Skip to content

Commit 680fdde

Browse files
lukelowrylukelowry
committed
Read OPF cases from JSON and rebuild components on ports
1 parent d5c7d68 commit 680fdde

102 files changed

Lines changed: 102595 additions & 12928 deletions

File tree

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

‎CHANGELOG.md‎

Lines changed: 1 addition & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -10,9 +10,7 @@
1010
- Added `enable` and `disable` solver events that put a branch in or out of service.
1111
- `verify()` returns `Model::ConfigurationChecks` listing every configuration problem instead of an error count.
1212
- Added `Model::ParameterReader`
13-
- Added the `OptimalPowerFlow` model family, the `SteadyState` Ipopt interface, and the `EconomicDispatch` example for MATPOWER cases.
14-
- The MATPOWER reader accepts rows without semicolons, solution columns, and decimal `baseMVA` and costs.
15-
- Fixed the study output check for a dangling symlink.
13+
- Added the `OptimalPowerFlow` model family, the `SteadyState::IpoptInterface`, and the `EconomicDispatch` example.
1614

1715
## v0.2
1816

‎GridKit/AutomaticDifferentiation/CMakeLists.txt‎

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,3 @@
11
install(FILES DependencyTracking/Variable.hpp DependencyTracking/VariableImplementation.hpp
22
DependencyTracking/VariableOperators.hpp
33
DESTINATION include/GridKit/AutomaticDifferentiation/DependencyTracking)
4-
install(FILES Enzyme/CooEntries.hpp DESTINATION include/GridKit/AutomaticDifferentiation/Enzyme)

‎GridKit/AutomaticDifferentiation/Enzyme/CooEntries.hpp‎

Lines changed: 0 additions & 38 deletions
This file was deleted.

‎GridKit/AutomaticDifferentiation/Enzyme/D2LDx2.hpp‎

Lines changed: 63 additions & 40 deletions
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,6 @@
77

88
#include <array>
99
#include <cstddef>
10-
#include <type_traits>
1110

1211
#include <GridKit/AutomaticDifferentiation/Enzyme/EnzymeDefinitions.hpp>
1312
#include <GridKit/AutomaticDifferentiation/Enzyme/LowerSparseStorage.hpp>
@@ -22,56 +21,61 @@ namespace GridKit
2221
/**
2322
* @brief Enzyme automatic differentiation Hessian evaluator: Lagrangian Hessian, d2L/dx2
2423
*
25-
* Lower triangle of the Hessian of
26-
* \f$L = \sigma f(x) + \lambda^T g(x)\f$ by one forward sweep per local
27-
* variable over the reverse-mode gradient of \f$L\f$. Only structural
28-
* entries are stored, in global indices.
24+
* Hessian of \f$L = \sigma f(x) + \lambda_c^T c(x) + \lambda_d^T d(x)\f$
25+
* by one forward sweep per local variable over the reverse-mode gradient
26+
* of \f$L\f$. Only structural entries are stored, in global indices.
2927
*
30-
* @tparam ModelT - model type with `VARIABLE_SIZE` local variables and
31-
* `CONSTRAINT_SIZE` local constraints
28+
* @tparam ModelT - model type with internal and external local variables
29+
* and equality constraints, and internal inequality constraints
3230
*/
3331
template <typename ModelT>
3432
struct D2LDx2
3533
{
36-
using ScalarT = typename ModelT::ScalarT;
37-
using IdxT = typename ModelT::IdxT;
38-
using RealT = typename ModelT::RealT;
39-
40-
static_assert(std::is_same_v<ScalarT, double> && std::is_same_v<IdxT, size_t>,
41-
"D2LDx2 supports double values and size_t indices");
42-
4334
/**
4435
* @param[in] model - Pointer to the model to be differentiated
4536
* @param[in] n_var - Number of local variables
4637
* @param[in] variable_indices - Global index of each local variable
4738
* @param[in] x - Local variables
4839
* @param[in] sigma - Objective factor
49-
* @param[in] lambda - Local constraint multipliers
50-
* @param[in,out] entries - Hessian entries
40+
* @param[in] lambda_c - Local equality constraint multipliers
41+
* @param[in] lambda_d - Local inequality constraint multipliers
42+
* @param[out] rows - Row indices
43+
* @param[out] cols - Column indices
44+
* @param[out] vals - Values
45+
* @param[out] nnz - Number of nonzeros
5146
*/
47+
template <typename ScalarT, typename IdxT, typename RealT>
5248
static void eval(const ModelT* model,
5349
const size_t n_var,
5450
const IdxT* variable_indices,
5551
const ScalarT* x,
5652
const RealT sigma,
57-
const RealT* lambda,
58-
CooEntries* entries)
53+
const RealT* lambda_c,
54+
const RealT* lambda_d,
55+
IdxT* rows,
56+
IdxT* cols,
57+
ScalarT* vals,
58+
IdxT& nnz)
5959
{
60-
std::array<ScalarT, ModelT::VARIABLE_SIZE> gradient{};
60+
std::array<ScalarT, ObjectiveT::NUM_INPUTS> gradient{};
6161
for (size_t var_i = 0; var_i < n_var; ++var_i)
6262
{
6363
// Sparse storage. @see LowerSparseStorage.hpp
6464
ScalarT* seed = __enzyme_todense<ScalarT*>((void*) ident_load<ScalarT, IdxT>,
6565
(void*) ident_store<ScalarT, IdxT>,
6666
var_i);
67-
ScalarT* dgradient = __enzyme_todense<ScalarT*>((void*) mapped_load,
68-
(void*) lower_store,
67+
ScalarT* dgradient = __enzyme_todense<ScalarT*>((void*) sparse_load<ScalarT, IdxT>,
68+
(void*) sparse_store<ScalarT, IdxT>,
6969
var_i,
70+
1.0, // value scaling
7071
variable_indices,
7172
variable_indices,
72-
entries);
73+
rows,
74+
cols,
75+
vals,
76+
&nnz);
7377

