DCU Compile error

[Cstandardlib]

Describe the bug

When compiling abacus with ROCM, errors are encounterd:

c++: error: unrecognized command line option ‘--offload-arch=gfx906’; did you mean ‘--offload-abi=ilp32’?

And there are warnings about hipsolver
included in abacus-develop/source/module_base/module_container/base/macros/rocm.h:

In file included from /public/software/compiler/dtk/24.04.2/include/hipsolver/internal/hipsolver-types.h:20,
                 from /public/software/compiler/dtk/24.04.2/include/hipsolver/hipsolver.h:20,
                 from /work/home/abacus-develop/source/module_base/module_container/base/macros/rocm.h:7,
/public/software/compiler/dtk/24.04.2/include/hipblas.h:16:161: note: #pragma message: : warning : This file is deprecated. Use the header file from /opt/dtk-24.04.2/include/hipblas/hipblas.h by using #include <hipblas/hipblas.h>
 #pragma message(": warning : This file is deprecated. Use the header file from /opt/dtk-24.04.2/include/hipblas/hipblas.h by using #include <hipblas/hipblas.h>")

Expected behavior

Latest ABACUS should compile on this ROCM environment.

To Reproduce

No response

Environment

Environment

module list
Currently Loaded Modulefiles:
 1) mpi/hpcx/2.12.0/gcc-8.3.1   2) compiler/dtk/24.04.2   3) compiler/cmake/3.23.3

Compile & build

• Compiler: gcc-8.3.1

PACKAGES=/work/home/packages
cmake -B build -DBUILD_TESTING=ON -DCMAKE_INSTALL_PREFIX=~/.local/ \
        -DUSE_OPENMP=ON -DENABLE_LCAO=OFF \
        -DFFTW3_DIR=${PACKAGES}/fftw-3.3.10/build/ \
        -DLAPACK_DIR=${PACKAGES}/OpenBLAS-0.3.28/build/lib \
        -DSCALAPACK_DIR=${PACKAGES}/scalapack-2.2.0/ \
        -DUSE_ROCM=ON
cmake --build build -j`nproc`

Dependencies:

• fftw-3.3.10
• OpenBLAS-0.3.28
• scalapack-2.2.0

Additional Context

No response

Task list for Issue attackers (only for developers)

• Verify the issue is not a duplicate.
• Describe the bug.
• Steps to reproduce.
• Expected behavior.
• Error message.
• Environment details.
• Additional context.
• Assign a priority level (low, medium, high, urgent).
• Assign the issue to a team member.
• Label the issue with relevant tags.
• Identify possible related issues.
• Create a unit test or automated test to reproduce the bug (if applicable).
• Fix the bug.
• Test the fix.
• Update documentation (if necessary).
• Close the issue and inform the reporter (if applicable).

[Cstandardlib]

Error only occurs when TESTING is ON. If TESTING is OFF, it will compile normally.
2 errors:

1. first

abacus-develop/source/module_elecstate/test/elecstate_print_test.cpp:43:1: error: use of undeclared identifier 'InfoNonlocal'
InfoNonlocal::~InfoNonlocal(){}
^
abacus-develop/source/module_elecstate/test/elecstate_base_test.cpp:42:1: error: use of undeclared identifier 'InfoNonlocal'
InfoNonlocal::~InfoNonlocal(){}
^
2 errors generated when compiling for gfx928.
2 errors generated when compiling for gfx926.
2 errors generated when compiling for gfx906.
gmake[2]: *** [source/module_elecstate/test/CMakeFiles/elecstate_print.dir/build.make:76: source/module_elecstate/test/CMakeFiles/elecstate_print.dir/elecstate_print_test.cpp.o] Error 1
gmake[2]: *** Waiting for unfinished jobs....

