Xolotl: A New Plasma Facing Component Simulator
Download
Report
Transcript Xolotl: A New Plasma Facing Component Simulator
Xolotl: A New
Plasma Facing
Component
Simulator
Scott Forest Hull II
Jr. Software Developer
Oak Ridge National Laboratory
[email protected]
Project Overview:
The overall goal of the project:
•Develop and deploy high performance simulations
capable of predicting the lifetime and durability of
tungsten-based plasma-facing components in a “hot”
fusion plasma environment.
•Modeling surface morphology evolution in erosion or redeposition regimes.
•Recycling of hydrogenic species.
•Xolotl (SHO-lottle) – new code to simulate 3D continuum
reaction-diffusion for long-time morphological and
chemical evolution.
2
Managed by UT-Battelle
for the U.S. Department of Energy
Presentation_name
Project Overview (Continued)
• This code is being built from scratch!
• ORNL tasked with Xolotl Development.
• Funded by Fusion Energy Sciences
(FES) and Advanced Scientific Computing
Research ASCR/SciDAC.
• Participants of various institutions and PI, including
Oak Ridge National Laboratory (ORNL), Los Alamos
National Lab (LANL), Pacific Northwest National
Laboratory (PNNL), Argonne National Laboratory (ANL),
and others.
• Proposal funding is for Fiscal Year 2012-2016.
3
Managed by UT-Battelle
for the U.S. Department of Energy
Presentation_name
Science Lesson
• Develop a new simulator to predict PFC lifetime and
performance.
• Integrating and applying discrete particle-based, as well
as continuum-based, multi-scale modeling techniques
to provide scientific discovery of the mechanism
controlling PFC and bulk materials evolution under
fusion plasma and 14-MeV neutron exposure.
• Will work on various HPC machines, including
Jaguar/Titan – targeting HPC CPU/GPU hybrids as a
main focus.
• Solving coupled reaction-diffusion problem.
4
Managed by UT-Battelle
for the U.S. Department of Energy
Presentation_name
Parallel Programming Model
Exact model determined during design phase, but some
ideas:
•OpenMP, MPI, Pthreads – also investigating OpenACC.
•Languages: C++
•Runtime libraries and frameworks include Integrated
Plasma Simulator (IPS) and Mesh-Oriented datABase
(MOAB).
•Parallel build with CMake and Ctest
•OpenCL and CUDA – explored for new data structures
and reveal tasks within a timestep.
•Hybrid MPI + X – expand spatial scales of the problems.
5
Managed by UT-Battelle
for the U.S. Department of Energy
Presentation_name
Performance and Scaling
• Since this is a new project, bottlenecks and scaling can only be
predicted.
• The programs will scale to a height of the largest available
systems on the current and future market.
• Collect application-specific data at a relatively coarse-grained
level.
• Use tools, like HPCToolKit and mpiP, for application analysis.
6
Managed by UT-Battelle
for the U.S. Department of Energy
Presentation_name
Tools
• Debug with logging and GDB. Valgrind as necessary.
• VisIt and ParaView for visualization tools.
• NEAMS integrated Computational Environment (NiCE)
for leveraging an existing system for data and workflow
management.
• Eclipse IDE, IBM’s Rational Software Architecture (RSA)
for programming environment and development.
• Development Techniques:
– Unified Modeling Language (UML)
– Test Driven Development (TDD)
7
Managed by UT-Battelle
for the U.S. Department of Energy
Presentation_name
Roadmap
• Release 1.0 on Sourceforge of Plasma Facing
Components (PFC) Code within the first year.
• Code Refinement in sequential years for Xolotl:
– Xolotl simulations for PFC surface and near surface
evaluations.
– Xolotl simulations to evaluate bulk neutron-induced defect
cluster and gas bubble response.
• Work on 10,000 cores by year three.
8
Managed by UT-Battelle
for the U.S. Department of Energy
Presentation_name
Questions
Questions? Email me: [email protected]
THANK YOU!
9
Managed by UT-Battelle
for the U.S. Department of Energy
Presentation_name