74-
__enzyme_fwddiff<void>((void*) lagrangianGradient,
78+
__enzyme_fwddiff<void>((void*) lagrangianGradient<ScalarT, RealT>,
7579
enzyme_const,
7680
model,
7781
enzyme_dup,
@@ -83,23 +87,38 @@ namespace GridKit
8387
enzyme_const,
8488
sigma,
8589
enzyme_const,
86-
lambda);
90+
lambda_c,
91+
enzyme_const,
92+
lambda_d);
8793
}
8894
}
8995

9096
private:
91-
__attribute__((always_inline)) inline static ScalarT lagrangian(const ModelT* model,
92-
const ScalarT* x,
93-
const RealT sigma,
94-
const RealT* lambda)
97+
using ObjectiveT = ModelWrapper<ModelT, MemberFunctions::Objective>;
98+
using EqualitiesT = ModelWrapper<ModelT, MemberFunctions::Equalities>;
99+
using InequalitiesT = ModelWrapper<ModelT, MemberFunctions::Inequalities>;
100+
101+
template <typename ScalarT, typename RealT>
102+
[[gnu::always_inline]] inline static ScalarT lagrangian(const ModelT* model,
103+
const ScalarT* x,
104+
const RealT sigma,
105+
const RealT* lambda_c,
106+
const RealT* lambda_d)
95107
{
96-
std::array<ScalarT, ModelT::CONSTRAINT_SIZE> g{};
97-
ModelWrapper<ModelT, MemberFunctions::Constraints>::eval(model, x, g.data());
108+
std::array<ScalarT, EqualitiesT::NUM_OUTPUTS> c{};
109+
EqualitiesT::eval(model, x, c.data());
98110

99-
ScalarT value = sigma * ModelWrapper<ModelT, MemberFunctions::Objective>::eval(model, x);
100-
for (size_t i = 0; i < ModelT::CONSTRAINT_SIZE; ++i)
111+
std::array<ScalarT, InequalitiesT::NUM_OUTPUTS> d{};
112+
InequalitiesT::eval(model, x, d.data());
113+
114+
ScalarT value = sigma * ObjectiveT::eval(model, x);
115+
for (size_t i = 0; i < EqualitiesT::NUM_OUTPUTS; ++i)
116+
{
117+
value += lambda_c[i] * c[i];
118+
}
119+
for (size_t i = 0; i < InequalitiesT::NUM_OUTPUTS; ++i)
101120
{
102-
value += lambda[i] * g[i];
121+
value += lambda_d[i] * d[i];
103122
}
104123
return value;
105124
}
@@ -109,13 +128,15 @@ namespace GridKit
109128
*
110129
* Only the tangent of `gradient` is used, so it is not reset.
111130
*/
112-
__attribute__((always_inline)) inline static void lagrangianGradient(const ModelT* model,
113-
const ScalarT* x,
114-
ScalarT* gradient,
115-
const RealT sigma,
116-
const RealT* lambda)
131+
template <typename ScalarT, typename RealT>
132+
[[gnu::always_inline]] inline static void lagrangianGradient(const ModelT* model,
133+
const ScalarT* x,
134+
ScalarT* gradient,
135+
const RealT sigma,
136+
const RealT* lambda_c,
137+
const RealT* lambda_d)
117138
{
118-
__enzyme_autodiff<void>((void*) lagrangian,
139+
__enzyme_autodiff<void>((void*) lagrangian<ScalarT, RealT>,
119140
enzyme_const,
120141
model,
121142
enzyme_dup,
@@ -124,7 +145,9 @@ namespace GridKit
124145
enzyme_const,
125146
sigma,
126147
enzyme_const,
127-
lambda);
148+
lambda_c,
149+
enzyme_const,
150+
lambda_d);
128151
}
129152
};
130153
} // namespace Sparse

