SLIDE 1
April 4-7, 2016 | Silicon Valley
Tom Fogal, 4/6/2016
2
NAVIGATING THE IN SITU VISUALIZATION LANDSCAPE Tom Fogal, 4/6/2016 - - PowerPoint PPT Presentation
NAVIGATING THE IN SITU VISUALIZATION LANDSCAPE Tom Fogal, 4/6/2016 - - PowerPoint PPT Presentation
April 4-7, 2016 | Silicon Valley NAVIGATING THE IN SITU VISUALIZATION LANDSCAPE Tom Fogal, 4/6/2016 VISUALIZATION & ANALYSIS 22 27 31 38 42 17 22 27 31 38 13 17 22 27 31 6 13 17 22 27 2 6 13 17 22 2 POST HOC VISUALIZATION AND ANALYSIS
SLIDE 2
SLIDE 3
3
POST HOC VISUALIZATION AND ANALYSIS
Compute timestep Write to disk Compute timestep Compute timestep Write to disk Vis & analysis
…
SLIDE 4
4
IN SITU VISUALIZATION & ANALYSIS
Compute timestep Write to disk Compute timestep Vis & analysis Write to disk Vis & analysis Compute timestep Write to disk Vis & analysis Compute timestep
…
SLIDE 5
5
PYFR IN SITU
“When running PyFR at scale, it generates very large data sets that need analyzing for acoustics. The traditional post hoc method is simply not fit for purpose – in situ visualization and processing are
- critical. We see a potential for 50x
speedups with in situ, which significantly accelerates our scientific discovery”
First prototype of ParaView in-situ visualization capabilities in pyFR (CFD) simulations, predicting jet engine acoustics Both compute and visualization running
- n Titan GPUs and streaming to a remote
location
- Dr. Peter Vincent
Imperial College Courtesy A.S. Ayer
SLIDE 6
6
HOW?
SLIDE 7
7
IN SITU AXES
In situ terminology project: Childs et al.
Integration type Proximity Access Division of execution Operation controls
SLIDE 8
11
VISIT’S LIBSIM
Local Computer Remote Supercomputer
Rank 2 Rank 3 Rank 0 Rank 1
Simulation Libsim VisIt runtime library Data Adaptor Commands
Geometry & images Images
Viewer
SLIDE 9
12
LIBSIM AUTOMATIC IN SITU ADDITIONS
Timestep Changed() Update Plots() Save Window() Simulate
Y N
VisIt Is Connect ed()?
SLIDE 10
13
LIBSIM CONTROL
Simulate AttemptTo Complete Connection() set callbacks Process Engine Command()
1 2
VisIt Detect Input()
SLIDE 11
15
PARAVIEW’S CATALYST
Solver Adaptor
SLIDE 12
16
CATALYST AUTOMATIC IN SITU ADDITIONS
Simulate Processor:: New() (meta)Data Description setup Convert to VTK data Data Descript ? Data Description SetGrid(vtk) Processor:: CoProcess
Y N
SLIDE 13
17
PARAVIEW CINEMA
Θ φ
SLIDE 14
18
NVIDIA INDEX
SLIDE 15
19
IN SITU VIA CUDA IPC
Zero-copy sharing between processes
Simulation process Simulation data Visualization process
SLIDE 16
20
DAMARIS/VIZ, ADIOS
SLIDE 17
21
VTK-M
libsim
Simulations
GUI / Parallel Management Base Vis Library In Situ Vis Library
Catalyst
ParaView
SLIDE 18
22
FURTHER RESOURCES
http://visitusers.org/index.php?title=VisIt-tutorial-in-situ http://www.paraview.org/in-situ/ https://users.nccs.gov/~pnorbert/ADIOS-VisualizationSchema-1.1.pdf http://damaris.gforge.inria.fr/ http://www.nvidia-arc.com/products/index.html
SLIDE 19
23
OTHER TALKS OF INTEREST
10:00 Wednesday: HPC visualization hangout, Pod B 15:00 Wednesday: VTK-m, LL21D 16:00 Wednesday: Raytracing (Sparse) Scientific data in OptiX™, LL21D 10:30 Thursday: Marriot Salon 1 15:00 Thursday: Marriot Salon 1
SLIDE 20
April 4-7, 2016 | Silicon Valley
THANK YOU
JOIN THE NVIDIA DEVELOPER PROGRAM AT developer.nvidia.com/join