Pull request #16026 updatedBUILD_MSYS2_UCRT64falseBUILD_ALPINEfalseDISABLE_ALL_CMAKE_BUILDSfalseENABLE_MSYS2_UCRT64_CMAKE_BUILDfalseENABLE_MACOS_CMAKE_BUILDfalseENABLE_RUST_PARTESTfalseRUST_PARTEST_SIMCODETARGETwasm-jit

4829e91122a5a5c5af0924c632a64a193afab2131829e91122a5a5c5af0924c632a64a193afab2131PR-16026829e91122a5a5c5af0924c632a64a193afab2131829e91122a5a5c5af0924c632a64a193afab2131PR-16026

829e91122a5a5c5af0924c632a64a193afab2131829e91122a5a5c5af0924c632a64a193afab2131PR-16026https://github.com/OpenModelica/OpenModelica.gitfalse#49984514000538OpenModelica » OMCompiler: Avoid unnecessary full initialization (#16026) #44false4206425FAILURE1783764460434https://test.openmodelica.org/jenkins/job/OpenModelica/view/change-requests/job/PR-16026/4/testsuite/simulation/modelica/NBackend/ModelicaTest/ModelicaTest.Media.TestOnly.FlueGas.mosOMCompiler/Compiler/Template/CodegenC.tplOMCompiler/Compiler/NBackEnd/Classes/NBEquation.moOMCompiler/Compiler/NBackEnd/Modules/2_Pre/NBInline.moOMCompiler/Compiler/NBackEnd/Classes/NBackendDAE.motestsuite/simulation/modelica/NBackend/ModelicaTest/ReferenceFiles/ModelicaTest.Media.TestOnly.FlueGas_res.mat324c3bf3be874f4cc8ea778387b63a6ebb5cd9d51783718086000https://test.openmodelica.org/jenkins/user/38032125+kabdelhak38032125+kabdelhak38032125+kabdelhak@users.noreply.github.com [NB] Fix FlueGas regressions from #15045 (resizable sparsity pattern) (#16036) * [NB,C] Fix SLICE subscripts in resizable Jacobian sparsity templates Add case {SLICE(exp=sliceExp)} to resizableSparsityRowCount and resizableFillDepsForRow in CodegenC.tpl, mirroring the existing SLICE handling already present in resizableColCount. Without this, NBackend Jacobian sparsity for array variables like $pDER_ODE_JAC.$DER.medium2.X[{1,2,3,4}] fell to the else branch and called indexSubRecursive with a SLICE subscript, which only handles INDEX(). Also adds the cross-backend test for ModelicaTest.Media.TestOnly.FlueGas with a committed reference file from the old backend. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * [NB] Fix FlueGas regressions from #15045 (resizable sparsity pattern) Three fixes to make ModelicaTest.Media.TestOnly.FlueGas compile with --newBackend: 1. NBackendDAE.mo: propagate `complete` flag through lowerInstNode so that iterator variables like `j` are lowered via lowerComponentReference (fixes: Variable not found in hashTable) 2. NBEquation.mo / NBInline.mo: add optional funcCrefOpt to EqData.mapExp so that inline() can lower component references on both the expression and cref sides simultaneously (fixes: iterator `i` not a VAR_NODE) 3. CodegenC.tpl: add case {SLICE(exp=sliceExp)} to resizableSparsityRowCount and resizableFillDepsForRow, mirroring the existing SLICE handling in resizableColCount. Without this, array Jacobian variables like $pDER_ODE_JAC.$DER.medium2.X[{1,2,3,4}] caused indexSubRecursive to receive a SLICE subscript which only handles INDEX() Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com> 2026-07-10 23:14:46 +0200324c3bf3be874f4cc8ea778387b63a6ebb5cd9d5[NB] Fix FlueGas regressions from #15045 (resizable sparsity pattern) (#16036)editOMCompiler/Compiler/NBackEnd/Classes/NBEquation.moeditOMCompiler/Compiler/NBackEnd/Modules/2_Pre/NBInline.moeditOMCompiler/Compiler/Template/CodegenC.tpladdtestsuite/simulation/modelica/NBackend/ModelicaTest/ModelicaTest.Media.TestOnly.FlueGas.mosaddtestsuite/simulation/modelica/NBackend/ModelicaTest/ReferenceFiles/ModelicaTest.Media.TestOnly.FlueGas_res.mateditOMCompiler/Compiler/NBackEnd/Classes/NBackendDAE.moOMCompiler/Compiler/OpenModelica.rs/openmodelica_util/src/System.rsOMCompiler/Compiler/OpenModelica.rs/openmodelica_error/src/ErrorExt.rsOMCompiler/Compiler/OpenModelica.rs/openmodelica_util/Cargo.tomlOMCompiler/Compiler/Util/System.moOMCompiler/Compiler/FrontEnd/Parser.mo09e7ddf104e4ac232b78114ace34a0223dec83031783751522000https://test.openmodelica.org/jenkins/user/sjoelundMartin