New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix empty array slices in Cpp runtime #7649
Conversation
`stop=0` in `start:step:stop` was used to mark `end` so far. It is needed for empty slices though. See e.g. ModelicaTest.Math.TestMatrices2b (MSL 4.0.0) raising ERROR : init : SimManager: Wrong slice exceeding array size
osx-fmuchecker hangs since an hour. I used this opportunity to extend the test case -- this only works since the merge of #7650 (initialization of reductions). |
* Fix empty array slices in Cpp runtime `stop=0` in `start:step:stop` was used to mark `end` so far. It is needed for empty slices though. See e.g. ModelicaTest.Math.TestMatrices2b (MSL 4.0.0) raising ERROR : init : SimManager: Wrong slice exceeding array size * Add missing access operators for 3-5 dims to ArraySliceConst * Prevent reduction of ArraySlice for index set of size 1
* fix issues with OpenSuse (#7636) - installs liblpsolve55 in lib64, fix the copy - add the path to ffi include when building the Compile runtime to fix: ./../../3rdParty/libffi/install/include/ffi.h:58:10: error: 'ffitarget.h' file not found with <angled> include; use "quotes" instead * Flattening fixes (#7626) - Flatten types in statements. - Flatten expressions after evaluating bindings in EvaluateConstants, since they come from the instance tree and not the flat model. - Create typed bindings instead of flat bindings in some places where the generated binding might not actually be flat. - Lift the result in evalComponentBinding if necessary to ensure it has the same dimensions as the cref being evaluated. - Apply the array hack for dimensions in Dimension.fromExp instead of during typing to simplify problematic dimensions in more cases. - Don't double delete components. * Fix loading of subpackages in case of encryption (#7639) * Fix loading of subpackages in case of encryption * Return the class parts * Expand calls with multiple iterators (#7642) - Expand reductions/array constructors with multiple iterators into nested calls when generating the DAE structure, since the code generation can't handle multiple iterators. * Simplify min/max reductions more (#7644) - Evaluate min/max reductions if possible. - Fix Expression.typeOf for relations, to avoid using the wrong default value when evaluating reductions. * Fix product in old frontend (#7646) - Don't expand product of an expression with unknown dimensions. * Implement product operator and DIV_SCALAR_ARRAY in Cpp runtime (#7641) * Implement product operator in Cpp runtime * Update OMCompiler/SimulationRuntime/cpp/Core/Math/ArrayOperations.cpp Co-authored-by: phannebohm <phannebohm@fh-bielefeld.de> * Update test * Implement DIV_SCALAR_ARRAY in Cpp runtime See e.g. PowerSystems.Examples.AC3ph.Precalculation.EqCircFromTransDat Co-authored-by: phannebohm <phannebohm@fh-bielefeld.de> * Implement const element access in DynArrayDim3 (#7647) See e.g. PowerSystems.Examples.AC3ph.Precalculation.EqCircFromTransDat * Fix Cpp writeLhsCref to not use ArraySliceConst on lhs (#7648) * Fix Cpp writeLhsCref to not use ArraySliceConst on lhs Call 'daeExpCref(isLhs=true, ...)' instead of 'daeExpCrefRhs(...)' in scalarLhsCref. See e.g. PowerSystems.Examples.AC3ph.Precalculation.TransDatFromEqCirc raising: ERROR : init : SimManager: Can't assign array to ArraySliceConst * Whitespaces * Fix initialization of reductions in Cpp runtime (#7650) * Fix initialization of reductions in Cpp and C runtime `foldExpPre` had been declared but not used. This resulted in Cpp errors. See e.g. ModelicaTest.Fluid.TestComponents.Machines.TestWaterPumpCharacteristics or PowerSystems.Examples.AC3ph.Precalculation.Z_matrixTrDat3 ERROR : init : SimManager: Right and left array must have the same size for element wise addition * Revert fix of initialization of reductions for C runtime * Don't dump equations into Cpp FMU sources (#7643) * Fix empty array slices in Cpp runtime (#7649) * Fix empty array slices in Cpp runtime `stop=0` in `start:step:stop` was used to mark `end` so far. It is needed for empty slices though. See e.g. ModelicaTest.Math.TestMatrices2b (MSL 4.0.0) raising ERROR : init : SimManager: Wrong slice exceeding array size * Add missing access operators for 3-5 dims to ArraySliceConst * Prevent reduction of ArraySlice for index set of size 1 * Fix unlifting of array types in ffi runtime (#7638) Co-authored-by: perost <perost86@gmail.com> Co-authored-by: Adeel Asghar <adeel.asghar@liu.se> Co-authored-by: rfranke <rfranke@users.noreply.github.com> Co-authored-by: phannebohm <phannebohm@fh-bielefeld.de>
@rfranke, with this commit the results of PowerSystems_cpp became better than those of PowerSystems with the C runtime, 99 vs 94 successful simulations out of 118 models. I guess this is also due to the fact that PowerSystems_cpp is now testing @sjoelund, @adrpo, would you mind updating the test script for the C runtime to also use Thanks! |
OK, so we have the following branches tested with the master version of OMC
If this is not a mess, I don't know what is 😄 May I at least suggest that we delete the obsolete directories, for which new tests are no longer generated, or at least that we move them in a @rfranke, what do you think? For the renaming, we need to completely overhaul the whole thing, see #5276, I'll try to take care of that this autumn. |
It's fine for me as it is -- as we decided to package the latest version with OpenModelica. The printed library version 1.0.1 makes clear that PowerSystems_cpp should not be related to PowerSystems above, but to PowerSystems_latest below. |
Libraries listed there are not the ones tested. Those are found in the overview reports, and the directory has a lot of old stuff in them. |
That is not immediately apparent if you happen to land on that page.
OK, I guess I should start from there instead
That's my point, I would suggest to clean it up from obsolet stuff. |
stop=0
instart:step:stop
was used to markend
so far.It is needed for empty slices though.
See e.g. ModelicaTest.Math.TestMatrices2b (MSL 4.0.0) raising