diff --git a/source/source_basis/module_ao/CMakeLists.txt b/source/source_basis/module_ao/CMakeLists.txt index b453b3fb45..09132e9811 100644 --- a/source/source_basis/module_ao/CMakeLists.txt +++ b/source/source_basis/module_ao/CMakeLists.txt @@ -15,8 +15,8 @@ if(ENABLE_LCAO) endif() if(BUILD_TESTING) - if(ENABLE_MPI) - add_subdirectory(test) - endif() + # These tests support both MPI and serial builds: MPI-only cases are + # guarded by __MPI, while the remaining cases use the serial layout. + add_subdirectory(test) endif() endif() diff --git a/source/source_basis/module_ao/parallel_orbitals.h b/source/source_basis/module_ao/parallel_orbitals.h index b786339bcf..ed10b012e5 100644 --- a/source/source_basis/module_ao/parallel_orbitals.h +++ b/source/source_basis/module_ao/parallel_orbitals.h @@ -27,13 +27,6 @@ class Parallel_Orbitals Parallel_2D& para_2d() { return para_2d_; } const Parallel_2D& para_2d() const { return para_2d_; } - - // Transitional implicit upcast. Keeping it lets the const call sites that - // pass a Parallel_Orbitals where a const Parallel_2D& is expected compile - // while they are migrated. It is deleted once every such site calls - // para_2d() explicitly; no new site may rely on it. - operator const Parallel_2D&() const { return para_2d_; } - #ifdef __MPI /// BLACS context of the 2D process grid, completing the getter family /// (get_row_size / get_col_size / get_local_size / get_block_size / diff --git a/source/source_basis/module_ao/test/CMakeLists.txt b/source/source_basis/module_ao/test/CMakeLists.txt index 9e53fbc4cf..501c0434a8 100644 --- a/source/source_basis/module_ao/test/CMakeLists.txt +++ b/source/source_basis/module_ao/test/CMakeLists.txt @@ -72,10 +72,12 @@ addtest( abacus_copy_test_files(parallel_orbitals_test.sh) find_program(BASH bash) -add_test( - NAME MODULE_AO_parallel_orbitals_test_para - COMMAND ${BASH} parallel_orbitals_test.sh - "$" - WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}) +if(ENABLE_MPI) + add_test( + NAME MODULE_AO_parallel_orbitals_test_para + COMMAND ${BASH} parallel_orbitals_test.sh + "$" + WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}) +endif() abacus_copy_test_files(lcao_H2O) diff --git a/source/source_basis/module_ao/test/parallel_orbitals_test.cpp b/source/source_basis/module_ao/test/parallel_orbitals_test.cpp index 1f399d51a4..7152acd40b 100644 --- a/source/source_basis/module_ao/test/parallel_orbitals_test.cpp +++ b/source/source_basis/module_ao/test/parallel_orbitals_test.cpp @@ -1,6 +1,12 @@ #include "gtest/gtest.h" #include "../parallel_orbitals.h" +#include +#include +#include +#include +#include + //-------------------------------------------------------------- // unit test of class "Parallel_Orbitals" //-------------------------------------------------------------- @@ -15,9 +21,40 @@ * - get_indexes_col * - get_indexes_row(iat) * - get_indexes_col(iat) - * + * + * Composition guard (folded in from the former + * test_parallel_orbitals_composition.cpp): + * - the composed Parallel_2D layout is exposed only through para_2d(); + * - derived atom and band metadata can be resynchronized after layout changes; + * - copy and move semantics remain explicit; + * - MPI and serial builds exercise their respective initialization paths. + * * the test framework is based on parallel_2d_test.cpp */ +static_assert(!std::is_convertible::value, + "Parallel_Orbitals must expose Parallel_2D only through para_2d()"); +static_assert(!std::is_copy_constructible::value, + "Parallel_Orbitals must not