This year at SC14 in New Orleans, the SCI Institute presented demos at three locations: Utah, KAUST and DOE booths. Demonstrations all involved remote data visualization streaming from CHPC in Utah, Argonne National Laboratory, and KAUST (Saudi Arabia).
Molecular and Materials Visualizations, in collaboration with Intel and Argonne National Laboratory
|
|
|
|
| Ken-ichi Nomura (University of Southern California), discusses multi-field molecular orbital data from a LiAl-water DFT simulation conducted at Argonne National Laboratory. Live visualizations were shown using ANL's VL3-nanovol software. | Aaron Knoll demonstrates various particle and molecular visualization methods on CPU, GPU, and Intel Xeon Phi, under development as part of the Intel Parallel Computing Center at the University of Utah, and in collaboration with Argonne National Laboratory. | Live remote rendering of multi-timestep 180 million particle detonation-to-deflagration simulation in Uintah, from a large-memory workstation at CHPC. Data courtesy of Jacqueline Beckvermit, University of Utah. Newer Uintah simulations are being computed on Mira at Argonne National Laboratory, as part of the ALCF INCITE program. |
Uintah CFD simulation framework
At SC14 we demonstrated a live run of the Uintah CFD simulation framework using the ash cluster at CHPC to simulate the mixing that occurs when two fluids moving in opposite directions begin to interact. The characteristic effect can be seen when thin layers of clouds in one layer of the atmosphere start to curl into a different velocity region of air above or below them.
Simulation creator Tony Saad from the University of Utah Institute for Clean and Secure Energy (ICSE, Dr. Phil Smith's group), describres the process as follows:
The fluid on top is moving in the (+X, +Y) direction while the fluid at the bottom is moving in the (-X,-Y) direction. The interface is perturbed a bit to trigger the instability, which is reminiscent of what is called the Kelvin-Helmholtz instability that you often see in the atmosphere around clouds although this simulation is at constant density and the streams are moving in opposite directions.
 |
 |
 |
|
|
Screenshots from the demonstation at 128^3 (left) of an advanced timestep of the simulation, and another at 512^3 (right) at an earlier timestep.We ran the simulation at up to 512^3 resolution, outputting four double precision variables at each timestep, which added up to about 10Tb of data produced during the live demos (from KAUST, Saudi Arabia). |
|||
Paper Presentations - publications in Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, IEEE Press
 |
 |
| Aaditya Landge presented "In-situ feature extraction of large scale combustion simulations using segmented merge trees". | Sidharth Kumar presented "Efficient I/O and storage of adaptive-resolution data". |
Parallel IDX I/O library for large-scale simulations
We also presented Parallel IDX (PIDX), a high performance I/O library that stores simulation output in a hierarchical-Z (HZ) ordering, translating data from Cartesian coordinates to a one-dimensional array ordered by locality at different resolution levels.