‎GridKit/AutomaticDifferentiation/Enzyme/DfDx.hpp‎

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -22,24 +22,24 @@ namespace GridKit
2222
*
2323
* One reverse sweep.
2424
*
25-
* @tparam ModelT - model type with `VARIABLE_SIZE` local variables
25+
* @tparam ModelT - model type with internal and external local variables
2626
*/
2727
template <typename ModelT>
2828
struct DfDx
2929
{
30-
using ScalarT = typename ModelT::ScalarT;
31-
using RealT = typename ModelT::RealT;
32-
3330
/**
3431
* @param[in] model - Pointer to the model to be differentiated
3532
* @param[in] x - Local variables
3633
* @param[out] gradient - Gradient with respect to the local variables
3734
*/
35+
template <typename ScalarT, typename RealT>
3836
static void eval(const ModelT* model, const ScalarT* x, RealT* gradient)
3937
{
40-
std::fill_n(gradient, ModelT::VARIABLE_SIZE, 0.0);
38+
using WrapperT = ModelWrapper<ModelT, MemberFunctions::Objective>;
39+
40+
std::fill_n(gradient, WrapperT::NUM_INPUTS, 0.0);
4141

42-
__enzyme_autodiff<void>((void*) ModelWrapper<ModelT, MemberFunctions::Objective>::eval,
42+
__enzyme_autodiff<void>((void*) WrapperT::template eval<ScalarT>,
4343
enzyme_const,
4444
model,
4545
enzyme_dup,

‎GridKit/AutomaticDifferentiation/Enzyme/DgDx.hpp‎

Lines changed: 29 additions & 22 deletions
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,6 @@
77

88
#include <array>
99
#include <cstddef>
10-
#include <type_traits>
1110

1211
#include <GridKit/AutomaticDifferentiation/Enzyme/EnzymeDefinitions.hpp>
1312
#include <GridKit/AutomaticDifferentiation/Enzyme/LowerSparseStorage.hpp>
@@ -25,47 +24,55 @@ namespace GridKit
2524
* One forward sweep per local variable. Only structural entries are
2625
* stored, in global indices.
2726
*
28-
* @tparam ModelT - model type with `CONSTRAINT_SIZE` local constraints
27+
* @tparam ModelT - model type with internal and external local variables
28+
* @tparam function - constraint functions, `Equalities` or `Inequalities`
2929
*/
30-
template <typename ModelT>
30+
template <typename ModelT, MemberFunctions function>
3131
struct DgDx
3232
{
33-
using ScalarT = typename ModelT::ScalarT;
34-
using IdxT = typename ModelT::IdxT;
35-
36-
static_assert(std::is_same_v<ScalarT, double> && std::is_same_v<IdxT, size_t>,
37-
"DgDx supports double values and size_t indices");
38-
3933
/**
4034
* @param[in] model - Pointer to the model to be differentiated
4135
* @param[in] n_var - Number of local variables
42-
* @param[in] constraint_indices - Global row of each local constraint
43-
* @param[in] variable_indices - Global column of each local variable
36+
* @param[in] row_indices - Global row of each local constraint function
37+
* @param[in] col_indices - Global column of each local variable
4438
* @param[in] x - Local variables
45-
* @param[in,out] entries - Jacobian entries
39+
* @param[out] rows - Row indices
40+
* @param[out] cols - Column indices
41+
* @param[out] vals - Values
42+
* @param[out] nnz - Number of nonzeros
4643
*/
44+
template <typename ScalarT, typename IdxT>
4745
static void eval(const ModelT* model,
4846
const size_t n_var,
49-
const IdxT* constraint_indices,
50-
const IdxT* variable_indices,
47+
const IdxT* row_indices,
48+
const IdxT* col_indices,
5149
const ScalarT* x,
52-
CooEntries* entries)
50+
IdxT* rows,
51+
IdxT* cols,
52+
ScalarT* vals,
53+
IdxT& nnz)
5354
{
54-
std::array<ScalarT, ModelT::CONSTRAINT_SIZE> g{};
55+
using WrapperT = ModelWrapper<ModelT, function>;
56+
57+
std::array<ScalarT, WrapperT::NUM_OUTPUTS> g{};
5558
for (size_t var_i = 0; var_i < n_var; ++var_i)
5659
{
5760
// Sparse storage. @see LowerSparseStorage.hpp
5861
ScalarT* seed = __enzyme_todense<ScalarT*>((void*) ident_load<ScalarT, IdxT>,
5962
(void*) ident_store<ScalarT, IdxT>,
6063
var_i);
61-
ScalarT* dg = __enzyme_todense<ScalarT*>((void*) mapped_load,
62-
(void*) mapped_store,
64+
ScalarT* dg = __enzyme_todense<ScalarT*>((void*) sparse_load<ScalarT, IdxT>,
65+
(void*) sparse_store<ScalarT, IdxT>,
6366
var_i,
64-
constraint_indices,
65-
variable_indices,
66-
entries);
67+
1.0, // value scaling
68+
row_indices,
69+
col_indices,
70+
rows,
71+
cols,
72+
vals,
73+
&nnz);
6774

68-
__enzyme_fwddiff<void>((void*) ModelWrapper<ModelT, MemberFunctions::Constraints>::eval,
75+
__enzyme_fwddiff<void>((void*) WrapperT::template eval<ScalarT>,
6976
enzyme_const,
7077
model,
7178
enzyme_dup,

0 commit comments

Comments
 (0)