be copy constructible"); +static_assert(!std::is_copy_assignable::value, + "Parallel_Orbitals must not be copy assignable"); +static_assert(!std::is_move_constructible::value, + "Parallel_Orbitals must not be move constructible"); +static_assert(!std::is_move_assignable::value, + "Parallel_Orbitals must not be move assignable"); +static_assert(std::is_default_constructible::value, + "Parallel_Orbitals must be default constructible"); + +static_assert(!std::is_copy_constructible::value, + "Parallel_2D must not be copy constructible"); +static_assert(!std::is_copy_assignable::value, + "Parallel_2D must not be copy assignable"); +static_assert(std::is_move_constructible::value, + "Parallel_2D must be move constructible"); +static_assert(std::is_move_assignable::value, + "Parallel_2D must be move assignable"); +static_assert(std::is_default_constructible::value, + "Parallel_2D must be default constructible"); + class TestParaO : public testing::Test { protected: @@ -161,7 +198,7 @@ TEST_F(TestParaO, Serial) //1. set dim0 and dim1 //2. set_serial po.para_2d().set_serial(gr, gc); - EXPECT_EQ(po.para_2d().dim0 * po.para_2d().dim1, 1); + EXPECT_EQ(po.para_2d().get_dim0() * po.para_2d().get_dim1(), 1); EXPECT_EQ(po.para_2d().get_row_size(), gr); EXPECT_EQ(po.para_2d().get_col_size(), gc); EXPECT_EQ(po.para_2d().get_local_size(), gr * gc); @@ -196,6 +233,180 @@ TEST_F(TestParaO, Serial) } #endif +//-------------------------------------------------------------- +// composition guard for class "Parallel_Orbitals" +//-------------------------------------------------------------- +class ParallelOrbitalsCompositionTest : public testing::Test +{ +protected: + std::ofstream ofs_running; + std::ofstream ofs_warning; + std::vector iat2iwt{0, 10, 20}; + const int nat = 3; + + /// Refresh state derived from the current para_2d() layout. + void run_derived_layer(Parallel_Orbitals& po, const int nlocal_global, const int nbands_global) + { + po.set_atomic_trace(iat2iwt.data(), nat, nlocal_global); +#ifdef __MPI + ASSERT_EQ(po.set_nloc_wfc_Eij(nbands_global, ofs_running, ofs_warning), 0); + po.set_desc_wfc_Eij(nlocal_global, nbands_global, po.para_2d().get_row_size()); +#endif + } + + /// Assert that derived state matches the current para_2d() layout. + void assert_sync_with(const Parallel_Orbitals& po, const int nlocal_global, const int nbands_global) + { + // The trailing sentinels must follow the new distribution. + EXPECT_EQ(po.atom_begin_row[nat], po.para_2d().nrow); + EXPECT_EQ(po.atom_begin_col[nat], po.para_2d().ncol); + // Each atom must either start locally or be marked absent on this rank. + for (int iat = 0; iat < nat; ++iat) + { + if (po.atom_begin_row[iat] == -1) + { + EXPECT_EQ(po.get_nrow_atom(iat), 0); + } + else + { + EXPECT_GE(po.atom_begin_row[iat], 0); + EXPECT_LT(po.atom_begin_row[iat], po.para_2d().nrow); + } + if (po.atom_begin_col[iat] == -1) + { + EXPECT_EQ(po.get_ncol_atom(iat), 0); + } + else + { + EXPECT_GE(po.atom_begin_col[iat], 0); + EXPECT_LT(po.atom_begin_col[iat], po.para_2d().ncol); + } + } +#ifdef __MPI + // Band metadata must be re-derived from the current distribution. + EXPECT_EQ(po.nrow_bands, po.para_2d().nrow); + EXPECT_EQ(po.nloc_wfc, static_cast(po.ncol_bands) * po.para_2d().nrow); + EXPECT_EQ(po.nloc_Eij, static_cast(po.ncol_bands) * po.ncol_bands); + EXPECT_EQ(po.get_wfc_global_nbands(), po.desc_wfc[3]); + EXPECT_EQ(po.get_wfc_global_nbasis(), po.desc_wfc[2]); + // The embedded descriptor context must match the composed