/*------------------------------------------------------------------------------------------------*/ /* Copyright (C) by the DBCSR developers group - All rights reserved */ /* This file is part of the DBCSR library. */ /* */ /* For information on the license, see the LICENSE file. */ /* For further information please visit https://dbcsr.cp2k.org */ /* SPDX-License-Identifier: GPL-2.0+ */ /*------------------------------------------------------------------------------------------------*/ #include <algorithm> #include <cstdint> #include <cstdio> #include <cstdlib> #include <dbcsr.h> #include <iostream> #include <mpi.h> #include <random> #include <string> #include <tensors/dbcsr_tensor.h> #include <vector> //-------------------------------------------------------------------------------------------------! // Example: tensor contraction (13|2)x(54|21)=(3|45) // tensor1 x tensor2 = tensor3 //-------------------------------------------------------------------------------------------------! std::vector<int> random_dist(int dist_size, int nbins) { std::vector<int> dist(dist_size); for (int i = 0; i < dist_size; i++) dist[i] = i % nbins; return dist; } void printvec(std::vector<int>& v) { for (auto i : v) { std::cout << i << " "; } std::cout << '\n' << std::endl; } void fill_random(dbcsr_t_tensor tensor, std::vector<std::vector<int>> nzblocks) { int myrank, mpi_size; int dim = nzblocks.size(); MPI_Comm_rank(MPI_COMM_WORLD, &myrank); MPI_Comm_size(MPI_COMM_WORLD, &mpi_size); std::random_device rd; std::mt19937_64 gen(rd()); std::uniform_real_distribution<> dis(-1.0, 1.0); if (myrank == 0) std::cout << "Filling Tensor..." << std::endl; if (myrank == 0) std::cout << "Dimension: " << dim << std::endl; int nblocks = nzblocks[0].size(); std::vector<std::vector<int>> mynzblocks(dim); std::vector<int> idx(dim); for (int i = 0; i != nblocks; ++i) { // make index out of nzblocks for (int j = 0; j != dim; ++j) idx[j] = nzblocks[j][i]; int proc = -1; c_dbcsr_t_get_stored_coordinates(tensor, idx.data(), &proc); if (proc == myrank) { for (int j = 0; j != dim; ++j) mynzblocks[j].push_back(idx[j]); } } std::vector<int*> dataptr(4, nullptr); for (int i = 0; i != dim; ++i) { dataptr[i] = mynzblocks[i].size() == 0 ? nullptr : &mynzblocks[i][0]; } if (myrank == 0) std::cout << "Reserving blocks..." << std::endl; if (mynzblocks[0].size() != 0) c_dbcsr_t_reserve_blocks_index(tensor, mynzblocks[0].size(), dataptr[0], dataptr[1], dataptr[2], dataptr[3]); auto fill_rand = [&](std::vector<double>& blk) { for (auto& e : blk) { e = dis(gen); } }; dbcsr_t_iterator iter = nullptr; c_dbcsr_t_iterator_start(&iter, tensor); std::vector<int> loc_idx(dim); std::vector<int> blk_sizes(dim); std::vector<double> block(1); int blk = 0; int blk_proc = 0; while (c_dbcsr_t_iterator_blocks_left(iter)) { c_dbcsr_t_iterator_next_block(iter, loc_idx.data(), &blk, &blk_proc, blk_sizes.data(), nullptr); int tot = 1; for (int i = 0; i != dim; ++i) { tot *= blk_sizes[i]; } block.resize(tot); fill_rand(block); c_dbcsr_t_put_block(tensor, loc_idx.data(), blk_sizes.data(), block.data(), nullptr, nullptr); } c_dbcsr_t_iterator_stop(&iter); MPI_Barrier(MPI_COMM_WORLD); } int main(int