Issue building tester, issues with undefined reference

[pelahi]

Description
I am building slate with HIP/ROCM 5.0.2 and am encountering errors building the tester executables. I configured using cmake and was able build all the libraries but get the following set of undefined references when building tester

/usr/bin/ld: CMakeFiles/tester.dir/matrix_generator.cc.o: in function `slate::BaseMatrix<float>::tileCopyDataLayout(slate::Tile<float>*, slate::Tile<float>*, blas::Layout, bool)':
matrix_generator.cc:(.text._ZN5slate10BaseMatrixIfE18tileCopyDataLayoutEPNS_4TileIfEES4_N4blas6LayoutEb[_ZN5slate10BaseMatrixIfE18tileCopyDataLayoutEPNS_4TileIfEES4_N4blas6LayoutEb]+0x786): undefined reference to `void slate::device::transpose<float>(bool, long, long, float*, long, float*, long, blas::Queue&)'
/usr/bin/ld: matrix_generator.cc:(.text._ZN5slate10BaseMatrixIfE18tileCopyDataLayoutEPNS_4TileIfEES4_N4blas6LayoutEb[_ZN5slate10BaseMatrixIfE18tileCopyDataLayoutEPNS_4TileIfEES4_N4blas6LayoutEb]+0x8a7): undefined reference to `void slate::device::transpose<float>(bool, long, long, float*, long, float*, long, blas::Queue&)'
/usr/bin/ld: CMakeFiles/tester.dir/matrix_generator.cc.o: in function `slate::BaseMatrix<float>::tileLayoutConvert(std::set<std::tuple<long, long>, std::less<std::tuple<long, long> >, std::allocator<std::tuple<long, long> > >&, int, blas::Layout, bool)':
matrix_generator.cc:(.text._ZN5slate10BaseMatrixIfE17tileLayoutConvertERSt3setISt5tupleIJllEESt4lessIS4_ESaIS4_EEiN4blas6LayoutEb[_ZN5slate10BaseMatrixIfE17tileLayoutConvertERSt3setISt5tupleIJllEESt4lessIS4_ESaIS4_EEiN4blas6LayoutEb]+0x831): undefined reference to `void slate::device::transpose_batch<float>(bool, long, float**, long, long, blas::Queue&)'
/usr/bin/ld: matrix_generator.cc:(.text._ZN5slate10BaseMatrixIfE17tileLayoutConvertERSt3setISt5tupleIJllEESt4lessIS4_ESaIS4_EEiN4blas6LayoutEb[_ZN5slate10BaseMatrixIfE17tileLayoutConvertERSt3setISt5tupleIJllEESt4lessIS4_ESaIS4_EEiN4blas6LayoutEb]+0x8c9): undefined reference to `void slate::device::transpose_batch<float>(bool, long, long, float**, long, float**, long, long, blas::Queue&)'
/usr/bin/ld: matrix_generator.cc:(.text._ZN5slate10BaseMatrixIfE17tileLayoutConvertERSt3setISt5tupleIJllEESt4lessIS4_ESaIS4_EEiN4blas6LayoutEb[_ZN5slate10BaseMatrixIfE17tileLayoutConvertERSt3setISt5tupleIJllEESt4lessIS4_ESaIS4_EEiN4blas6LayoutEb]+0x923): undefined reference to `void slate::device::gecopy<float, float>(long, long, float const* const*, long, float**, long, long, blas::Queue&)'
/usr/bin/ld: CMakeFiles/tester.dir/matrix_generator.cc.o: in function `slate::BaseMatrix<double>::tileCopyDataLayout(slate::Tile<double>*, slate::Tile<double>*, blas::Layout, bool)':
matrix_generator.cc:(.text._ZN5slate10BaseMatrixIdE18tileCopyDataLayoutEPNS_4TileIdEES4_N4blas6LayoutEb[_ZN5slate10BaseMatrixIdE18tileCopyDataLayoutEPNS_4TileIdEES4_N4blas6LayoutEb]+0x786): undefined reference to `void slate::device::transpose<double>(bool, long, long, double*, long, double*, long, blas::Queue&)'

This is just a sample of the error messages.

Steps To Reproduce
I am just loading a few modules to provide the dependences, ensure that cmake searchs for them and then just build using make.

Environment

• d9e94b9

• How installed:

  • git clone

• How compiled:

  • CMake: cmake -Dc_api=1 -Dfortran_api=1 -Duse_cmake_find_blas=ON -Duse_cmake_find_lapack=ON -DCMAKE_CXX_COMPILER=CC -DCMAKE_Fortran_COMPILER=ftn -DCMAKE_CXX_FLAGS=-DLAPACK_FORTRAN_ADD_

• Compiler: g++ (GCC) 12.1.0 20220506 (HPE)

• BLAS library (e.g., MKL, ESSL, OpenBLAS) & version: openblas/0.3.15

• LAPACK: netlib-lapack/3.9.1

• SCALAPACK: netlib-scalapack/2.1.0

• CUDA / ROCm / oneMKL version (e.g., nvcc --version): hipcc, HIP version: 5.4.22804-474e8620

• MPI library & version (MPICH, Open MPI, Intel MPI, IBM Spectrum, Cray MPI, etc. Sometimes mpicxx -v gives info.): MPICH Version: 3.4a2 (cray-mpich)

• OS: cray, sles15sp4

• Hardware (CPUs, GPUs, nodes): NA as issue is compilation.