Block decomposition: further streamlining of function prototypes

parent bd6eee93
......@@ -4832,15 +4832,15 @@ DynamicModel::computingPass(bool jacobianExo, int derivsOrder, int paramsDerivsO
tie(blocks, equation_lag_lead, variable_lag_lead, n_static, n_forward, n_backward, n_mixed)
= select_non_linear_equations_and_variables(is_equation_linear);
equation_type_and_normalized_equation = equationTypeDetermination(first_order_endo_derivatives, 0);
equationTypeDetermination(first_order_endo_derivatives, 0);
prologue = 0;
epilogue = 0;
block_type_firstequation_size_mfs = reduceBlocksAndTypeDetermination(blocks, equation_type_and_normalized_equation, n_static, n_forward, n_backward, n_mixed, linear_decomposition);
reduceBlocksAndTypeDetermination(blocks, equation_type_and_normalized_equation, n_static, n_forward, n_backward, n_mixed, linear_decomposition);
computeChainRuleJacobian();
blocks_linear = BlockLinear();
determineLinearBlocks();
collect_block_first_order_derivatives();
......@@ -4861,7 +4861,7 @@ DynamicModel::computingPass(bool jacobianExo, int derivsOrder, int paramsDerivsO
first_order_endo_derivatives = collect_first_order_derivatives_endogenous();
equation_type_and_normalized_equation = equationTypeDetermination(first_order_endo_derivatives, mfs);
equationTypeDetermination(first_order_endo_derivatives, mfs);
cout << "Finding the optimal block decomposition of the model ..." << endl;
......@@ -4869,13 +4869,13 @@ DynamicModel::computingPass(bool jacobianExo, int derivsOrder, int paramsDerivsO
tie(blocks, equation_lag_lead, variable_lag_lead, n_static, n_forward, n_backward, n_mixed) = computeBlockDecompositionAndFeedbackVariablesForEachBlock(static_jacobian, equation_type_and_normalized_equation, false, true);
block_type_firstequation_size_mfs = reduceBlocksAndTypeDetermination(blocks, equation_type_and_normalized_equation, n_static, n_forward, n_backward, n_mixed, linear_decomposition);
reduceBlocksAndTypeDetermination(blocks, equation_type_and_normalized_equation, n_static, n_forward, n_backward, n_mixed, linear_decomposition);
printBlockDecomposition(blocks);
computeChainRuleJacobian();
blocks_linear = BlockLinear();
determineLinearBlocks();
collect_block_first_order_derivatives();
......
......@@ -728,13 +728,14 @@ ModelTree::computePrologueAndEpilogue(const jacob_map_t &static_jacobian_arg)
}
}
equation_type_and_normalized_equation_t
ModelTree::equationTypeDetermination(const map<tuple<int, int, int>, expr_t> &first_order_endo_derivatives, int mfs) const
void
ModelTree::equationTypeDetermination(const map<tuple<int, int, int>, expr_t> &first_order_endo_derivatives, int mfs)
{
expr_t lhs;
BinaryOpNode *eq_node;
EquationType Equation_Simulation_Type;
equation_type_and_normalized_equation_t V_Equation_Simulation_Type(equations.size());
equation_type_and_normalized_equation.clear();
equation_type_and_normalized_equation.resize(equations.size());
for (unsigned int i = 0; i < equations.size(); i++)
{
int eq = equation_reordered[i];
......@@ -768,9 +769,8 @@ ModelTree::equationTypeDetermination(const map<tuple<int, int, int>, expr_t> &fi
}
}
}
V_Equation_Simulation_Type[eq] = { Equation_Simulation_Type, dynamic_cast<BinaryOpNode *>(res.second) };
equation_type_and_normalized_equation[eq] = { Equation_Simulation_Type, dynamic_cast<BinaryOpNode *>(res.second) };
}
return V_Equation_Simulation_Type;
}
pair<lag_lead_vector_t, lag_lead_vector_t>
......@@ -1072,14 +1072,14 @@ ModelTree::printBlockDecomposition(const vector<pair<int, int>> &blocks) const
<< " and " << Nb_feedback_variable << " feedback variable(s)." << endl;
}
block_type_firstequation_size_mfs_t
void
ModelTree::reduceBlocksAndTypeDetermination(const vector<pair<int, int>> &blocks, const equation_type_and_normalized_equation_t &Equation_Type, const vector<unsigned int> &n_static, const vector<unsigned int> &n_forward, const vector<unsigned int> &n_backward, const vector<unsigned int> &n_mixed, bool linear_decomposition)
{
int i = 0;
int count_equ = 0, blck_count_simult = 0;
int Blck_Size, MFS_Size;
int Lead, Lag;
