Presentation is loading. Please wait.

Presentation is loading. Please wait.

CMRS Review, PPPL, 5 June 2003 &. 4 projects in high energy and nuclear physics 5 projects in fusion energy science 14 projects in biological and environmental.

Similar presentations


Presentation on theme: "CMRS Review, PPPL, 5 June 2003 &. 4 projects in high energy and nuclear physics 5 projects in fusion energy science 14 projects in biological and environmental."— Presentation transcript:

1 CMRS Review, PPPL, 5 June 2003 &

2 4 projects in high energy and nuclear physics 5 projects in fusion energy science 14 projects in biological and environmental research 10 projects in basic energy sciences “The partial differential equation entered theoretical physics as a handmaid, but has gradually become mistress.” – A. Einstein & ISICs PDEs are dense in the SciDAC portfolio

3 CMRS Review, PPPL, 5 June 2003 Scope for TOPS l Design and implementation of “solvers” for PDE-derived systems n Linear solvers n Eigensolvers n Nonlinear solvers n Time integrators n Optimizers l Software integration l Performance optimization Optimizer Linear solver Eigensolver Time integrator Nonlinear solver Indicates dependence Sens. Analyzer (w/ sens. anal.)

4 CMRS Review, PPPL, 5 June 2003 Status of CMRS-TOPS collaboration CMRS team has provided TOPS with discretization of model 2D multicomponent Hall magnetic reconnection evolution code in PETSc ’s DA/DMMG format, using automatic differentiation for Jacobian objects l TOPS has implemented fully nonlinearly implicit Newton-GMRES-MG-SOR parallel solver (with deflation of nullspace in CMRS’s doubly periodic formulation) l Both first- and second-order implicit temporal integration available l CMRS and TOPS reproduce the same dynamics on the same grids with the same time-stepping, up to a finite-time singularity due to collapse of current sheet (that falls below presently uniform mesh resolution) l TOPS code, being implicit, can choose timesteps an order of magnitude larger, with potential for higher ratio in more physically realistic parameter regimes, though it is slower in wall-clock time for small CFL Plan: tune PETSc solver by profiling, blocking, reuse, etc. l Plan: identify the numerical complexity benefits from implicitness (in suppressing fast timescales) and quantify (explicit versus implicit) l Plan (with APDEC team): incorporate AMR

5 CMRS Review, PPPL, 5 June 2003 Equilibrium: Model equations: (Porcelli et al., 1993, 1999) 2D Hall MHD sawtooth instability ( PETSc examples /snes/ex29.c and /sles/ex31.c ) figures c/o A. Bhattacharjee, CMRS Vorticity, early time Vorticity, later time zoom

6 CMRS Review, PPPL, 5 June 2003 PETSc ’s DMMG in Hall MR application l Mesh and time refinement studies of CMRS Hall magnetic reconnection model problem (4 mesh sizes, dt=0.1 (near CFL limit for fastest wave) on left, dt=0.8 on right) l Measure of functional inverse to thickness of current sheet versus time, for 0<t<200 (nondimensional), where singularity occurs around t=215

7 CMRS Review, PPPL, 5 June 2003 PETSc ’s DMMG in Hall MR app., cont. l Implicit timestep increase studies of CMRS Hall magnetic reconnection model problem, on finest (192  192) mesh of previous slide, in absolute magnitude, rather than semi-log

8 CMRS Review, PPPL, 5 June 2003 Newton-Krylov-Schwarz – a parallel PDE “workhorse” Newton nonlinear solver asymptotically quadratic Krylov accelerator spectrally adaptive Schwarz preconditioner parallelizable

9 CMRS Review, PPPL, 5 June 2003 PETSc code CMRS code Application Initialization Function Evaluation Jacobian Evaluation Post- Processing PCKSP PETSc Main Routine Linear Solvers (SLES) Nonlinear Solvers (SNES) Timestepping Solvers (TS) CMRS/PETSc Library Interactions ADIC generated code


Download ppt "CMRS Review, PPPL, 5 June 2003 &. 4 projects in high energy and nuclear physics 5 projects in fusion energy science 14 projects in biological and environmental."

Similar presentations


Ads by Google