HMLQCD

From HP-SEE Wiki

(Difference between revisions)
Jump to: navigation, search
(Short Description)
(Short Description)
Line 26: Line 26:
QCD lattices must be scaled appropriately in order to get physical results. This scalability requirement needs HPC techniques.
QCD lattices must be scaled appropriately in order to get physical results. This scalability requirement needs HPC techniques.
-
The project will test for the first time local chiral actions for the calculation of the hadron masses.
+
The project will test for the first time local chiral actions for the calculation of the hadron masses. On the algorithmic side the project will test new solvers for overlap and domain wall fermions.
== Problems Solved ==
== Problems Solved ==

Revision as of 14:05, 11 July 2011

Contents

General Information

  • Application's name: Hadron Masses from Lattice QCD
  • Virtual Research Community: Computational Physics
  • Scientific contact: Artan Boriçi
  • Technical contact: Artan Boriçi
  • Developers: A.Boriçi, D.Xhako, R.Zeqirllari
  • Web site: -

Short Description

LQCD is a quantum field theory whose correlation functions are described by means of vacuum expectation values. These are path integrals whose measure is defined on four dimensional hypercubic lattices. The computation of path integrals is performed via Markov Chain Monte Carlo sampling of the underlying positive definite measure. The lattice QCD measure is a non-local function on the degrees of freedom which makes the evolution in configuration space very slow with large autocorrelation times of certain observables. At any Markov step several huge ans sparse linear systems have to be solved. Once the gauge field configurations are produced, one stores them in the disk for further analysis. The mass spectrum analysis involves computation of quark propagators, which are the solution of huge linear systems of Dirac operators on the lattice. As a typical example on 32^3 by 64 lattices one needs thousands of Monte Carlo steps to compute one statistically independent configuration. One Krylov solver needs typically hundreds of iterations and one multiplication by the Wilson-Dirac operator needs 1Gflops. QCD lattices must be scaled appropriately in order to get physical results. This scalability requirement needs HPC techniques.

The project will test for the first time local chiral actions for the calculation of the hadron masses. On the algorithmic side the project will test new solvers for overlap and domain wall fermions.

Problems Solved

Lattice QCD has become an indispensable tool both for particle and nuclear physics. It has fundamental role in describing elementary particle interactions from first principles.

Scientific and Social Impact

Using local chiral fermions saves two orders of magnitude computing resources which is translated in physical results with unprecedented accuracy close to the chiral limit. Increased social support for scientific communities.

Collaborations and Beneficiaries

- TIRLatt (Tirana lattice QCD group) - International lattice QCD groups will benefit from expanded libraries of QCDLAB and fermiQCD

Technical Features and HP-SEE Implementation

  • Primary programming language: C/C++/Matlab/Octave (Matlab compiler)
  • Parallel programming paradigm: MPI/MPITB(Octave)/Parallel Computing Toolbox (Matlab)
  • Main parallel code: C/C++/Matlab
  • Pre/post processing code: -
  • Application tools and libraries:
  • Number of cores required: Limited to available number of CPUs
  • Minimum RAM/core required: $GB
  • Storage space during a single run: 1TB
  • Long-term data storage: 2TB

Usage Example

-

Publications

  • ...
Personal tools