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

TODO:
– better understand how the tolerance of GMRES and BiCGStab influence the
  result (hardcoded to 1e-12, needed for GIMF in the block-decomposed version);
  also put that value in options_ and maybe give it an interface?
– implement ITERSTACK for the block-decomposed version
– do the same in bytecode
– leave ILU as an alternative preconditionner?
– handle the error flag of gmres and bicgstab?
– remove the persistent variables in sim1.m
– turn various settings as user-accessible options

[skip ci]

By the way, rename “stop” variable to “converged” for clarity.

SMC: Add support for generate_trace_plots.m

Closes #1932

See merge request Dynare/dynare!2311

Incidentally:
– fix a bug where path for endogenous after failure to converge was not centered
– ensure that sim1_linear.m is actually tested, by adding a test for
  linear_approximation + stack_solve_algo=0

Ref. #1859

⇒ Generated pseudo-random numbers will not change if we add new unit tests.

Ref. #1859

Ref. #1859

[skip ci]

Ref. !2319

One input argument is required

 forecast and smoothing: disallow unsupported options

See merge request Dynare/dynare!2317

Fixes crash of det_cond_forecast with expected shocks without surprises.

 hssmc.m: make sure no stale files are kept

See merge request Dynare/dynare!2316

 dsge_simulated_theoretical_correlation.m

See merge request Dynare/dynare!2315

Would cause crash in subsequent loading of output files as naming would be different

Make sure ivar is set; usually not triggered due to stored decision rules

rand_multivariate_student.m: suppress spurious output

See merge request Dynare/dynare!2314

Ref. !2313

Marco Ratto authored 7 months ago and Sébastien Villemot committed 7 months ago

- when init2shocks is used, list of exo names must be re-set before retrieving decomposition info for auxiliary variable
- when decomposition results are squeezed, steady state info in oo_ must also be temporarily squeezed, other wise plotted results will be wrong

[skip CI]

The problem has apparently been fixed in some update to R2024a.

This is a revert of commit 18131492.

Remove persistent variables in priordens so that we can safely use SMC
with the parallel toolbox. A more satisfactory solution would be to
use the dprior class, but this is still a work in progress.