# Compiler tag (GCC/CLANG/ICC/ICX/ONEAPI/NVCC) TAG ?= NVCC # Instruction set (SSE/AVX/AVX_FMA/AVX2/AVX512) ISA ?= AVX512 # Optimization scheme (lammps/gromacs/clusters_per_bin) OPT_SCHEME ?= gromacs # Enable likwid (true or false) ENABLE_LIKWID ?= true # SP or DP DATA_TYPE ?= DP # AOS or SOA DATA_LAYOUT ?= AOS # Assembly syntax to generate (ATT/INTEL) ASM_SYNTAX ?= ATT # Debug DEBUG ?= true # Explicitly store and load atom types (true or false) EXPLICIT_TYPES ?= false # Trace memory addresses for cache simulator (true or false) MEM_TRACER ?= false # Trace indexes and distances for gather-md (true or false) INDEX_TRACER ?= false # Compute statistics COMPUTE_STATS ?= true # Configurations for lammps optimization scheme # Use omp simd pragma when running with half neighbor-lists ENABLE_OMP_SIMD ?= true # Use kernel with explicit SIMD intrinsics USE_SIMD_KERNEL ?= false # Configurations for gromacs optimization scheme # Use reference version USE_REFERENCE_VERSION ?= false # Enable XTC output XTC_OUTPUT ?= false # Check if cj is local when decreasing reaction force HALF_NEIGHBOR_LISTS_CHECK_CJ ?= false # Configurations for CUDA # Use CUDA host memory to optimize transfers USE_CUDA_HOST_MEMORY ?= false USE_SUPER_CLUSTERS ?= true #Feature options OPTIONS = -DALIGNMENT=64 #OPTIONS += More options