It would previously forget to write the indices for the highest derivation
order (e.g. if order=3, then it would only write sparce indices up to 2nd
order).
Ref. dynare#1859
Such a decomposition has to be simulated with periods as the outer loop and
blocks as the inner loop.
It is enabled by default for purely backward/forward/static models, as long as
the “block” option is not given. In that case, “mfs” is also set to 3 by
default (until that value becomes the global default).
M_.time_recursive_block_decomposition is set to “true” when that decomposition
has been performed, “false” otherwise for the traditional decomposition (the
latter has to be simulated with blocks as the outer loop and periods as the
inner loop).
The files are created under <basename>/+debug/dynamic_resid.m and
<basename>/+debug/static_resid.m.
Their purpose is to evaluate separately the LHS and RHS of each equation.
The new representation is only supported for MATLAB/Octave, C and Julia output
for the time being. Bytecode and JSON are unsupported.
This commit adds new fields in M_.
This is a preliminary step for dynare#1859.
The previous system would spawn as many threads as there are object files to be
compiled (which could lead to hundreds of threads for large block-decomposed
models). This could pose a memory usage problem (even when not just waiting,
threads require memory for their own stack).
– DataTree::packageDir() now takes a std::string_view, returns a
std::filesystem::path, and no longer creates that directory
– DataTree::writeToFileIfModified() now takes a std::filesystem::path as
argument
– Do not call DataTree::writeToFileIfModified() for generating MATLAB/Octave
files, since it does not work (the directory inside which the file is written
has been deleted by the preprocessor just before)
– Consistently use DataTree::packageDir() everywhere (for compatibility with
planner_objective)
Currently two threads are used (one for the dynamic MEX, one for the static
MEX). When the sparse representation is implemented, four threads will be used.
Closes: #41
This is unsafe since the find() method can return a past-the-end iterator,
which should be tested for.
Replace most instances by calls to the std::map::at() method (which throws if
the key is unknown), and which is incidentally more readable.
Elements of params_derivatives could be accessed without them being defined,
thus causing illegal memory read.
Ensure that these elements always exist. By the way, use std::map::at() instead
of std::map::find() to trigger an exception instead of an illegal memory read
in that case.
When computing the derivatives or block decomposition of the planner objective,
the epilogue or the original Ramsey model, the preprocessor would talk about
dynamic/static model, which was confusing. It now uses the right terminology.
If block decomposition fails, error out if “block” option was passed, but not
otherwise.
This commit does not modify the generated files.
This is a preliminary step for dynare#1859.
There were actually two distinct bugs, leading to incorrect results in some
corner cases:
– in the “evaluate” mode of the bytecode MEX, the temporary terms of the
derivatives “evaluate” blocks were not evaluated at runtime; but these
temporary terms may be needed for residuals of subsequent blocks;
– when the bytecode MEX was only computing residuals of the model (and not 1st
order derivatives), the temporary terms of the derivatives were not evaluated
at runtime; but these temporary terms may be needed for residuals of subsequent
blocks.
The “temporary_terms_union.insert(it)” statement introduced in the previous
commit was in the wrong block.
By the way, replace the “switch” by an “if constexpr”.
– factorize common code between the static and the dynamic version
– reorganise language-specific code into dedicated functions
– use a function template in the main helper to do some computations
at compile-time (using constexpr features)
We follow the advice given by Josuttis in his book about move semantics.
Deleting those member fuctions can be a bug if we want to allow copy semantics,
because overload resolution will no longer fallback to the copy
constructor/assignment operator when given an rvalue.
In particular, this explain why it was not possible to delete the move
assignment operator of the StaticModel class.
Sébastien Villemot
MichelJuillard
Willi Mutschler