NERSCPowering Scientific Discovery for 50 Years

John Ludlow

FES Requirements Worksheet

1.1. Project Information - Atomic and Molecular Physics for Controlled Fusion Energy

Document Prepared By

John Ludlow

Project Title

Atomic and Molecular Physics for Controlled Fusion Energy

Principal Investigator

John Ludlow

Participating Organizations

Auburn University 
Oak Ridge National Laboratory 
Los Alamos National Laboratory 
University of Strathclyde 
Queen's University of Belfast

Funding Agencies

 DOE SC  DOE NSA  NSF  NOAA  NIH  Other: EAEC, IAEA

2. Project Summary & Scientific Objectives for the Next 5 Years

Please give a brief description of your project - highlighting its computational aspect - and outline its scientific objectives for the next 3-5 years. Please list one or two specific goals you hope to reach in 5 years.

We plan to apply existing and develop new theoretical and computational methods, which will make use of the world's latest advanced computing platforms, to calculate atomic and molecular collision processes of key importance to ongoing research in controlled fusion energy. We are supported by grants from the US Department of Energy, the European Atomic Energy Community, and the International Atomic Energy Agency. We currently have computing allocations at NERSC in Oakland, CA, NICS in Knoxville, TN, NCCS in Oak Ridge, TN, and ALCF in Argonne, IL; with pending proposals to HECToR in Edinburgh, UK and PRACE in Julich, DE. 
 
Over the next five years we shall continue to provide the fundamental collisional rates required for light element fusion related species, which includes electron-impact excitation and ionization of atoms and diatomic molecules and their ions, dielectronic recombination of atomic ions, and heavy particle impact excitation, ionization, and charge transfer with atoms and diatomic molecules. We will also develop our collision codes to gradually progress to heavy element fusion related species, such as Mo, Xe, and W. We will continue to employ a variety of perturbative distorted-wave and non-perturbative close-coupling methods depending on the complexity and the accuracy required for the atomic or molecular system under consideration.

3. Current HPC Usage and Methods

3a. Please list your current primary codes and their main mathematical methods and/or algorithms. Include quantities that characterize the size or scale of your simulations or numerical experiments; e.g., size of grid, number of particles, basis sets, etc. Also indicate how parallelism is expressed (e.g., MPI, OpenMP, MPI/OpenMP hybrid)

Our most computationally demanding codes use the R-Matrix with Pseudo-States (RMPS) and Time-Dependent Close-Coupling (TDCC) methods to solve the Schrodinger and Dirac equations for a range of collisional processes in fusion related atoms and molecules. 
 
The RMPS method is a basis set approach to solving the time-independent Schrodinger and Dirac equations. Based upon accurate structure calculations, the method requires the formation and diagonalization of large symmetric Hermitian matrices for all eigenvalues and eigenvectors. Current problems need up to 50,000 cores. 
 
The TDCC method is a lattice approach to solving the time-dependent Schrodinger and Dirac equations. The method employs a mixture of explicit and implicit propagation techniques on multi-dimensional grids related to the number of active electrons in the problem. The TDCC-2d codes for electron single ionization use 2500 cores, the TDCC-3d cores for electron double ionization use 125,000 cores, the TDCC-4d cores for bare ion double ionization use 50,000 cores, and a planned TDCC-6D for bare ion double charge transfer will need 1,000,000 cores. 
 
Both the RMPS and TDCC suite of codes are written in Fortran using MPI to implement parallelism. The codes are developed within our group and utilize externally written library routines, from Lapack, ScaLapack, NAG, and parallel NAG. Visualization employs library routines from NCAR and GNUPLOT, and MPEG movies for the time-dependent codes. 

3b. Please list known limitations, obstacles, and/or bottlenecks that currently limit your ability to perform simulations you would like to run. Is there anything specific to NERSC?

The RMPS codes will have to explore furthur degrees of parallelism in the formation of the Hamiltonian matrices. There is at least an order of magnitude difference in the scale of the calculations between light fusion related species and heavy species, such as W. 
 
The TDCC codes will evolve to include a higher degree of dimensionality in collisional problems and be extended to address more complicated molecular problems. Lattice calculations involve a high level of communication between processors, especially as these multi-dimensional problems shall be distributed over an even greater number of processors. 

