# Compile NAMD-2.14 [HiPac 3 Release Note and informations](https://hackmd.io/2deAUP8zQ-u1giVioCOdaQ) Notice: With icc 2021.4.0 : must have icc and gcc-9 and g++-9 Verify with `icc --version` With mkl : must have mkl To verify `mkl_help`(Not sure) With cuda-11.* : must have cuda-11.*(below cuda-12.0) To verify `nvcc --version` With fftw3-3.3.10. : must have fftw3 To verify `fftw-wisdom --version` And check the output `fftw-wisdom tool for FFTW version 3.3.10. ` With verbs : must have infiniteband hardware and Mellanox OFED Driver To verify `ibv_devinfo` And should not be `No IB devices found` with ucx : must have ucx (im not sure) Im not sure how to verify. NOTE: If you want cpu only, remove ``` --with-cuda --cuda-prefix /usr/local/cuda \ --cuda-gencode arch=compute_70,code=sm_70 \ --cuda-dlink arch=compute_70,code=sm_70 ``` ## verbs cuda ###### namd-2.14_Linux-x86_64-icc_verbs-linux-x86_64-smp-icc_fftw3_cuda11.8_cuda-gencode_cuda-dlink ```ls /us ./build charm++ verbs-linux-x86_64 icc smp --with-production ``` ``` ./config \ Linux-x86_64-icc \ --charm-arch verbs-linux-x86_64-smp-icc \ --with-fftw3 --fftw-prefix /usr \ --with-mkl --mkl-prefix /opt/intel/oneapi/mkl/ \ --cxx icpc --cxx-opts "-O3 -xHost -ipo" \ --cc icc --cc-opts "-O3 -xHost -ipo" \ --with-cuda --cuda-prefix /usr/local/cuda \ --cuda-gencode arch=compute_70,code=sm_70 \ --cuda-dlink arch=compute_70,code=sm_70 ``` ###### namd-2.14_Linux-x86_64-g++_verbs-linux-x86_64-smp-gcc_fftw3_cuda11.8_cuda-gencode_cuda-dlink verbs + cuda must have smp ``` ./build charm++ verbs-linux-x86_64 gcc smp --with-production ``` ``` ./config \ Linux-x86_64-g++ \ --charm-arch verbs-linux-x86_64-smp-gcc \ --with-fftw3 --fftw-prefix /usr \ --with-mkl --mkl-prefix /opt/intel/oneapi/mkl/ \ --cxx g++ --cxx-opts "-O3" \ --cc gcc --cc-opts "-O3" \ --with-cuda --cuda-prefix /usr/local/cuda \ --cuda-gencode arch=compute_70,code=sm_70 \ --cuda-dlink arch=compute_70,code=sm_70 ``` ## ucx cuda ###### namd-2.14_Linux-x86_64-icc_ucx-linux-x86_64-smp-icc_fftw3_cuda11.8_cuda-gencode_cuda-dlink ucx + cuda must have smp ``` ./build charm++ ucx-linux-x86_64 icc smp --with-production ``` ``` ./config \ Linux-x86_64-icc \ --charm-arch ucx-linux-x86_64-smp-icc \ --with-fftw3 --fftw-prefix /usr \ --with-mkl --mkl-prefix /opt/intel/oneapi/mkl/ \ --cxx icpc --cxx-opts "-O3 -xHost -ipo" \ --cc icc --cc-opts "-O3 -xHost -ipo" \ --with-cuda --cuda-prefix /usr/local/cuda \ --cuda-gencode arch=compute_70,code=sm_70 \ --cuda-dlink arch=compute_70,code=sm_70 ``` ## single_node cuda ###### namd-2.14_Linux-x86_64-icc_multicore-linux-x86_64-icc_fftw3_cuda11.8_cuda-gencode_cuda-dlink ``` ./build charm++ multicore-linux-x86_64 icc --with-production ``` ``` ./config \ Linux-x86_64-icc \ --charm-arch multicore-linux-x86_64-icc \ --with-fftw3 --fftw-prefix /usr \ --with-mkl --mkl-prefix /opt/intel/oneapi/mkl/ \ --cxx icpc --cxx-opts "-O3 -xHost -ipo" \ --cc icc --cc-opts "-O3 -xHost -ipo" \ --with-cuda --cuda-prefix /usr/local/cuda \ --cuda-gencode arch=compute_70,code=sm_70 \ --cuda-dlink arch=compute_70,code=sm_70 ```