Open drmichaeltcvx opened 10 months ago
This shouldn't work, but try eliminating the branch of this ifdef conditional you're not using in your compilation and see if that fixes it:
It is the only location in any of the StackVariables
types involved in this specific issue where the size of the struct is dependent on the compilation context, so theoretically the same type instantiated in two different scopes with different visibility of the CALC_FEM_SHAPE_IN_KERNEL
preproc macro would result in different struct layouts/sizes.
It is a subtype instead of the base type, which is what is actually reporting the error, so I don't expect it to change anything, but it is worthwhile to eliminate as a possibility.
Describe the bug Nvidia linker reports that a symbol (class) appears to have different sizes in coreComponents (libphysicsSolvers.a).
It seems that the definitions for _ZN4geos13finiteElement18ImplicitKernelBaseINS_20CellElementSubRegionENS_12constitutive11PorousSolidINS3_16ElasticIsotropicEEENS0_25H1_Wedge_Lagrange1_Gauss6ELi3ELi3EE14StackVariablesC1Ev$571
are not matching in two locations :
../lib/libphysicsSolvers.a:PoromechanicsEFEMKernels_CellElementSubRegion_PorousSolid-ElasticIsotropic-_H1_Wedge_Lagrange1_Gauss6.cpp.o , and
../lib/libphysicsSolvers.a:SolidMechanicsFixedStressThermoPoroElasticKernels_CellElementSubRegion_PorousSolid-ElasticIsotropic-_H1_Wedge_Lagrange1_Gauss6.cpp.o
To Reproduce Steps to reproduce the behavior:
Expected behavior GPU-enabled GEOS used to build fine before ~ Nov 10
Screenshots
Platform (please complete the following information):
Additional context Add any other context about the problem here.