layout. + EXPECT_EQ(po.desc_wfc[1], po.get_blacs_ctxt()); + EXPECT_EQ(po.desc_wfc[2], nlocal_global); + EXPECT_EQ(po.desc_wfc[3], nbands_global); + EXPECT_EQ(po.desc_wfc[8], po.para_2d().get_row_size()); +#endif + } +}; + +TEST_F(ParallelOrbitalsCompositionTest, GetterEqualsPara2d) +{ + Parallel_Orbitals po; + const int mg = 50; + const int ng = 50; +#ifdef __MPI + ASSERT_EQ(po.para_2d().init(mg, ng, 2, MPI_COMM_WORLD, 1), 0); +#else + po.para_2d().set_serial(mg, ng); +#endif + const Parallel_2D& p2d = po.para_2d(); + EXPECT_EQ(p2d.get_row_size(), p2d.nrow); + EXPECT_EQ(p2d.get_col_size(), p2d.ncol); + EXPECT_EQ(p2d.get_local_size(), p2d.nloc); + EXPECT_EQ(p2d.get_local_size(), static_cast(p2d.nrow) * p2d.ncol); + EXPECT_EQ(p2d.get_block_size(), p2d.nb); + EXPECT_EQ(p2d.get_dim0(), p2d.dim0); + EXPECT_EQ(p2d.get_dim1(), p2d.dim1); +#ifdef __MPI + EXPECT_EQ(p2d.get_desc(), &p2d.desc[0]); + EXPECT_EQ(p2d.desc[1], p2d.blacs_ctxt); + EXPECT_EQ(p2d.desc[2], mg); + EXPECT_EQ(p2d.desc[3], ng); + EXPECT_EQ(p2d.desc[4], p2d.nb); + EXPECT_EQ(p2d.desc[5], p2d.nb); + EXPECT_EQ(p2d.desc[8], std::max(p2d.nrow, 1)); + // PO 对组合成员唯一的转发 getter:必须与 para_2d() 内部一致 + EXPECT_EQ(po.get_blacs_ctxt(), p2d.blacs_ctxt); +#endif +} + +#ifdef __MPI +TEST_F(ParallelOrbitalsCompositionTest, DerivedLayerSyncsAfterReinit) +{ + Parallel_Orbitals po; + // First use a small distribution. + ASSERT_EQ(po.para_2d().init(30, 30, 1, MPI_COMM_WORLD, 1), 0); + const int old_nrow = po.para_2d().nrow; + run_derived_layer(po, 30, 16); + EXPECT_EQ(po.nrow_bands, old_nrow); + EXPECT_EQ(po.atom_begin_row[nat], old_nrow); + + // Reinitialize only the composed layout and leave derived state stale. + ASSERT_EQ(po.para_2d().init(50, 50, 1, MPI_COMM_WORLD, 1), 0); + const int new_nrow = po.para_2d().nrow; + ASSERT_NE(new_nrow, old_nrow); + EXPECT_NE(po.nrow_bands, new_nrow); + EXPECT_NE(po.atom_begin_row[nat], new_nrow); + + // Refreshing derived state restores the invariant. + run_derived_layer(po, 50, 16); + assert_sync_with(po, 50, 16); +} + +TEST_F(ParallelOrbitalsCompositionTest, DerivedLayerSyncsAfterSet) +{ + Parallel_Orbitals po; + ASSERT_EQ(po.para_2d().init(40, 40, 2, MPI_COMM_WORLD, 1), 0); + run_derived_layer(po, 40, 16); + EXPECT_EQ(po.nrow_bands, po.para_2d().nrow); + + // Reinitialize through set() while reusing the existing BLACS context. + const int blacs = po.para_2d().blacs_ctxt; + ASSERT_EQ(po.para_2d().set(60, 60, 2, blacs), 0); + ASSERT_NE(po.para_2d().nrow, po.nrow_bands); + run_derived_layer(po, 60, 16); + assert_sync_with(po, 60, 16); +} +#else +TEST_F(ParallelOrbitalsCompositionTest, DerivedLayerSyncsAfterSetSerial) +{ + Parallel_Orbitals po; + po.para_2d().set_serial(30, 30); + const int old_nrow = po.para_2d().nrow; + po.set_atomic_trace(iat2iwt.data(), nat, 30); + EXPECT_EQ(po.atom_begin_row[nat], old_nrow); + + // Reinitialize only the composed layout and leave derived state stale. + po.para_2d().set_serial(50, 50); + const int new_nrow = po.para_2d().nrow; + ASSERT_NE(new_nrow, old_nrow); + EXPECT_NE(po.atom_begin_row[nat], new_nrow); + + // Refresh derived state; serial global-to-local maps are identities. + po.set_atomic_trace(iat2iwt.data(), nat, 50); + EXPECT_EQ(po.atom_begin_row[nat], po.para_2d().nrow); + EXPECT_EQ(po.atom_begin_col[nat], po.para_2d().ncol); + for (int iat = 0; iat < nat; ++iat) + { + EXPECT_EQ(po.atom_begin_row[iat], iat2iwt[iat]); + EXPECT_EQ(po.atom_begin_col[iat], iat2iwt[iat]); + } +} +#endif + +TEST_F(ParallelOrbitalsCompositionTest, ValueSemanticsSmoke) +{ + Parallel_2D