sub header2

[View PDF]

 

H. Karimabadi, U-California
W. H. Matthaeus, U-Delaware
W. Daughton, LANL
B. Loring, LBNL

Objectives

  • Analyze the full spectrum of heating modalities from those measured in units of Earth radius down to those measured on the scales of electron interactions
  • Interactive remote visualization and analysis of the entire dataset across the entire range of physical scales.

Approach

  • Apply fully kinetic PIC explicit modeling of electrons and ions in hot magnetized plasma to study heating in the solar wind.
  • VPIC ran for ˜72 hr w/ 50k cores on Jaguar to simulate  4x1010 particles on a 16384x8192 grid producing over 7 TB

Cropped region showing coherent structures late in the simulation

  • Coherent Structures, Intermittent Turbulence and Dissipation in High-Temperature Plasmas, Physics of Plasmas, Jan 2013
  • Intermittent Dissipation at Kinetic Scales in Collisionless Plasma Turbulence, Physical Review Letters, Nov 2012
  • Coherent Structures, Intermittent Turbulence and Dissipation in High-Temperature Plasmas, SC13 Scientific Visualization Showcase, Nov 2013
  • In-situ visualization for global hybrid simulations, Proc. XSEDE '13, Aug 2013

Results

  • Developed parallel surface LIC algorithm for data parallel “multi-block” composite datasets supporting VTK and ParaView. Fixed VTK so that it can run  fully with or without GPUs. The work is released in VTK 6.1 and PV 4.1.
  • Our analysis of the unprecedented PIC simulation revealed a previously unknown heating modality: waves driven by motion of coherent structures. We analyzed all of the heating modalities to put the discovery in its proper context
  • “Partnering with the [visualization] group at Berkeley Lab has been critical in developing tools to analyze our data sets. There is an urgent need to develop accurate forecasting models; a severe space-weather event can have dire financial and national security consequences.” lead scientist Dr. H. Karimabadi UCSD.

H. Karimabadi1, W. H. Matthaeus2, W. Daughton3, B. Loring4, et al.

[1] Department of Electrical and Computer Engineering, University of California, San Diego. [2] University of Delaware, Department of Physics and Astronomy. [3] Los Alamos National Laboratory. [4] Lawrence Berkeley National Laboratory

 Notes:

This slide reports the primary scientific result achieved in the collaboration, namely a detailed numerical and visual analysis of the heating modalities in hot turbulent plasmas, such as the solar wind. We also report here on the algorithmic developments in the area of data-parallel vector field visualization which enabled the analysis of the large dataset. Without our work on the algorithmic side the data could not be processed using existing tools (VisIt, ParaView, VTK) it was simply too large for the existing serial algorithm. And, previously published parallel algorithms are incompatible with the data parallel, composite data,  pipelines deployed in VTK, ParaView and VisIt. Our algorithm includes the novel use of image processing filters at various places in the internal pipeline and includes optimizations that significantly improve the remote interactivity necessary for interactive exploration large datasets. For example, the high contrast LIC and bright pseudocoloring achieved in the above figure illustrate the new filtering. This is a cropped region of a 2560x1920 rendering of the full 222M triangle surface, made on Edison interactively using 512 cores. Another quote from the lead scientist, Dr. H Karimabadi: "We are now using the new LIC routinely to search for development of complex structures."