argc, char* argv[]) { MPI_Init(&argc, &argv); int mpi_size, mpi_rank; MPI_Comm_size(MPI_COMM_WORLD, &mpi_size); MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank); c_dbcsr_init_lib(MPI_COMM_WORLD, nullptr); dbcsr_t_pgrid pgrid_3d = nullptr; dbcsr_t_pgrid pgrid_4d = nullptr; std::vector<int> dims4(4); std::vector<int> dims3(3); MPI_Fint fcomm = MPI_Comm_c2f(MPI_COMM_WORLD); c_dbcsr_t_pgrid_create(&fcomm, dims3.data(), dims3.size(), &pgrid_3d, nullptr); c_dbcsr_t_pgrid_create(&fcomm, dims4.data(), dims4.size(), &pgrid_4d, nullptr); if (mpi_rank == 0) { std::cout << "pgrid3-dimensions:" << std::endl; printvec(dims3); std::cout << "pgrid4-dimensions:" << std::endl; printvec(dims4); } // block sizes std::vector<int> blk1, blk2, blk3, blk4, blk5; // blk indices of non-zero blocks std::vector<int> nz11, nz12, nz13, nz21, nz22, nz24, nz25, nz33, nz34, nz35; blk1 = {3, 9, 12, 1}; blk2 = {4, 2, 3, 1, 9, 2, 32, 10, 5, 8, 7}; blk3 = {7, 3, 8, 7, 9, 5, 10, 23, 2}; blk4 = {8, 1, 4, 13, 6}; blk5 = {4, 2, 22}; nz11 = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3}; nz12 = {2, 4, 4, 4, 5, 5, 6, 7, 9, 10, 10, 0, 0, 3, 6, 6, 8, 9, 1, 1, 4, 5, 7, 7, 8, 10, 10, 1, 3, 4, 4, 7}; nz13 = {6, 2, 4, 8, 5, 7, 1, 7, 2, 1, 2, 0, 3, 5, 1, 6, 4, 7, 2, 6, 0, 3, 2, 6, 7, 4, 7, 8, 5, 0, 1, 6}; nz21 = {0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3}; nz22 = {0, 2, 3, 5, 9, 1, 1, 3, 4, 4, 5, 5, 5, 6, 6, 8, 8, 8, 9, 10, 0, 2, 2, 3, 4, 5, 7, 8, 10, 10, 0, 2, 3, 5, 9, 10}; nz24 = {2, 4, 1, 2, 1, 2, 4, 0, 0, 3, 1, 2, 3, 0, 3, 2, 3, 3, 1, 0, 2, 0, 0, 2, 3, 2, 3, 1, 1, 2, 0, 0, 2, 1, 4, 4}; nz25 = {0, 2, 1, 0, 0, 1, 2, 0, 2, 0, 1, 2, 1, 0, 2, 1, 2, 1, 0, 1, 2, 0, 1, 2, 1, 1, 1, 2, 0, 1, 0, 2, 1, 0, 2, 1}; nz33 = {1, 3, 4, 4, 4, 5, 5, 7}; nz34 = {2, 1, 0, 0, 2, 1, 3, 4}; nz35 = {2, 1, 0, 1, 2, 1, 0, 0}; // (13|2)x(54|21)=(3|45) // distribute blocks std::vector<int> dist11 = random_dist(blk1.size(), dims3[0]); std::vector<int> dist12 = random_dist(blk2.size(), dims3[1]); std::vector<int> dist13 = random_dist(blk3.size(), dims3[2]); std::vector<int> dist21 = random_dist(blk1.size(), dims4[0]); std::vector<int> dist22 = random_dist(blk2.size(), dims4[1]); std::vector<int> dist23 = random_dist(blk4.size(), dims4[2]); std::vector<int> dist24 = random_dist(blk5.size(), dims4[3]); std::vector<int> dist31 = random_dist(blk3.size(), dims3[0]); std::vector<int> dist32 = random_dist(blk4.size(), dims3[1]); std::vector<int> dist33 = random_dist(blk5.size(), dims3[2]); if (mpi_rank == 0) { std::cout << "dist11:" << std::endl; printvec(dist11); std::cout << "dist12:" << std::endl; printvec(dist12); std::cout << "dist13:" << std::endl; printvec(dist13); std::cout << "dist21:" << std::endl; printvec(dist21); std::cout << "dist22:" << std::endl; printvec(dist22); std::cout << "dist23:" << std::endl; printvec(dist23); std::cout << "dist24:" << std::endl; printvec(dist24); std::cout << "dist31:" << std::endl; printvec(dist31); std::cout << "dist32:" << std::endl; printvec(dist32); std::cout << "dist33:" << std::endl; printvec(dist33); } dbcsr_t_distribution dist1 = nullptr; dbcsr_t_distribution dist2 = nullptr; dbcsr_t_distribution dist3 = nullptr; // (13|2)x(54|21)=(3|45) std::vector<int> map11, map12, map21, map22, map31, map32; map11 = {0, 2}; map12 = {1}; map21 = {3, 2}; map22 = {1, 0}; map31 = {0}; map32 = {1, 2}; if (mpi_rank == 0) std::cout << "Creating dist objects..." << '\n' << std::endl; // create distribution objects c_dbcsr_t_distribution_new( &dist1, pgrid_3d, dist11.data(), dist11.size(), dist12.data(), dist12.size(), dist13.data(), dist13.size(), nullptr, 0); c_dbcsr_t_distribution_new(&dist2, pgrid_4d, dist21.data(), dist21.size(), dist22.data(), dist22.size(), dist23.data(), dist23.size(), dist24.data(), dist24.size()); c_dbcsr_t_distribution_new( &dist3, pgrid_3d, dist31.data(), dist31.size(), dist32.data(), dist32.size(), dist33.data(), dist33.size(), nullptr, 0); MPI_Barrier(MPI_COMM_WORLD); // create tensors // (13|2)x(54|21)=(3|45) dbcsr_t_tensor tensor1 = nullptr; dbcsr_t_tensor tensor2 = nullptr; dbcsr_t_tensor tensor3 = nullptr; if (mpi_rank == 0) std::cout << "Creating tensors..." << std::endl; c_dbcsr_t_create_new(&tensor1, "(13|2)", dist1, map11.data(), map11.size(), map12.data(), map12.size(), nullptr, blk1.data(), blk1.size(), blk2.data(), blk2.size(), blk3.data(), blk3.size(), nullptr, 0); c_dbcsr_t_create_new(&tensor2, "(54|21)", dist2, map21.data(), map21.size(), map22.data(), map22.size(), nullptr, blk1.data(), blk1.size(), blk2.data(), blk2.size(), blk4.data(), blk4.size(), blk5.data(), blk5.size()); c_dbcsr_t_create_new(&tensor3, "(3|45)", dist3, map31.data(), map31.size(), map32.data(), map32.size(), nullptr, blk3.data(), blk3.size(), blk4.data(), blk4.size(), blk5.data(), blk5.size(), nullptr, 0); MPI_Barrier(MPI_COMM_WORLD); // fill the tensors if (mpi_rank == 0) std::cout << "Tensor 1" << '\n' << std::endl; fill_random(tensor1, {nz11, nz12, nz13}); if (mpi_rank == 0) std::cout << "Tensor 2" << '\n' << std::endl; fill_random(tensor2, {nz21, nz22, nz24, nz25}); if (mpi_rank == 0) std::cout << "Tensor 3" << '\n' << std::endl; fill_random(tensor3, {nz33, nz34, nz35}); // contracting // (13|2)x(54|21)=(3|45) MPI_Barrier(MPI_COMM_WORLD); if (mpi_rank == 0) std::cout << "Contracting..." << std::endl; // cn : indices to be contracted // noncn : indices not to be contracted // mapn : how nonc indices map to tensor 3 std::vector<int> c1, nonc1, c2, nonc2, map1, map2; c1 = {0, 1}; nonc1 = {2}; c2 = {0, 1}; nonc2 = {2, 3}; map1 = {0}; map2 = {1, 2}; int unit_nr = -1; if (mpi_rank == 0) unit_nr = 6; bool log_verbose = true; // tensor_3(map_1, map_2) := 0.2 * tensor_1(notcontract_1, contract_1) // * tensor_2(contract_2, notcontract_2) // + 0.8 * tensor_3(map_1, map_2) c_dbcsr_t_contract_r_dp(0.2, tensor1, tensor2, 0.8, tensor3, c1.data(), c1.size(), nonc1.data(), nonc1.size(), c2.data(), c2.size(), nonc2.data(), nonc2.size(), map1.data(), map1.size(), map2.data(), map2.size(), nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, &unit_nr, &log_verbose); c_dbcsr_t_destroy(&tensor1); c_dbcsr_t_destroy(&tensor2); c_dbcsr_t_destroy(&tensor3); c_dbcsr_t_pgrid_destroy(&pgrid_3d, nullptr); c_dbcsr_t_pgrid_destroy(&pgrid_4d, nullptr); c_dbcsr_t_distribution_destroy(&dist1); c_dbcsr_t_distribution_destroy(&dist2); c_dbcsr_t_distribution_destroy(&dist3); c_dbcsr_finalize_lib(); MPI_Finalize(); return 0; }