2. second

/usr/bin/ld: ../../../libcontainer_rocm.a(container_rocm_generated_lapack.hip.cu.o): in function `container::kernels::lapack_trtri<float, container::DEVICE_GPU>::operator()(char const&, char const&, int const&, float*, int const&)':
lapack.hip.cu:(.text._ZN9container7kernels12lapack_trtriIfNS_10DEVICE_GPUEEclERKcS5_RKiPfS7_[_ZN9container7kernels12lapack_trtriIfNS_10DEVICE_GPUEEclERKcS5_RKiPfS7_]+0x82): undefined reference to `container::kernels::lapack_trtri<float, container::DEVICE_CPU>::operator()(char const&, char const&, int const&, float*, int const&)'

[Cstandardlib]

DCU version uses CPU lapack_trtri and the CMakeLists in source/module_base/module_container/ATen/ops/test/CMakeLists.txt does not include this dependency.

template <typename T>
struct lapack_trtri<T, DEVICE_GPU> {
    void operator()(
        const char& uplo,
        const char& diag,
        const int& dim,
        T* Mat,
        const int& lda) 
    {
        // TODO: trtri is not implemented in this method yet
        // Cause the trtri in cuSolver is not stable for ABACUS!
        // hipSolverConnector::trtri(hipsolver_handle, uplo, diag, dim, Mat, lda);
        // hipSolverConnector::potri(hipsolver_handle, uplo, diag, dim, Mat, lda);
        std::vector<T> H_Mat(dim * dim, static_cast<T>(0.0));
        hipMemcpy(H_Mat.data(), Mat, sizeof(T) * H_Mat.size(), hipMemcpyDeviceToHost);
        lapack_trtri<T, DEVICE_CPU>()(uplo, diag, dim, H_Mat.data(), lda);
        hipMemcpy(Mat, H_Mat.data(), sizeof(T) * H_Mat.size(), hipMemcpyHostToDevice);
    }
};

I will add lapack.cpp in source/module_base/module_container/ATen/ops/test/CMakeLists.txt .

[Cstandardlib]

Dependency failure in module_elecstate tests:

/usr/bin/ld: /usr/bin/ld: DWARF error: invalid or unhandled FORM value: 0x25
CMakeFiles/elecstate_fp_energy.dir/__/elecstate_energy_terms.cpp.o: in function `elecstate::ElecState::get_hartree_energy()':
elecstate_energy_terms.cpp:(.text+0x4): undefined reference to `elecstate::H_Hartree_pw::hartree_energy'
/usr/bin/ld: CMakeFiles/elecstate_fp_energy.dir/__/elecstate_energy_terms.cpp.o: in function `elecstate::ElecState::get_etot_efield()':
elecstate_energy_terms.cpp:(.text+0x14): undefined reference to `elecstate::Efield::etotefield'
/usr/bin/ld: CMakeFiles/elecstate_fp_energy.dir/__/elecstate_energy_terms.cpp.o: in function `elecstate::ElecState::get_etot_gatefield()':
elecstate_energy_terms.cpp:(.text+0x24): undefined reference to `elecstate::Gatefield::etotgatefield'
/usr/bin/ld: CMakeFiles/elecstate_fp_energy.dir/__/elecstate_energy_terms.cpp.o: in function `elecstate::ElecState::get_solvent_model_Ael()':
elecstate_energy_terms.cpp:(.text+0x34): undefined reference to `surchem::Ael'
/usr/bin/ld: CMakeFiles/elecstate_fp_energy.dir/__/elecstate_energy_terms.cpp.o: in function `elecstate::ElecState::get_solvent_model_Acav()':
elecstate_energy_terms.cpp:(.text+0x44): undefined reference to `surchem::Acav'
/usr/bin/ld: CMakeFiles/elecstate_fp_energy.dir/__/elecstate_energy_terms.cpp.o: in function `elecstate::ElecState::get_dftu_energy()':
elecstate_energy_terms.cpp:(.text+0x54): undefined reference to `GlobalC::dftu'
clang-15: error: linker command failed with exit code 1 (use -v to see invocation)
gmake[2]: *** [source/module_elecstate/test/CMakeFiles/elecstate_fp_energy.dir/build.make:287: source/module_elecstate/test/elecstate_fp_energy] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:5723: source/module_elecstate/test/CMakeFiles/elecstate_fp_energy.dir/all] Error 2