Support more quadratic solvers #2574
Conversation
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
pet-mit
added
the
breaking change
src/solver/optimisation/opt_pilotage_optimisation_quadratique.cpp
Outdated
Show resolved
Hide resolved
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
pet-mit
changed the title
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
| @@ -48,7 +48,10 @@ namespace | |||
| { | |||
| void printSolvers() | |||
There was a problem hiding this comment.
In the scope of this PR ? (not sure)
printSolvers() :
- Should be renamed into : logAllAvailableSolvers()
- why do we print to std::cout and not in usual logs ? Does it even make sense ?
Moreover printSolvers is called from handleOptions which has a very non expressive and probably wrong name and, besides printing, does very strange things.
src/tests/end-to-end/binding_constraints/test_binding_constraints.cpp
Outdated
Show resolved
Hide resolved
| INCLUDE | ||
| "${CMAKE_SOURCE_DIR}/solver/utils" |
There was a problem hiding this comment.
INCLUDE
"${CMAKE_SOURCE_DIR}/solver/utils"is this useful ?
I mean : if we link with Antares::solverUtils, related include directory should be automatic
There was a problem hiding this comment.
without it, test file "basis_status.cpp" cannot find "basis_status_impl.h" which is located under src/solver/utils (with sources, not with headers in include folder)
|
|
||
| BOOST_FIXTURE_TEST_CASE(solver_not_supported, QpFixture) | ||
| { | ||
| options.quadraticSolver = "sirius"; |
There was a problem hiding this comment.
When I look at the fixture (the part that does not "check" anything), I wish it could be turned into an class QuadraticProblem.
If this is possible, in unit tests we could have :
BOOST_FIXTURE_TEST_CASE(simple_qp_one_var, QpFixture)
{
QuadraticProblem QP(/*whatever needed*/);
QP.addVariable("x", 0, 1, -0.5, 1);
QP.solve();
BOOST_CHECK(QP.success());
BOOST_TEST(QP.solution() == {0.25});
....
}Instead, we have to manipulate problemeAResoudre things, we don't want to read about as a user.
There was a problem hiding this comment.
I actually want the checks to be explicitely on problemAResoudre, which holds both the inputs and the outputs of this ortools_quadratic_wrapper
Signed-off-by: Peter Mitri <peter.mitri@rte-france.com>
…eature/new_quad_solvers
| @@ -69,7 +70,7 @@ toc_depth: 2 | |||
| #### Build | |||
| * vcpkg (linux, sirius) (#2078) (#2090) (#2145) | |||
| * Remove src/antares-deps (#2182) | |||
| * Use OR-Tools v9.11-rte1.1 (#2437) | |||
| * Use OR-Tools v9.11-rte1.2 (#2574) | |||
There was a problem hiding this comment.
I see that there is or-tools 9.12 now, could we use it ?
There was a problem hiding this comment.
we are not yet able to make an RTE release for sirius support (CI issues with Windows)
|
| @@ -17,7 +17,8 @@ hide: | |||
| | --adequacy | Force the simulation in [adequacy](04-parameters.md#mode) mode | | |||
| | --parallel | Enable [parallel](optional-features/multi-threading.md) computation of MC years | | |||
| | --force-parallel=VALUE | Override the max number of years computed [simultaneously](optional-features/multi-threading.md) | | |||
| | --solver=VALUE | The optimization solver to use. Possible values are: `sirius` (default), `coin`, `xpress`, `scip` | | |||
| | --linear-solver=VALUE | The optimization solver to use for linear problems. Possible values are: `sirius` (default), `coin`, `xpress`, `scip` | | |||
There was a problem hiding this comment.
Breaking change in the API should be expressed as a change in major version number.
| @@ -25,7 +25,7 @@ target_link_libraries(${PROJ} | |||
| Antares::linear-problem-api | |||
| Antares::logs | |||
| Antares::solverUtils | |||
| ortools::ortools | |||
| $<LINK_LIBRARY:WHOLE_ARCHIVE,ortools::ortools> | |||
There was a problem hiding this comment.
Can you add a comment for using this generator expression and not a simple link
There was a problem hiding this comment.
And maybe even write an ADR on the subject
This PR adds the possibility for the user to choose quadratic (QP) solvers other than sirius, dor adequacy problems.
It does so by interfacing Antares QP description (PROBLEME_ANTARES_A_RESOUDRE) to the new OR-Tools MathOpt API.
The new "quadratic-solver" parameter allows the user to switch solver:
Breaking change: