Some 32-bit platforms need linking to libatomic

[barracuda156]

For instance, build fails on ppc:

[ 33%] Linking CXX shared library libngla.dylib
cd /opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_math_ngsolve/ngsolve/work/build/linalg && /opt/local/bin/cmake -E cmake_link_script CMakeFiles/ngla.dir/link.txt --verbose=ON
/opt/local/bin/g++-mp-13 -pipe -Os -DNDEBUG -I/opt/local/include -D_GLIBCXX_USE_CXX11_ABI=0 -arch ppc -mmacosx-version-min=10.6 -dynamiclib -Wl,-headerpad_max_install_names -L/opt/local/lib -Wl,-headerpad_max_install_names -Wl,-rpath,/opt/local/lib/libgcc -o libngla.dylib -install_name /opt/local/lib/libngla.dylib CMakeFiles/ngla.dir/basematrix.cpp.o CMakeFiles/ngla.dir/basevector.cpp.o CMakeFiles/ngla.dir/multivector.cpp.o CMakeFiles/ngla.dir/blockjacobi.cpp.o CMakeFiles/ngla.dir/cg.cpp.o CMakeFiles/ngla.dir/chebyshev.cpp.o CMakeFiles/ngla.dir/commutingAMG.cpp.o CMakeFiles/ngla.dir/eigen.cpp.o CMakeFiles/ngla.dir/jacobi.cpp.o CMakeFiles/ngla.dir/order.cpp.o CMakeFiles/ngla.dir/pardisoinverse.cpp.o CMakeFiles/ngla.dir/sparsecholesky.cpp.o CMakeFiles/ngla.dir/sparsematrix.cpp.o CMakeFiles/ngla.dir/sparsematrix_dyn.cpp.o CMakeFiles/ngla.dir/special_matrix.cpp.o CMakeFiles/ngla.dir/superluinverse.cpp.o CMakeFiles/ngla.dir/mumpsinverse.cpp.o CMakeFiles/ngla.dir/elementbyelement.cpp.o CMakeFiles/ngla.dir/arnoldi.cpp.o CMakeFiles/ngla.dir/paralleldofs.cpp.o CMakeFiles/ngla.dir/python_linalg.cpp.o CMakeFiles/ngla.dir/umfpackinverse.cpp.o CMakeFiles/ngla.dir/__/parallel/parallelvvector.cpp.o CMakeFiles/ngla.dir/__/parallel/parallel_matrices.cpp.o  -Wl,-rpath,"\$ORIGIN" -Wl,-rpath,"\$ORIGIN/" /opt/local/lib/libumfpack.dylib /opt/local/lib/libcholmod.dylib /opt/local/lib/libamd.dylib /opt/local/lib/libccolamd.dylib /opt/local/lib/libsuitesparseconfig.dylib ../basiclinalg/libngbla.dylib ../ngstd/libngstd.dylib -Wl,-framework,Accelerate /opt/local/lib/libnglib.dylib /opt/local/lib/libngcore.dylib /opt/local/Library/Frameworks/Python.framework/Versions/3.11/lib/libpython3.11.dylib 
Undefined symbols:
  "___atomic_compare_exchange_8", referenced from:
      __ZNSt6atomicIdE21compare_exchange_weakERddSt12memory_order in basevector.cpp.o
      __ZNSt6atomicIdE21compare_exchange_weakERddSt12memory_order in sparsecholesky.cpp.o
      __ZNSt6atomicIdE21compare_exchange_weakERddSt12memory_order in sparsematrix.cpp.o
      __ZZNK4ngla22ElementByElementMatrixIdE7MultAddEdRKNS_10BaseVectorERS2_ENKUlRKN6ngcore8TaskInfoEE_clES9_ in elementbyelement.cpp.o
      __ZZNK4ngla22ElementByElementMatrixIdE12MultTransAddEdRKNS_10BaseVectorERS2_ENKUlRKN6ngcore8TaskInfoEE_clES9_ in elementbyelement.cpp.o
  "___atomic_load_8", referenced from:
      __ZN4ngla10BaseVector11AddIndirectEN6ngcore9FlatArrayIimEEN5ngbla10VectorViewIdmSt17integral_constantIiLi1EEEEb in basevector.cpp.o
      __ZZNK4ngla15BaseVectorPtrMV13InnerProductDERKNS_11MultiVectorEENKUliE_clEi in basevector.cpp.o
      __ZN5ngbla9AtomicAddERSt7complexIdES1_ in basevector.cpp.o
      __ZN5ngbla9AtomicAddERSt7complexIdES1_ in basevector.cpp.o
      __ZNK4ngla14SparseCholeskyIdddE10SolveBlockEiN5ngbla10VectorViewIdmSt17integral_constantIiLi1EEEE in sparsecholesky.cpp.o
      __ZN5ngbla9AtomicAddERSt7complexIdES1_ in sparsecholesky.cpp.o
      __ZN5ngbla9AtomicAddERSt7complexIdES1_ in sparsecholesky.cpp.o
      __ZZNK4ngla14SparseCholeskyIdddE14SolveReorderedEN5ngbla10VectorViewIdmSt17integral_constantIiLi1EEEEENKUliE_clEi in sparsecholesky.cpp.o
      __ZZNK4ngla14SparseCholeskyIdddE14SolveReorderedEN5ngbla10VectorViewIdmSt17integral_constantIiLi1EEEEENKUliE_clEi in sparsecholesky.cpp.o
      __ZZNK4ngla14SparseCholeskyIdddE14SolveReorderedEN5ngbla10VectorViewIdmSt17integral_constantIiLi1EEEEENKUliE1_clEi in sparsecholesky.cpp.o
      __ZN5ngbla9AtomicAddILi1EdNS_9MinusExprINS_3VecILi1EKdEEEEEEvRNS2_IXT_ET0_EET1_ in sparsecholesky.cpp.o
      __ZN5ngbla9AtomicAddILi2EdNS_9MinusExprINS_3VecILi2EKdEEEEEEvRNS2_IXT_ET0_EET1_ in sparsecholesky.cpp.o
      __ZN5ngbla9AtomicAddILi3EdNS_9MinusExprINS_3VecILi3EKdEEEEEEvRNS2_IXT_ET0_EET1_ in sparsecholesky.cpp.o
      __ZN5ngbla9AtomicAddERSt7complexIdES1_ in sparsematrix.cpp.o
      __ZN5ngbla9AtomicAddERSt7complexIdES1_ in sparsematrix.cpp.o
      __ZN5ngbla9AtomicAddILi1ELi1EdNS_3MatILi1ELi1EdEEEEvRNS1_IXT_EXT0_ET1_EET2_ in sparsematrix.cpp.o
      __ZN5ngbla9AtomicAddILi2ELi2EdNS_3MatILi2ELi2EdEEEEvRNS1_IXT_EXT0_ET1_EET2_ in sparsematrix.cpp.o
      __ZN5ngbla9AtomicAddILi1ELi2EdNS_3MatILi1ELi2EdEEEEvRNS1_IXT_EXT0_ET1_EET2_ in sparsematrix.cpp.o
      __ZN5ngbla9AtomicAddILi2ELi1EdNS_3MatILi2ELi1EdEEEEvRNS1_IXT_EXT0_ET1_EET2_ in sparsematrix.cpp.o
      __ZN5ngbla9AtomicAddILi3ELi3EdNS_3MatILi3ELi3EdEEEEvRNS1_IXT_EXT0_ET1_EET2_ in sparsematrix.cpp.o
      __ZN5ngbla9AtomicAddILi1ELi3EdNS_3MatILi1ELi3EdEEEEvRNS1_IXT_EXT0_ET1_EET2_ in sparsematrix.cpp.o
      __ZN5ngbla9AtomicAddILi3ELi1EdNS_3MatILi3ELi1EdEEEEvRNS1_IXT_EXT0_ET1_EET2_ in sparsematrix.cpp.o
      __ZN4ngla14SparseMatrixTMIdE16AddElementMatrixEN6ngcore9FlatArrayIimEES4_N5ngbla10MatrixViewIdLNS5_8ORDERINGE1ENS5_14undefined_sizeES8_mEEb in sparsematrix.cpp.o
      __ZN4ngla14SparseMatrixTMIdE25AddElementMatrixSymmetricEN6ngcore9FlatArrayIimEEN5ngbla10MatrixViewIdLNS5_8ORDERINGE1ENS5_14undefined_sizeES8_mEEb in sparsematrix.cpp.o
      __ZZNK4ngla22ElementByElementMatrixIdE7MultAddEdRKNS_10BaseVectorERS2_ENKUlRKN6ngcore8TaskInfoEE_clES9_ in elementbyelement.cpp.o
      __ZZNK4ngla22ElementByElementMatrixIdE12MultTransAddEdRKNS_10BaseVectorERS2_ENKUlRKN6ngcore8TaskInfoEE_clES9_ in elementbyelement.cpp.o
ld: symbol(s) not found

Configure should check for 8-byte atomics support without libatomic, if it is not detected, then pass -latomic ldflag.

Related: NGSolve/netgen#169