[spritche]

Hi all,

I'm attempting to build a single precision version of OpenFOAM-v3.0+, and am running into difficulties. I've verified that the double precision builds complete as expected.

For comparison reasons, I was building two versions in parallel, one with the Intel 2015 (composer_xe_2015.0.090) compiler, and the other with gcc 4.8.5. Both compilers have run into the same issue with a function call to Foam::max.

Intel 2015:

icpc -std=c++0x -fp-trap=common -fp-model precise -Dlinux64 -DWM_ARCH_OPTION=64 -DWM_SP -DWM_LABEL_SIZE=32 -Wall -Wextra -Wnon-virtual-dtor -Wno-unused-parameter -Wno-invalid-offsetof -diag-disable 2304 -xHost -O2  -DNoRepository -I../turbulenceModels/lnInclude -I/codes/OpenFOAM/OpenFOAM-v3.0+_IntelSP/src/transportModels -I/codes/OpenFOAM/OpenFOAM-v3.0+_IntelSP/src/finiteVolume/lnInclude -I/codes/OpenFOAM/OpenFOAM-v3.0+_IntelSP/src/meshTools/lnInclude  -IlnInclude -I. -I/codes/OpenFOAM/OpenFOAM-v3.0+_IntelSP/src/OpenFOAM/lnInclude -I/codes/OpenFOAM/OpenFOAM-v3.0+_IntelSP/src/OSspecific/POSIX/lnInclude   -fPIC -c turbulentTransportModels/turbulentTransportModels.C -o /codes/OpenFOAM/OpenFOAM-v3.0+_IntelSP/platforms/linux64IccSPInt32Opt/cfs/build/src/TurbulenceModels/incompressible/turbulentTransportModels/turbulentTransportModels.o
../turbulenceModels/lnInclude/SpalartAllmarasDES.C(189): error: no instance of overloaded function "Foam::max" matches the argument list
            argument types are: (double, Foam::tmp<Foam::GeometricField<Foam::scalar, Foam::fvPatchField, Foam::volMesh>>)
                     /max(SMALL, (fv1*max(1e-10, 1 - ft2)))
                                     ^

gcc 4.8.5:

/codes/OpenFOAM/OpenFOAM-v3.0+_gccSP/src/TurbulenceModels/incompressible
g++ -m64 -Dlinux64 -DWM_ARCH_OPTION=64 -DWM_SP -DWM_LABEL_SIZE=32 -Wall -Wextra -Wold-style-cast -Wnon-virtual-dtor -Wno-unused-parameter -Wno-invalid-offsetof -O3  -DNoRepository -ftemplate-depth-100 -I../turbulenceModels/lnInclude -I/codes/OpenFOAM/OpenFOAM-v3.0+_gccSP/src/transportModels -I/codes/OpenFOAM/OpenFOAM-v3.0+_gccSP/src/finiteVolume/lnInclude -I/codes/OpenFOAM/OpenFOAM-v3.0+_gccSP/src/meshTools/lnInclude  -IlnInclude -I. -I/codes/OpenFOAM/OpenFOAM-v3.0+_gccSP/src/OpenFOAM/lnInclude -I/codes/OpenFOAM/OpenFOAM-v3.0+_gccSP/src/OSspecific/POSIX/lnInclude   -fPIC -c turbulentTransportModels/turbulentTransportModels.C -o /codes/OpenFOAM/OpenFOAM-v3.0+_gccSP/platforms/linux64GccSPInt32Opt/cfs/build/src/TurbulenceModels/incompressible/turbulentTransportModels/turbulentTransportModels.o
In file included from ../turbulenceModels/lnInclude/SpalartAllmarasDES.H:241:0,
                 from turbulentTransportModels/turbulentTransportModels.C:112:
../turbulenceModels/lnInclude/SpalartAllmarasDES.C: In member function 'Foam::tmp<Foam::GeometricField<float, Foam::fvPatchField, Foam::volMesh> > Foam::LESModels::SpalartAllmarasDES<BasicTurbulenceModel>::psi(const volScalarField&, const volScalarField&) const':
../turbulenceModels/lnInclude/SpalartAllmarasDES.C:189:55: error: no matching function for call to 'max(double, Foam::tmp<Foam::GeometricField<float, Foam::fvPatchField, Foam::volMesh> >&)'
                    /max(SMALL, (fv1*max(1e-10, 1 - ft2)))
                                                       ^

If not obvious in the above code blocks, the errors are for the inner max call after 'fv1*'.

Does anyone happen to know a good workaround to this issue, or have experience building the single precision version of OpenFOAM-v3.0+ successfully?

Any assistance is appreciated.

Thanks,
Steve

[spritche]

Just so that others are aware, this is a known bug. The fix can be found on develop.openfoam.com here.

The solution was to force 1e-10 to be a scalar:

/max(SMALL, (fv1*max(scalar(1e-10), 1 - ft2)))

in the file:

src/TurbulenceModels/turbulenceModels/DES/SpalartAllmarasDES/SpalartAllmarasDES.C


Thanks,
Steve