Getting nuclear DFT ready for the exascale age

1
Getting nuclear DFT ready for the exascale age
http//unedf.org

• N. Schunck
• Department of Physics ? Astronomy, University of
  Tennessee, Knoxville, TN-37996, USA
• Physics Division, Oak Ridge National Laboratory,
  Oak Ridge, TN-37831, USA

J. Dobaczewski, G. Fann, R. Harrison, J.
McDonnell, W. Nazarewicz, N. Nikolov, H. H. Nam,
J. Pei, J. Sarich, J. Sheikh, W. Shelton, A.
Staszczak, M. Stoitsov
The 3rd LACM-EFES-JUSTIPEN Workshop JIHIR, Oak
Ridge National Laboratory, February 23-25, 2009
2
Nuclear DFT Why supercomputing?
1
DFT A global theory
Principle average out individual degrees of
freedom

• Correlations must be added ad hoc
• Lack of quantitative predictions at the 100 keV
  level

• No limit theory from light nuclei to the
  physics of neutron stars
• Rich physics
• Fast and reliable

Ground-state of even nucleus can be computed in a
matter of minutes on a standard laptop why
bother with supercomputing?

• Why super-computers
• Large-scale problems fission, shape coexistence,
  time-dependent problems
• Systematic restoration of broken symmetries and
  correlations made easy (QRPA, GCM)
• Optimization of extended functionals on larger
  sets of experimental data

Supercomputers DFT at full power
3
Classes of DFT Solvers
2
Non-linear integro-differential fixed point
problem

• Coordinate-space direct integration of the HFB
  equations
• Accurate provide  exact  result
• Slow and CPU/memory intensive for 2D-3D
  geometries
• Configuration space expansion of the solutions
  on a basis (usually HO)
• Fast and amenable to beyond mean-field extensions
• Truncation effects source of divergences/renormal
  ization issues
• Wrong asymptotic unless different bases are used
  (WS, PTG, Gamow, etc.)

Computational package used and developed at ORNL
and estimate of the resources needed for a
standard HFB calculation
4
Recent achievements
3
Microscopic description of nuclear fission
Even-even, odd-even and odd-odd mass tables
Systematics of odd-proton states in odd nuclei
Cf. Talks by N. Nikolov, J. Pei, J. Sheikh, A.
Staszczak, M. Stoitsov, S. Wild and J.
Moré Online ressources http//massexplorer.org/ h
ttp//unedf.org/
5
Petascale and beyond
4

• Hardware constraints
• Many cores (100,000) stacked into sockets -
  Currently 4 cores/socket, evolution toward 8
  cores/socket and more
• Small-memory per core (shared memory per socket)
• Short, crash-prone, expensive runtime
• Consequences on the architecture of DFT solvers
• Optimize time of one HFB calculation reduce
  number of iterations, use symmetries smartly by
  improving/interfacing codes, parallelization,
  etc.
• Work on parallel wrapper load balancing,
  checkpoints, error control mechanisms

6
Computing platform for DFT applications
5
Interfacing codes
Parallelize solver
Load balancing
7
Summary and Outlook
6

• Nuclear DFT adapted to global studies of nuclear
  properties (at the scale of the mass table and
  beyond)
• Existing codes have demonstrated that they can be
  ported and run on current leadership class
  computers.
• Future evolution of super-computing is dictated
  by factors (economical, technological, political)
  that have little to do with nuclear physics we
  must adapt

• Physics
• Get new functionals of spectroscopic quality
• Clarify the role of correlations can they be
  included at the level of DFT? Divergences?

• Computing
• Improve our algorithms faster, more reliable,
  more stable
• Learn how to take advantage of redundancy (load
  balancing)

• Community
• Change the way we program collaborative work
  with CS/AM and among physicists
• Be happy with it brand new physics within reach