nov. 14, 2012 hank childs, lawrence berkeley jeremy meredith, oak ridge pat mccormick, los alamos...
TRANSCRIPT
Nov. 14, 2012
Hank Childs, Lawrence Berkeley
Jeremy Meredith, Oak Ridge
Pat McCormick, Los Alamos
Chris Sewell, Los Alamos
Ken Moreland, Sandia
Panel at IEEE/ACM SuperComputing 2012
Visualization Frameworks for Multi-Core and Many-core
Architectures
Panel Motivation
A terrible mismatch between what we have and what we will need!
No threading, C++,
parallelism through MPI (but one MPI task
per core)
State of most visualization software
today
Seamless support for multi-core and many-
core nodesVS
Upcoming requirements stemming from HPC
trends
Millions of lines of code, hundreds to thousands person
years of investment.
Multiple new efforts recently started.
This panel’s purpose is to inform about these efforts: their
goals and strategies.
Future Requirements Opinions vary on
requirements. However… Must run on future
architectures. Must be capable of in
situ processing. Must be capable of
supporting massive data sets (scale and complexity).
Fortunately, lessons learned from previous era: Interoperability, data flow networks, data
models, execution models
To date, our community has used a combination of
libraries and tools. Libraries:
Provide data model, execution model and algorithms
Examples: AVS, OpenDX, VTK, more…
Tools: Incorporate libraries (for data model,
execution model, and algorithms) Provide user interface, parallel handling Examples: EnSight, FieldView, ParaView,
VisIt, VAPOR, more…
This software is not vaporware
Tutorial and code sprint, Kitware HQ, Clifton Park, NY,
September 2012
Hedgehogs of gradient fields along an isosurface in PISTON. Implemented by Childs (LBNL) and Sewell (LANL).
Prototype integration of VisIt and DAX, with DAX calculating derived quantities. Implemented by Harrison (LLNL).
Transform operator (UI + functionality) in EAVL/EAVLab. Implemented by Whitlock (LLNL).
Panel Format
Overview (8 minutes) Hank Childs, Lawrence Berkeley
EAVL (16 minutes) Jeremy Meredith, Oak Ridge
DSLs (16 minutes) Pat McCormick, Los Alamos
PISTON (16 minutes) Chris Sewell, Los Alamos
DAX (16 minutes) Ken Moreland, Sandia
Question & Answer (18 minutes)
Questions for the panelists
What fundamental problem are you trying to solve?
What are your plans to deal with exascale-specific issues (massive concurrency, distributed memory, memory overhead, fault tolerance)?
What is your philosophy for dealing with ambiguity of the exascale architecture?
How is your technology implemented? What is the long-term result for this
effort? (Production software? Research prototype?)