#include #include #include #include #include "util/timer.h" #include using T = float; using DESCRIPTOR = descriptor::D3Q19; void simulate(descriptor::Cuboid cuboid, std::size_t nStep) { auto current = cuda::device::current::get(); Lattice lattice(cuboid); CellMaterials materials(cuboid, [&cuboid](uint3 p) -> int { if (p.x == 0 || p.x == cuboid.nX-1 || p.y == 0 || p.y == cuboid.nY-1 || p.z == 0) { return 2; // boundary cell } else if (p.z == cuboid.nZ-1) { return 3; // lid cell } else { return 1; // bulk } }); auto bulk_mask = materials.mask_of_material(1); auto box_mask = materials.mask_of_material(2); auto lid_mask = materials.mask_of_material(3); cuda::synchronize(current); for (std::size_t iStep=0; iStep < 100; ++iStep) { lattice.apply(Operator(BgkCollideO(), bulk_mask, 0.56), Operator(BounceBackO(), box_mask), Operator(BounceBackMovingWallO(), lid_mask, 0.05f, 0.f, 0.f)); lattice.stream(); } cuda::synchronize(current); auto start = timer::now(); for (std::size_t iStep=0; iStep < nStep; ++iStep) { lattice.apply(Operator(BgkCollideO(), bulk_mask, 0.56), Operator(BounceBackO(), box_mask), Operator(BounceBackMovingWallO(), lid_mask, 0.05f, 0.f, 0.f)); lattice.stream(); } cuda::synchronize(current); auto mlups = timer::mlups(cuboid.volume, nStep, start); std::cout << sizeof(T) << ", " << cuboid.nX << ", " << nStep << ", " << mlups << std::endl; } int main(int argc, char* argv[]) { if (cuda::device::count() == 0) { std::cerr << "No CUDA devices on this system" << std::endl; return -1; } if (argc != 3) { std::cerr << "Invalid parameter count" << std::endl; return -1; } const std::size_t n = atoi(argv[1]); const std::size_t steps = atoi(argv[2]); simulate({ n, n, n}, steps); return 0; }