block_type_firstequation_size_mfs_t block_type_size_mfs;
block_type_firstequation_size_mfs.clear();
BlockSimulationType Simulation_Type, prev_Type = UNKNOWN;
int eq = 0;
unsigned int l_n_static = 0, l_n_forward = 0, l_n_backward = 0, l_n_mixed = 0;
......@@ -1174,8 +1174,8 @@ ModelTree::reduceBlocksAndTypeDetermination(const vector<pair<int, int>> &blocks
if (i > 0)
{
bool is_lead = false, is_lag = false;
int c_Size = get<2>(block_type_size_mfs[block_type_size_mfs.size()-1]);
int first_equation = get<1>(block_type_size_mfs[block_type_size_mfs.size()-1]);
int c_Size = get<2>(block_type_firstequation_size_mfs[block_type_firstequation_size_mfs.size()-1]);
int first_equation = get<1>(block_type_firstequation_size_mfs[block_type_firstequation_size_mfs.size()-1]);
if (c_Size > 0 && ((prev_Type == EVALUATE_FORWARD && Simulation_Type == EVALUATE_FORWARD && !is_lead)
|| (prev_Type == EVALUATE_BACKWARD && Simulation_Type == EVALUATE_BACKWARD && !is_lag)))
{
......@@ -1193,41 +1193,40 @@ ModelTree::reduceBlocksAndTypeDetermination(const vector<pair<int, int>> &blocks
|| (prev_Type == EVALUATE_BACKWARD && Simulation_Type == EVALUATE_BACKWARD && !is_lag))
{
//merge the current block with the previous one
BlockSimulationType c_Type = get<0>(block_type_size_mfs[block_type_size_mfs.size()-1]);
BlockSimulationType c_Type = get<0>(block_type_firstequation_size_mfs[block_type_firstequation_size_mfs.size()-1]);
c_Size++;
block_type_size_mfs[block_type_size_mfs.size()-1] = { c_Type, first_equation, c_Size, c_Size };
if (block_lag_lead[block_type_size_mfs.size()-1].first > Lag)
Lag = block_lag_lead[block_type_size_mfs.size()-1].first;
if (block_lag_lead[block_type_size_mfs.size()-1].second > Lead)
Lead = block_lag_lead[block_type_size_mfs.size()-1].second;
block_lag_lead[block_type_size_mfs.size()-1] = { Lag, Lead };
block_type_firstequation_size_mfs[block_type_firstequation_size_mfs.size()-1] = { c_Type, first_equation, c_Size, c_Size };
if (block_lag_lead[block_type_firstequation_size_mfs.size()-1].first > Lag)
Lag = block_lag_lead[block_type_firstequation_size_mfs.size()-1].first;
if (block_lag_lead[block_type_firstequation_size_mfs.size()-1].second > Lead)
Lead = block_lag_lead[block_type_firstequation_size_mfs.size()-1].second;
block_lag_lead[block_type_firstequation_size_mfs.size()-1] = { Lag, Lead };
auto tmp = block_col_type[block_col_type.size()-1];
block_col_type[block_col_type.size()-1] = { get<0>(tmp)+l_n_static, get<1>(tmp)+l_n_forward, get<2>(tmp)+l_n_backward, get<3>(tmp)+l_n_mixed };
}
else
{
block_type_size_mfs.emplace_back(Simulation_Type, eq, Blck_Size, MFS_Size);
block_type_firstequation_size_mfs.emplace_back(Simulation_Type, eq, Blck_Size, MFS_Size);
block_lag_lead.emplace_back(Lag, Lead);
block_col_type.emplace_back(l_n_static, l_n_forward, l_n_backward, l_n_mixed);
}
}
else
{
block_type_size_mfs.emplace_back(Simulation_Type, eq, Blck_Size, MFS_Size);
block_type_firstequation_size_mfs.emplace_back(Simulation_Type, eq, Blck_Size, MFS_Size);
block_lag_lead.emplace_back(Lag, Lead);
block_col_type.emplace_back(l_n_static, l_n_forward, l_n_backward, l_n_mixed);
}
}
else
{
block_type_size_mfs.emplace_back(Simulation_Type, eq, Blck_Size, MFS_Size);
block_type_firstequation_size_mfs.emplace_back(Simulation_Type, eq, Blck_Size, MFS_Size);
block_lag_lead.emplace_back(Lag, Lead);
block_col_type.emplace_back(l_n_static, l_n_forward, l_n_backward, l_n_mixed);
}
prev_Type = Simulation_Type;
eq += Blck_Size;
}
return block_type_size_mfs;
}
vector<bool>
......@@ -1248,11 +1247,12 @@ ModelTree::equationLinear(const map<tuple<int, int, int>, expr_t> &first_order_e
return is_linear;
}
vector<bool>
ModelTree::BlockLinear() const
void
ModelTree::determineLinearBlocks()
{
unsigned int nb_blocks = getNbBlocks();
vector<bool> blocks_linear(nb_blocks, true);
blocks_linear.clear();
blocks_linear.resize(nb_blocks, true);
for (unsigned int block = 0; block < nb_blocks; block++)
{
BlockSimulationType simulation_type = getBlockSimulationType(block);
......@@ -1291,7 +1291,6 @@ ModelTree::BlockLinear() const
the_end:
;
}
return blocks_linear;
}
int
......