Sjölundmartin.sjolund@liu.seParallelize parsing in the Rust omc port (#16038) Route Parser.parallelParseFilesWork through a new builtin System.launchParallelTasksThreaded, whose Rust body runs the parse worklist on a rayon thread pool. The generic launchParallelTasks stays serial: three of its four call sites carry non-Send payloads (Rc-bearing SymbolTable, Arc<dyn Fn> thunks), so the new function's AnyInput/ AnyOutput: Send bounds opt in only the parse site. The func arrives as the Arc<dyn Fn + 'static> codegen emits; an internal SendSync newtype re-adds the marker the cast drops (sound: routed funcs are always zero-capture top-level fnptr!s). The pool is a process-wide OnceLock sized to the requested thread count, so it is neither oversubscribed nor rebuilt per call. The external "C" binding reuses the existing System_launchParallelTasks symbol, so the classic runtime is a plain alias and no new C code or mmtorust codegen change is needed. ErrorExt.moveMessagesToParentThread (previously a no-op) now drains a worker's thread-local message queue into a shared staging buffer that the parent splices back after the fan-out completes. loadModel(Modelica) parse phase: ~1.5s -> ~0.47s (~3x). wasm32 has no OS threads and falls back to the serial map. Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> 2026-07-11 08:32:02 +020009e7ddf104e4ac232b78114ace34a0223dec8303Parallelize parsing in the Rust omc port (#16038)editOMCompiler/Compiler/Util/System.moeditOMCompiler/Compiler/OpenModelica.rs/openmodelica_error/src/ErrorExt.rseditOMCompiler/Compiler/FrontEnd/Parser.moeditOMCompiler/Compiler/OpenModelica.rs/openmodelica_util/Cargo.tomleditOMCompiler/Compiler/OpenModelica.rs/openmodelica_util/src/System.rsOMCompiler/SimulationRuntime/fmi/export/openmodelica/fmu2_model_interface.cOMCompiler/SimulationRuntime/fmi/export/openmodelica/fmu3_model_interface.c829e91122a5a5c5af0924c632a64a193afab21311783622917000https://test.openmodelica.org/jenkins/user/schuellerschuellerschueller@phimeca.comOMCompiler: Avoid unnecessary full initialization In `updateIfNeeded()`, when called during initialization mode (e.g. from `fmi2GetReal` / `omcGetReal`), the code ran the full `initialization()` solver. This re-solved the initial equations for every tiny parameter perturbation the importer made during setup. For models where the initial equations depend on `time` (e.g. weather data, sin(time), tables), re-running the symbolic initialization solver with `startTime > 0` could produce different results or fail to converge, causing "fmi2GetReal: terminated by an assertion" and ultimately a CVodeError. Replace the `initialization()` call in `updateIfNeeded`'s init-mode path with the same light-weight equation-evaluation path used in the non-init case, plus `updateBoundParameters`. This updates external objects and re-evaluates algebraic/output equations without re-running the full initialization solver. Closes #12823 2026-07-09 20:48:37 +0200829e91122a5a5c5af0924c632a64a193afab2131OMCompiler: Avoid unnecessary full initializationeditOMCompiler/SimulationRuntime/fmi/export/openmodelica/fmu3_model_interface.ceditOMCompiler/SimulationRuntime/fmi/export/openmodelica/fmu2_model_interface.cgithttps://test.openmodelica.org/jenkins/user/38032125+kabdelhak38032125+kabdelhakhttps://test.openmodelica.org/jenkins/user/schuellerschuellerhttps://test.openmodelica.org/jenkins/user/sjoelundMartin Sjölundfalse3https://test.openmodelica.org/jenkins/job/OpenModelica/view/change-requests/job/PR-16026/3/