Title:
Parallel Computing of Particle Trajectory Sonification to Enable Real-Time Interactivity
Parallel Computing of Particle Trajectory Sonification to Enable Real-Time Interactivity
Author(s)
Yang, Jiajun
Hermann, Thomas
Hermann, Thomas
Advisor(s)
Editor(s)
Collections
Supplementary to
Permanent Link
Abstract
In this paper, we revisit, explore and extend the Particle Trajectory
Sonification (PTS) model, which supports cluster analysis of
high-dimensional data by probing a model space with virtual particles
which are ‘gravitationally’ attracted to a mode of the dataset’s
potential function. The particles’ kinetic energy progression of
as function of time adds directly to a signal which constitutes the
sonification. The exponential increase in computation power since
its conception in 1999 enables now for the first time to investigate
real-time interactivity in such complex interweaved dynamic
sonification models. We speeded up the computation of the PTS
model with (i) data optimization via vector quantization, and (ii)
parallel computing via OpenCL. We investigated the performance
of sonifying high-dimensional complex data under different approaches.
The results show a substantial increase in speed when
applying vector quantization and parallelism with CPU. GPU parallelism
provided a substantial speedup for very large number of
particles comparing to using CPU but did not show enough benefit
for a low number of particles due to copying overhead. A hybrid
OpenCL implementation is presented to maximize the benefits of
both worlds.
Sponsor
Date Issued
2017-06
Extent
Resource Type
Text
Resource Subtype
Proceedings
Rights Statement
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.