......@@ -295,13 +295,13 @@ protected:
//! Search the equations and variables belonging to the prologue and the epilogue of the model
void computePrologueAndEpilogue(const jacob_map_t &static_jacobian);
//! Determine the type of each equation of model and try to normalized the unnormalized equation using computeNormalizedEquations
equation_type_and_normalized_equation_t equationTypeDetermination(const map<tuple<int, int, int>, expr_t> &first_order_endo_derivatives, int mfs) const;
void equationTypeDetermination(const map<tuple<int, int, int>, expr_t> &first_order_endo_derivatives, int mfs);
//! Compute the block decomposition and for a non-recusive block find the minimum feedback set
/*! Returns a tuple (blocks, equation_lag_lead, variable_lag_lead, n_static,
n_forward, n_backward, n_mixed) */
tuple<vector<pair<int, int>>, lag_lead_vector_t, lag_lead_vector_t, vector<unsigned int>, vector<unsigned int>, vector<unsigned int>, vector<unsigned int>> computeBlockDecompositionAndFeedbackVariablesForEachBlock(const jacob_map_t &static_jacobian, const equation_type_and_normalized_equation_t &Equation_Type, bool verbose_, bool select_feedback_variable);
//! Reduce the number of block merging the same type equation in the prologue and the epilogue and determine the type of each block
block_type_firstequation_size_mfs_t reduceBlocksAndTypeDetermination(const vector<pair<int, int>> &blocks, const equation_type_and_normalized_equation_t &Equation_Type, const vector<unsigned int> &n_static, const vector<unsigned int> &n_forward, const vector<unsigned int> &n_backward, const vector<unsigned int> &n_mixed, bool linear_decomposition);
void reduceBlocksAndTypeDetermination(const vector<pair<int, int>> &blocks, const equation_type_and_normalized_equation_t &Equation_Type, const vector<unsigned int> &n_static, const vector<unsigned int> &n_forward, const vector<unsigned int> &n_backward, const vector<unsigned int> &n_mixed, bool linear_decomposition);
//! Determine the maximum number of lead and lag for the endogenous variable in a bloc
/*! Returns a pair { equation_lead,lag, variable_lead_lag } */
pair<lag_lead_vector_t, lag_lead_vector_t> getVariableLeadLagByBlock(const vector<int> &components_set, int nb_blck_sim) const;
......@@ -310,7 +310,7 @@ protected:
//! Print an abstract of the block structure of the model
void printBlockDecomposition(const vector<pair<int, int>> &blocks) const;
//! Determine for each block if it is linear or not
vector<bool> BlockLinear() const;
void determineLinearBlocks();
//! Remove equations specified by exclude_eqs
vector<int> includeExcludeEquations(set<pair<string, string>> &eqs, bool exclude_eqs,
vector<BinaryOpNode *> &equations, vector<int> &equations_lineno,
......
......@@ -1152,19 +1152,19 @@ StaticModel::computingPass(int derivsOrder, int paramsDerivsOrder, const eval_co
auto first_order_endo_derivatives = collect_first_order_derivatives_endogenous();
equation_type_and_normalized_equation = equationTypeDetermination(first_order_endo_derivatives, mfs);
equationTypeDetermination(first_order_endo_derivatives, mfs);
cout << "Finding the optimal block decomposition of the model ..." << endl;
auto [blocks, equation_lag_lead, variable_lag_lead, n_static, n_forward, n_backward, n_mixed] = computeBlockDecompositionAndFeedbackVariablesForEachBlock(static_jacobian, equation_type_and_normalized_equation, false, false);
block_type_firstequation_size_mfs = reduceBlocksAndTypeDetermination(blocks, equation_type_and_normalized_equation, n_static, n_forward, n_backward, n_mixed, false);
reduceBlocksAndTypeDetermination(blocks, equation_type_and_normalized_equation, n_static, n_forward, n_backward, n_mixed, false);
printBlockDecomposition(blocks);
computeChainRuleJacobian();
blocks_linear = BlockLinear();
determineLinearBlocks();
collect_block_first_order_derivatives();
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment