Documentation for PISM, the Parallel Ice Sheet Model

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Last revision Both sides next revision
petscoptions [2012/07/21 23:33]
Ed Bueler created from kspdiverged
petscoptions [2012/07/21 23:41]
Ed Bueler
Line 1: Line 1:
 +====== PETSc options appropriate to SSA solves in PISM ======
  
-Jed'​s ​original ​example included an example of comparing which were the efficient solvers for another ​invertible system that was invertible, using 8  MPI processes:+Jed Brown, [[http://​www.mcs.anl.gov/​petsc/​petsc-2/​miscellaneous/​index.html|a PETSc developer]] and an original ​PISM author, gave this example of comparing which were the efficient solvers for a saved, ​invertible ​linear ​system that came from an SSA solve. ​ Using 8  MPI processes:
 <​code>​ <​code>​
 # Additive Schwarz with overlap 1 # Additive Schwarz with overlap 1
-$ time mpiexec -n 8 ./ex10 -f ~/​pism-run/​SSAFD_ksperror_y-4141.petsc -ksp_type gmres -ksp_norm_type unpreconditioned -ksp_pc_side right -pc_type asm -sub_pc_type lu+$ time mpiexec -n 8 ./ex10 -f SSAFD_ksperror.petsc -ksp_type gmres -ksp_norm_type unpreconditioned -ksp_pc_side right -pc_type asm -sub_pc_type lu
 Number of iterations =  19 Number of iterations =  19
 Residual norm 92.3905 Residual norm 92.3905
Line 12: Line 13:
  
 # Umfpack instead of PETSc subdomain solves, slightly faster # Umfpack instead of PETSc subdomain solves, slightly faster
-$ time mpiexec -n 8 ./ex10 -f ~/​pism-run/​SSAFD_ksperror_y-4141.petsc -ksp_type gmres -ksp_norm_type unpreconditioned -ksp_pc_side right -pc_type asm -sub_pc_type lu -sub_pc_factor_mat_solver_package umfpack+$ time mpiexec -n 8 ./ex10 -f SSAFD_ksperror.petsc -ksp_type gmres -ksp_norm_type unpreconditioned -ksp_pc_side right -pc_type asm -sub_pc_type lu -sub_pc_factor_mat_solver_package umfpack
 Number of iterations =  19 Number of iterations =  19
 Residual norm 92.3905 Residual norm 92.3905
Line 21: Line 22:
  
 # More overlap, a little better # More overlap, a little better
-$ time mpiexec -n 8 ./ex10 -f ~/​pism-run/​SSAFD_ksperror_y-4141.petsc -ksp_type gmres -ksp_norm_type unpreconditioned -ksp_pc_side right -pc_type asm -sub_pc_type lu -sub_pc_factor_mat_solver_package umfpack -pc_asm_overlap 2+$ time mpiexec -n 8 ./ex10 -f SSAFD_ksperror.petsc -ksp_type gmres -ksp_norm_type unpreconditioned -ksp_pc_side right -pc_type asm -sub_pc_type lu -sub_pc_factor_mat_solver_package umfpack -pc_asm_overlap 2
 Number of iterations =  14 Number of iterations =  14
 Residual norm 139.529 Residual norm 139.529
Line 30: Line 31:
  
 # No overlap, slower and less robust # No overlap, slower and less robust
-$ time mpiexec -n 8 ./ex10 -f ~/​pism-run/​SSAFD_ksperror_y-4141.petsc -ksp_type gmres -ksp_norm_type unpreconditioned -ksp_pc_side right -pc_type bjacobi -sub_pc_type lu -sub_pc_factor_mat_solver_package umfpack+$ time mpiexec -n 8 ./ex10 -f SSAFD_ksperror.petsc -ksp_type gmres -ksp_norm_type unpreconditioned -ksp_pc_side right -pc_type bjacobi -sub_pc_type lu -sub_pc_factor_mat_solver_package umfpack
 Number of iterations =  32 Number of iterations =  32
 Residual norm 153.677 Residual norm 153.677
Line 38: Line 39:
 sys     ​0m0.916s sys     ​0m0.916s
  
-unpreconditioned ​norm, less accurate¬†+preconditioned ​norm, less accurate¬†
-$ time mpiexec -n 8 ./ex10 -f ~/​pism-run/​SSAFD_ksperror_y-4141.petsc -ksp_type gmres -ksp_norm_type preconditioned -ksp_pc_side left -pc_type bjacobi -sub_pc_type lu -sub_pc_factor_mat_solver_package umfpack+$ time mpiexec -n 8 ./ex10 -f SSAFD_ksperror.petsc -ksp_type gmres -ksp_norm_type preconditioned -ksp_pc_side left -pc_type bjacobi -sub_pc_type lu -sub_pc_factor_mat_solver_package umfpack
 Number of iterations =  18 Number of iterations =  18
 Residual norm 30777.5 Residual norm 30777.5
petscoptions.txt · Last modified: 2012/07/21 23:43 by Ed Bueler
© 2020 by PISM | webmaster