores on homogeneous indoor and outdoor benchmarks up to 10 kHz. To the best of our knowledge, this is the first time-domain parallel acoustic wave solver that can handle such large domains and frequencies.
Symposium on Solid & Physical Modeling, 2016 (First author)
We present a novel approach for optimizing acoustic parameters using sensitivity analysis for computer-aided design and analysis of architectural models. Our approach builds on recent low-dispersion wave-based acoustic solvers that can accurately compute the pressure field in large models. We present an efficient technique to compute the gradient of the pressure field using automatic differentiation and combine that with a quasi-Newtonian optimization method to automatically compute the optimal material properties. We highlight the performance on many complex CAD models to optimize the strength and clarity acoustic parameters, and thereby improve the acoustic characteristics of large models. To the best of our knowledge, this is the first practical and accurate approach for acoustic material optimization of large indoor CAD models.
ACM SAP, Proceedings of ACM Transactions on Applied Perception, 2016 (Co-author)
As sound propagation algorithms become faster and more accurate, the question arises as to whether the additional efforts to improve fidelity actually offer perceptual benefits over existing techniques. Could environmental sound effects go the way of music, where lower-fidelity compressed versions are actually favored by listeners? Here we address this issue with two acoustic phenomena that are known to have perceptual effects on humans and that, accordingly, might be expected to heighten their experience with simulated environments. We present two studies comparing listeners' perceptual response to both accurate and approximate algorithms simulating two key acoustic effects: diffraction and reverberation. For each effect, we evaluate whether increased numerical accuracy of a propagation algorithm translates into increased perceptual differentiation in interactive virtual environments. Our results suggest that auditory perception does benefit from the increased accuracy, with subjects showing better perceptual differentiation when experiencing the more accurate rendering method: The diffraction experiment shows a more linearly decaying sound field (with respect to the diffraction angle) for the accurate diffraction method, while the reverberation experiment shows that more accurate reverberation, after modest user experience, results in near-logarithmic response to increasing room volume.
A parallel time-domain wave simulator based on rectangular decomposition for distributed memory architectures
Applied Acoustics, 2015 (First author)
We present a parallel time-domain simulator to solve the acoustic wave equation for large acoustic spaces on a distributed memory architecture. Our formulation is based on the adaptive rectangular decomposition (ARD) algorithm, which performs acoustic wave propagation in three dimensions for homogeneous media. We propose an efficient parallelization of the different stages of the ARD pipeline; using a novel load balancing scheme and overlapping communication with computation, we achieve scalable performance on distributed memory architectures. Our solver can handle the full frequency range of human hearing (20 Hz–20 kHz) and scenes with volumes of thousands of cubic meters. We highlight the performance of our parallel simulator on a CPU cluster with up to a thousand cores and terabytes of memory. To the best of our knowledge, this is the fastest time-domain simulator for acoustic wave propagation in large, complex 3D scenes such as outdoor or architectural environments.