run_petscinv

#!/bin/bash

# Solves precomputed tomography matrices using our new petsc solver

mpirun -n 6 ./petscinv \
    -matrix_schedule 'samples/schedule' \
    -inparam_file 'samples/inparam' \
    -inversion_parameters 'vsh,vsv' \
    -reference_model 'prem_ani' \
    -number_of_layers 21 \
    -equatorial_increment 5.0 \
    -project_id 'my_inversion' \
    -output_format 'ascii_rtpv' \
    -solution_type 'normal' \
    -ksp_monitor \
    -grouped_varr \
    -ksp_type gmres



#
#   SOME NOTES:
#
# * Be aware that the rtol=the relative convergence tolerance 
#   (i.e. the relative decrease in the residual norm) value is
#   set based on the experience that for tomographic linear
#   systems nothing significantly changes beyond this criteria
#   One may need to check whether this is really the case before
#   defining a final model
# 
# * If you want to run a synthetic test just add the line. You
#   can either read models parameterized in the actual tomography
#   parameterization or as pointclouds
#   -synthetic_model 'samples/synth.pcn' \
#   -synthetic_model 'samples/synth_rtpv.pcn' \
#   -synthetic_model 'samples/synth_checkerboard.pcn' \

#
# * You can also just test variance reduction using a synthetic
#   input model, as a simple hypothesis test using
#   -varr_model_array 'samples/model_array.in' \
#
# * In case of an adaptive grid add the command line arguments
#   -grid_info 'grid/grid.info.rit.S.d' \
#   -adaptive_grid \
#
# * If you want to store grouped variance reductions for subsets
#   of data, just append the following
#   -grouped_varr \
#
# * You could think about playing around with the preconditioning
#   -pc_type jacobi 
#
# * To use LSQR on the rectangular system just set 
#   -type to 'lsqr' or 'rect'
#
# * To get some additional info on the matrix you invert add the flags
#   -info -mat_view_info -log_summary 
#
# * For testing you can specify a maximum number of iterations
#   -ksp_max_it 1000
#
# * Alternatively you can release the convergence criteria with
#   -ksp_rtol 0.0001 \
#   -ksp_atol 0.0001 \
#  
# * In the case of -type 'normal' please try different solvers
#   as gmres, bicg, cg, symmlq, minres using -ksp_type *
#
# * In case you want to use a 3D reference model, add the following
#   you need a file regions in the same folder
#   -reference_model './samples/prem_ani_crust_2.0/' \