3c. Please fill out the following table to the best of your ability. This table provides baseline data to help extrapolate to requirements for future years. If you are uncertain about any item, please use your best estimate to use as a starting point for discussions.

Facilities Used or Using

 NERSC  OLCF  ACLF  NSF Centers  Other:  

Architectures Used

 Cray XT  IBM Power  BlueGene  Linux Cluster  Other:  

Total Computational Hours Used per Year

 Core-Hours

NERSC Hours Used in 2009

 Core-Hours

Number of Cores Used in Typical Production Run

 

Wallclock Hours of Single Typical Production Run

 

Total Memory Used per Run

 GB

Minimum Memory Required per Core

 GB

Total Data Read & Written per Run

 GB

Size of Checkpoint File(s)

 GB

Amount of Data Moved In/Out of NERSC

 GB per  

On-Line File Storage Required (For I/O from a Running Job)

 TB and  Files

Off-Line Archival Storage Required

 TB and  Files

Please list any required or important software, services, or infrastructure (beyond supercomputing and standard storage infrastructure) provided by HPC centers or system vendors.

LAPACK, ScaLAPACK, NAG, parallel NAG, NCAR 

4. HPC Requirements in 5 Years

4a. We are formulating the requirements for NERSC that will enable you to meet the goals you outlined in Section 2 above. Please fill out the following table to the best of your ability. If you are uncertain about any item, please use your best estimate to use as a starting point for discussions at the workshop.

Computational Hours Required per Year

 

Anticipated Number of Cores to be Used in a Typical Production Run

 

Anticipated Wallclock to be Used in a Typical Production Run Using the Number of Cores Given Above

 

Anticipated Total Memory Used per Run

 GB

Anticipated Minimum Memory Required per Core

 GB

Anticipated total data read & written per run

 GB

 

 

Anticipated size of checkpoint file(s)

 GB

Anticipated Amount of Data Moved In/Out of NERSC

 GB per  

Anticipated On-Line File Storage Required (For I/O from a Running Job)

 TB and  Files

Anticipated Off-Line Archival Storage Required

 TB and  Files

4b. What changes to codes, mathematical methods and/or algorithms do you anticipate will be needed to achieve this project's scientific objectives over the next 5 years.

The use of parallel I/O. Further reduction in memory footprint of codes per core.

4c. Please list any known or anticipated architectural requirements (e.g., 2 GB memory/core, interconnect latency < 1 μs).

At least 8 GB memory per core will be needed for the RMPS codes. Low latency will be important for the TDCC codes to ensure efficient communication.

4d. Please list any new software, services, or infrastructure support you will need over the next 5 years.

We need ScaLAPACK routines for matrix diagonalization of non-symmetric real and complex matrices. We would also like to have ScaLAPACK FFT routines to supplement the current NAG parallel FFT routines. 

4e. It is believed that the dominant HPC architecture in the next 3-5 years will incorporate processing elements composed of 10s-1,000s of individual cores, perhaps GPUs or other accelerators. It is unlikely that a programming model based solely on MPI will be effective, or even supported, on these machines. Do you have a strategy for computing in such an environment? If so, please briefly describe it.

We shall continue to experiment with the OPEN_MP libraries to assess whether they help with nested Fortran do loops for small numbers of cores. For larger number of cores and GPUs,we will begin experiments with the CUDA Fortran approach to nested Fortran do loops. 

New Science With New Resources

To help us get a better understanding of the quantitative requirements we've asked for above, please tell us: What significant scientific progress could you achieve over the next 5 years with access to 50X the HPC resources you currently have access to at NERSC? What would be the benefits to your research field if you were given access to these kinds of resources?

Please explain what aspects of "expanded HPC resources" are important for your project (e.g., more CPU hours, more memory, more storage, more throughput for small jobs, ability to handle very large jobs).

With access to 50 times the current HPC resources, electron-impact excitation and ionization of the near neutral stages of heavy elements such as Mo, Xe, and W can be studied. This will benefit impurity transport research on existing tokamaks and future machines like ITER. Bare ion collisions with atoms and molecules, including double charge transfer, would also be possible. 
 
An order of magnitude more CPU hours, large memory nodes, large scratch space, and quicker throughput for large jobs are all important for future computational research.