BackendDAE modules "removedUnusedFunction" hides other bugs.

[Willi Braun]

The module postOptModule "removedUnusedFunction" should be neutral in terms of functionality to our testsuite, but it isn't.
Following results when it's deactivated:

Failed tests:
        ./openmodelica/cppruntime/libraries/msl32/Modelica.Electrical.QuasiStationary.SinglePhase.Examples.ParallelResonance.mos
        ./openmodelica/cppruntime/libraries/msl32/Modelica.Mechanics.MultiBody.Examples.Elementary.DoublePendulum.mos
        ./openmodelica/cppruntime/libraries/msl32/Modelica.Mechanics.MultiBody.Examples.Elementary.RollingWheelSetDriving.mos
        ./openmodelica/debugDumps/optdaedump.mos // doesn't matter
        ./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.Air.DryAirNasa.mos
        ./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir.mos
        ./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.IdealGases.Air.mos
        ./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.IdealGases.Nitrogen.mos
        ./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.IdealGases.SimpleNaturalGas.mos
        ./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.IdealGases.SimpleNaturalGasFixedComposition.mos
        ./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.Water.IdealSteam.mos
        ./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.Water.WaterIF97OnePhase_ph.mos
        ./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.Water.WaterIF97_pT.mos
        ./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.Water.WaterIF97_ph.mos
        ./simulation/libraries/msl32/Modelica.Fluid.Examples.AST_BatchPlant.Test.TankWithEmptyingPipe1.mos
        ./simulation/libraries/msl32/Modelica.Fluid.Examples.AST_BatchPlant.Test.TankWithEmptyingPipe2.mos
        ./simulation/libraries/msl32/Modelica.Fluid.Examples.AST_BatchPlant.Test.TanksWithEmptyingPipe1.mos
        ./simulation/libraries/msl32/Modelica.Fluid.Examples.AST_BatchPlant.Test.TanksWithEmptyingPipe2.mos
        ./simulation/libraries/msl32/Modelica.Fluid.Examples.AST_BatchPlant.Test.TwoTanks.mos
        ./simulation/libraries/msl32/Modelica.Fluid.Examples.BranchingDynamicPipes.mos
        ./simulation/libraries/msl32/Modelica.Fluid.Examples.TraceSubstances.RoomCO2.mos
        ./simulation/libraries/msl32/Modelica.Fluid.Examples.TraceSubstances.RoomCO2WithControls.mos
        ./simulation/modelica/algorithms_functions/ForIterator2.mos
        ./simulation/modelica/hpcom/Modelica.Fluid.Examples.BranchingDynamicPipes.mos // duplicated
24 of 2508 failed

So actually this are only 22 failing test, but as far as I see this are only 4 different issues. So will create separated tickets for them.
For:

./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.Air.DryAirNasa.mos
./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir.mos
./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.IdealGases.Air.mos
./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.IdealGases.Nitrogen.mos
./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.IdealGases.SimpleNaturalGas.mos
./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.IdealGases.SimpleNaturalGasFixedComposition.mos
./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.Water.IdealSteam.mos
./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.Water.WaterIF97OnePhase_ph.mos
./simulation/libraries/msl31/media/Modelica.Media.Examples.Tests.MediaTestModels.Water.WaterIF97_pT.mos
./simulation/libraries/msl31/media

the issue is:
State by sjoelund.se:

They are not missing functions: they are partial functions that should never have been added...

Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c: In function ‘omc_Modelica_Media_Interfaces_PartialMedium_specificEnthalpy__pTX’:
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4883: error: ‘Modelica_Media_Interfaces_PartialMedium_setState__pTX_rettype’ undeclared (first use in this function)
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4883: error: expected ‘;’ before ‘tmp2’
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4884: error: ‘Modelica_Media_Interfaces_PartialMedium_specificEnthalpy_rettype’ undeclared (first use in this function)
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4884: error: expected ‘;’ before ‘tmp3’
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4889: error: ‘tmp2’ undeclared (first use in this function)
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4890: error: ‘tmp3’ undeclared (first use in this function)
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c: In function ‘omc_Modelica_Media_Interfaces_PartialMixtureMedium_specificEnthalpy__pTX’:
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4908: error: ‘Modelica_Media_Interfaces_PartialMixtureMedium_setState__pTX_rettype’ undeclared (first use in this function)
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4908: error: expected ‘;’ before ‘tmp2’
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4909: error: ‘Modelica_Media_Interfaces_PartialMixtureMedium_specificEnthalpy_rettype’ undeclared (first use in this function)
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4909: error: expected ‘;’ before ‘tmp3’
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4914: error: ‘tmp2’ undeclared (first use in this function)
Modelica.Media.Examples.Tests.MediaTestModels.Air.MoistAir_functions.c:4915: error: ‘tmp3’ undeclared (first use in this function)

