-
- Downloads
WIP: Better preconditionners for stack_solve_algo={2,3} (GMRES + BiCGStab solvers)
Implement preconditionners as in UMFITER and ITERSTACK: – For UMFITER, on first iteration, do a full LU decomposition, and use it as a preconditionner in further iterations (similar to UMFITER in TROLL). – For ITERSTACK, build a preconditionner based on small LU decomposition repeated several times across the stack By the way, uniformize the preconditioners for solve_algo={7,8} across the different implementations. TODO: – better understand how the tolerance of GMRES and BiCGStab influence the result (1e-12 seems needed for GIMF in the block-decomposed version) – handle the error flag of gmres and bicgstab? (already done in bytecode) – turn various settings as user-accessible options – add tests for the three preconditioners [skip ci]
Showing
- matlab/default_option_values.m 11 additions, 0 deletionsmatlab/default_option_values.m
- matlab/optimization/newton_solve.m 2 additions, 1 deletionmatlab/optimization/newton_solve.m
- matlab/optimization/solve_one_boundary.m 34 additions, 57 deletionsmatlab/optimization/solve_one_boundary.m
- matlab/perfect-foresight-models/iterstack_preconditioner.m 41 additions, 0 deletionsmatlab/perfect-foresight-models/iterstack_preconditioner.m
- matlab/perfect-foresight-models/sim1.m 34 additions, 10 deletionsmatlab/perfect-foresight-models/sim1.m
- matlab/perfect-foresight-models/solve_block_decomposed_problem.m 2 additions, 2 deletions...perfect-foresight-models/solve_block_decomposed_problem.m
- matlab/perfect-foresight-models/solve_two_boundaries_stacked.m 20 additions, 96 deletions...b/perfect-foresight-models/solve_two_boundaries_stacked.m
- mex/sources/bytecode/Interpreter.cc 97 additions, 259 deletionsmex/sources/bytecode/Interpreter.cc
- mex/sources/bytecode/Interpreter.hh 17 additions, 9 deletionsmex/sources/bytecode/Interpreter.hh
- mex/sources/bytecode/bytecode.cc 32 additions, 2 deletionsmex/sources/bytecode/bytecode.cc
Loading
Please register or sign in to comment