Is there a way to make the compilation of exchcxx more memory efficient? At the moment the build fails on GHA/ADO CI machines when Cuda is enabled.
[7/11] Building CXX object src/CMakeFiles/exchcxx.dir/builtin_interface.cxx.o
[8/11] Building CXX object src/CMakeFiles/exchcxx.dir/cuda/libxc_device.cxx.o
##[warning]Free memory is lower than 5%; Currently used: 95.15%
##[warning]Free memory is lower than 5%; Currently used: 95.15%
[9/11] Building CUDA object src/CMakeFiles/exchcxx.dir/cuda/builtin.cu.o
FAILED: [code=9] src/CMakeFiles/exchcxx.dir/cuda/builtin.cu.o
$BUILD_PREFIX/bin/nvcc -forward-unknown-to-host-compiler -DEXCHCXX_HAS_CONFIG_H=1 -DUSING_Libxc -Dexchcxx_EXPORTS -I$SRC_DIR/include -I$SRC_DIR/_build/include -I$SRC_DIR/src -isystem $PREFIX/include -isystem $BUILD_PREFIX/targets/x86_64-linux/include -O3 -DNDEBUG -std=c++17 -arch=all -Xcompiler=-fPIC -Xcudafe --diag_suppress=partial_override -Xptxas -v -MD -MT src/CMakeFiles/exchcxx.dir/cuda/builtin.cu.o -MF src/CMakeFiles/exchcxx.dir/cuda/builtin.cu.o.d -x cu -c $SRC_DIR/src/cuda/builtin.cu -o src/CMakeFiles/exchcxx.dir/cuda/builtin.cu.o
$SRC_DIR/include/exchcxx/impl/builtin/kernels/deorbitalized.hpp(119): warning #177-D: variable "vtau_k" was declared but never referenced
double TAU_A, TAU_B, vrho_a_k, vrho_b_k, vsigma_aa_k, vsigma_bb_k, vlapl_a_k, vlapl_b_k, vtau_k, dummy;
^
detected during:
instantiation of "void ExchCXX::detail::device_eval_exc_vxc_helper_polar_kernel<KernelType>(int, ExchCXX::const_host_buffer_type, ExchCXX::const_host_buffer_type, ExchCXX::const_host_buffer_type, ExchCXX::const_host_buffer_type, ExchCXX::host_buffer_type, ExchCXX::host_buffer_type, ExchCXX::host_buffer_type, ExchCXX::host_buffer_type, ExchCXX::host_buffer_type) [with KernelType=ExchCXX::BuiltinSCANL_C]" at line 1595 of $SRC_DIR/src/cuda/builtin.cu
instantiation of "void ExchCXX::detail::device_eval_exc_vxc_helper_polar<KernelType>(int, ExchCXX::const_device_buffer_type, ExchCXX::const_device_buffer_type, ExchCXX::const_device_buffer_type, ExchCXX::const_device_buffer_type, ExchCXX::device_buffer_type, ExchCXX::device_buffer_type, ExchCXX::device_buffer_type, ExchCXX::device_buffer_type, ExchCXX::device_buffer_type, cudaStream_t) [with KernelType=ExchCXX::BuiltinSCANL_C]" at line 1039 of $SRC_DIR/include/exchcxx/impl/builtin/kernel_type.hpp
Remark: The warnings can be suppressed with "-diag-suppress <warning-number>"
nvcc error : 'ptxas' died due to signal 9 (Kill signal)
Is there a way to make the compilation of exchcxx more memory efficient? At the moment the build fails on GHA/ADO CI machines when Cuda is enabled.
See Pipelines - Run 20260102.9 logs
Runner info
Conda info