Transcript edwards_talk_scidac_09 - JLab Computer Center

Nuclear Physics in the SciDAC Era
Robert Edwards
Jefferson Lab
SciDAC 2009
Comparison of Chemistry & QCD : K. Wilson (1989 Capri):
“lattice gauge
theory could also require a
108 increase in computer power AND
spectacular algorithmic advances before useful interactions with
experiment ...”
• ab initio Chemistry
• ab initio QCD
1. 1930+50 = 1980
2. 0.1 flops  10 Mflops
3. Gaussian Basis functions
1. 1980 + 50 = 2030?*
2. 10 Mflops  1000 Tflops
3. Clever Multi-scale Variable?
“Almost 20 Years ahead of schedule!”
*Fast
Computers + Smart Algorithms + Rigorous QCD Theoretical Analysis
= ab initio predictions
Forces in Standard Model
Nuclei Weak
N=2 (Isospin)
Atoms: Maxwell
N=1(charge)
proton
electron
+
neutron
+
quarks
Standard Model: U(1) £ SU(2) £ SU(3)
Sub nuclear: Strong
N=3 (Color)
Quantum Chromo Dynamics - QCD
• QED: theory of electromagnetism
• QCD: theory of strong interactions – hadronic physics
QED
QCD
Photon, 
Gluons, G
Charged particles, e, , u, d,…
Quarks: u, d, s, c, b, t
2 charges: positive &
negative
Photon is neutral
e ' 1/137
3 charges: “red”, “green”,
“blue”
Gluons carry color charge
s ' O(1)
• Highly non-linear theory – can only use perturbation theory at high
energy
Quark+anti-Quark in Meson
Energy in glue
3 Color  3 quarks in Proton
QCD
• QCD: Dirac operator: Aº (vector
potential), m (mass), °º (4x4 matrices)
• Lattice QCD: finite difference
• Probability measure:
• Observables:
Gauge generation
How to produce gauge fields?
• Hamilton’s eq’s - 1st order coupled
diff. eq’s)
• Bummer!
– Must be “reversible”
– No adaptive time steps
Momentum
Total energy in
gauge/quark
fields
Cost Scaling
• Cost: reasonable statistics, box size and “physical” pion mass
• Extrapolate in lattice spacings: 10 ~ 100 PF-yr
PF-years
USQCD National Effort
• US Lattice QCD effort: Jefferson Laboratory, BNL and FNAL
FNAL
BNL
Weak matrix
elements
RHIC Physics
JLAB
Hadronic Physics
SciDAC – R&D Vehicle
Cluster Prototyping
Software R&D
Impact on DOE’s Nuclear Physics Program
SciDAC Software
QCD friendly API’s/libs
• Application codes
• High level (Linpack-like)
• Data parallel (C/C++)
• Linear algebra,
threading, comms
• Code generators
http://www.usqcd.org
QDP/C++ Expressions
Can form expressions:
ci(x) = Uij(x+nu) bj(x) + 2 di(x) for all sites x
QDP++ code (data-parallel)
multi1d<LatticeColorMatrix> U(Nd);
LatticeFermion c, b, d;
int nu, mu;
c = shift(u[mu],FORWARD,nu)*b + 2*d;
Template based
Shifts use QMP for face comms
Level-1 BLAS-like linear algebra core
Critical code: Dirac operator/inverter
• Critical codes: develop special API and libraries
• Example: Dirac operator
[
]
Threading/Multi-core
• Hybrid threads/MPI
• Impact:
– Coalesce
messages
– Better perf.
– Cachecoherency
latency
EXPENSIVE
Scaling on Cray XT4 (ORNL)
Socket level threading improved performance
threads+mpi
mpi
Work involving RENCI
Acceleration
• Deflation & multigrid – big speedups
JLab/W&M (SciDAC) + TOPS
Nuclear Physics & Jefferson Lab
CD-3
JLab Receives DOE Approval to Start Construction of
$310 Million Upgrade
•
•
Lab doubling beam energy
Adding new experimental Hall
Nuclear Structure
• Fundamental questions
– Size, shape, distribution of charge and current in hadrons
– Quark and gluon distributions
– How does nucleon spin arise from quarks and gluons?
– What role do strange quarks play in nucleon structure?
• Status
– Basic nucleon properties calculated with 5-10% precision.
– Pursuing higher precision, more demanding properties.
• NP 2014 milestone
– Perform lattice calculations in full QCD of nucleon form
factors, low moments of nucleon structure functions and low
moments of generalized parton distributions, including
flavor and spin dependence.
Nuclear Structure
Spin of the proton?
~41% quark spin (u+d)
~0% orbital
So: ~59% from glue
(&/or strange)
Most of mass & spin
not from quarks
Caveats:
• Missing terms
(disconnected)
Phys. Rev. D77 094502
Spectroscopy
Spectroscopy reveals fundamental aspects of hadronic physics.
– Essential degrees of freedom?
– Gluonic excitations in mesons - exotic states of matter?
• Status.
– Can extract excited nucleon energies & identify spins,
– Pursuing calculations in full QCD with realistic quark
masses.
• Crucial complement to 12 GeV program at JLab.
– Excited nucleon spectroscopy.
– GlueX: flagship search for gluonic excitations.
Nucleon spectrum
NP2012 milestone:
• Spectrum & E&M transitions up
to Q2 = 7 GeV2
Highly excited energies:
First ever lattice calculation
Pattern of states ->
Future work:
– Separate out decays
– Move to physical regime
Possible 5/2- state
½+
3/2+
5/2+
½-
Phys. Rev. D79 034505
3/2-
5/2-
Exotic matter?
QED
QCD
Can we observe exotic matter? Excited string
Spectroscopy
• Charmonium excited spectrum: J-+
• Exotic matter (1-+) radiative decay: large
Unknown in
experiment
GeV
Phys. Rev. D77 034501
& to appear PRD
If true with light quarks: Can observe at future JLab Hall D!!
Outlook
• Software infrastructure developed for Lattice QCD
– Enabled effective utilization of INCITE resources
• Lattice QCD’s impact on Nuclear Physics
– Nucleon structure (protons, neutrons)
– Spectroscopy
• Results relevant to U.S. DOE experimental programs
• Unifying Nuclear Physics research