For:

./simulation/libraries/msl32/Modelica.Fluid.Examples.BranchingDynamicPipes.mos
./simulation/libraries/msl32/Modelica.Fluid.Examples.TraceSubstances.RoomCO2.mos
./simulation/libraries/msl32/Modelica.Fluid.Examples.TraceSubstances.RoomCO2WithControls.mos

it seems that something in instantiation of function with constants is going wrong, since "_TinK", "_poly_Cp" and "_poly_rho" are constants in package.

Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c: In function 'omc_Modelica_Media_Incompressible_TableBased_h__T__der':
Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c:8790: error: '_TinK' undeclared (first use in this function)
Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c:8790: error: (Each undeclared identifier is reported only once
Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c:8790: error: for each function it appears in.)
Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c:8800: error: '_poly_Cp' undeclared (first use in this function)
Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c: In function 'omc_Modelica_Media_Incompressible_TableBased_h__T':
Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c:8829: error: '_TinK' undeclared (first use in this function)
Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c:8849: error: '_poly_Cp' undeclared (first use in this function)
Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c: In function 'omc_Modelica_Media_Incompressible_TableBased_h__pT':
Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c:8892: error: '_TinK' undeclared (first use in this function)
Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c:8912: error: '_poly_Cp' undeclared (first use in this function)
Modelica.Fluid.Examples.BranchingDynamicPipes_functions.c:8926: error: '_poly_rho' undeclared (first use in this function)

For ./simulation/modelica/algorithms_functions/ForIterator2.mos we have some code generation issues with arrays. But I have no idea what kind of issue the cppruntime tests have.

Does anybody something about that issue?
I hope it was not the propose of this module to fix this issue;)

[Martin Sjölund]

They are not missing functions: they are partial functions that should never have been added...

[Adrian Pop]

I think removedUnusedFunction phase should always be run at the start of the back-end.
Maybe a better solution would be to do this before the back-end, even before we dump the DAE.
I don't think is that hard to implement. Any takers? :)

We should do this at least until I fix the damn front-end as they will disappear after that.

[Lennart Ochel]

Should not be that hard to do. I can handle it after the Modelica conference. Or are there other volunteers?

[Adrian Pop]

Basically DAEUtil.traverseDAE + a function gets a DAE.CALL and returns the updated function tree. We'll see if I get some time this week to do it.

[Martin Sjölund]

It's still a useful phase because it removes a lot of code when compiling C-code if there was inlining.

[Lennart Ochel]

Replying to sjoelund.se:

It's still a useful phase because it removes a lot of code when compiling C-code if there was inlining.

Well, shouldn't the inline-module remove unused functions after they get inlined?

[Adrian Pop]

It's not possible to remove the function when you inlined one call. It might be used someplace else where it cannot be inlined. Inlining depends on the function call arguments. So after you inline, some functions might be possible to remove, some might not.

[Martin Sjölund]

Why? Some of the calls might not be inlined. Like LateInline called from both inside function and in model context.

[Martin Sjölund]

And to make things worse: Simplify can also remove function calls, like 0*f(x)

[Lennart Ochel]

That's true.

[Willi Braun]

Replying to adrpo:

I think removedUnusedFunction phase should always be run at the start of the back-end.
Maybe a better solution would be to do this before the back-end, even before we dump the DAE.
I don't think is that hard to implement. Any takers? :)

No, we can't move it to the end of the Front-End. It's correct to do it at the end of the back-end, since we might use functions within IndexReduction which are not used before(derivative annotation), also it's for jacobian, we might use derivative annotations.

[Willi Braun]

Replying to adrpo:

Basically DAEUtil.traverseDAE + a function gets a DAE.CALL and returns the updated function tree. We'll see if I get some time this week to do it.

There is really no need to hurry. I just wanted to make a note on the issues that were covered by that module, the module it self is note that bad.