source; + source.set_serial(4, 4); + EXPECT_EQ(source.nrow, 4); + + Parallel_2D moved(std::move(source)); + EXPECT_EQ(moved.nrow, 4); + + Parallel_2D assigned; + assigned = std::move(moved); + EXPECT_EQ(assigned.nrow, 4); +} + int main(int argc, char** argv) { #ifdef __MPI diff --git a/source/source_hamilt/module_hcontainer/test/test_hcontainer.cpp b/source/source_hamilt/module_hcontainer/test/test_hcontainer.cpp index db77bc121e..187697ebb8 100644 --- a/source/source_hamilt/module_hcontainer/test/test_hcontainer.cpp +++ b/source/source_hamilt/module_hcontainer/test/test_hcontainer.cpp @@ -460,8 +460,7 @@ TEST_F(HContainerTest, atompair_funcs) PO.atom_begin_row[i] = i*2; // nw = 2, value should be i*nw PO.atom_begin_col[i] = i*2; } - PO.para_2d().nrow = 4; - PO.para_2d().ncol = 4; + PO.para_2d().set_serial(4, 4); const int natom = 2; const int rx_shift = 1, ry_shift = 1, rz_shift = 1; hamilt::AtomPair atom_ij(0, 0, &PO, nullptr); diff --git a/source/source_hamilt/module_hcontainer/test/test_hcontainer_complex.cpp b/source/source_hamilt/module_hcontainer/test/test_hcontainer_complex.cpp index cde61f96eb..beee0acff3 100644 --- a/source/source_hamilt/module_hcontainer/test/test_hcontainer_complex.cpp +++ b/source/source_hamilt/module_hcontainer/test/test_hcontainer_complex.cpp @@ -418,8 +418,7 @@ TEST_F(HContainerTest, atompair_funcs) PO.atom_begin_row[i] = i*2; // nw = 2, value should be i*nw PO.atom_begin_col[i] = i*2; } - PO.para_2d().nrow = 4; - PO.para_2d().ncol = 4; + PO.para_2d().set_serial(4, 4); hamilt::AtomPair> atom_ij(0, 0, &PO, nullptr); hamilt::AtomPair> atom_ij2(0, 1, 1, 1, 1, &PO, nullptr); hamilt::AtomPair> atom_ij3(1, 0, PO.atom_begin_row.data(), PO.atom_begin_col.data(), 2, nullptr); diff --git a/source/source_hamilt/module_hcontainer/test/test_hcontainer_output.cpp b/source/source_hamilt/module_hcontainer/test/test_hcontainer_output.cpp index 027c762eac..de4f176ad6 100644 --- a/source/source_hamilt/module_hcontainer/test/test_hcontainer_output.cpp +++ b/source/source_hamilt/module_hcontainer/test/test_hcontainer_output.cpp @@ -72,8 +72,6 @@ TEST_F(OutputHContainerTest, Write) ParaV.atom_begin_row[i] = i * 2; ParaV.atom_begin_col[i] = i * 2; } - ParaV.para_2d().nrow = 4; - ParaV.para_2d().ncol = 4; std::ofstream ofs("output_hcontainer.log"); ParaV.para_2d().set_serial(4, 4); // std::cout << "ParaV.global2local_row = " << ParaV.global2local_row(0) << " " << ParaV.global2local_row(1) << " " diff --git a/source/source_hsolver/unittests/test_diago_pexsi.cpp b/source/source_hsolver/unittests/test_diago_pexsi.cpp index f4dc11b4dc..a8f458a460 100644 --- a/source/source_hsolver/unittests/test_diago_pexsi.cpp +++ b/source/source_hsolver/unittests/test_diago_pexsi.cpp @@ -175,13 +175,8 @@ template class PexsiPrepare exit(1); } - // set po variables - po.para_2d().ncol = hmtest.ncol; - po.para_2d().nrow = hmtest.nrow; - po.para_2d().nb = nb2d; - po.para_2d().blacs_ctxt = icontxt; - po.para_2d().dim0 = nprows; - po.para_2d().dim1 = npcols; + // set po variables from the same BLACS grid used for hmtest + po.para_2d().set(nlocal, nlocal, nb2d, icontxt); if (DETAILINFO && myrank == 0) { diff --git a/source/source_io/module_ctrl/ctrl_scf_lcao.cpp b/source/source_io/module_ctrl/ctrl_scf_lcao.cpp index 85dbd319e8..53287286d9 100644 --- a/source/source_io/module_ctrl/ctrl_scf_lcao.cpp +++ b/source/source_io/module_ctrl/ctrl_scf_lcao.cpp @@ -189,12 +189,8 @@ void ModuleIO::ctrl_scf_lcao(UnitCell& ucell, // Mode-2 MAT/ subdirectory is created lazily at write time. ModuleBase::ensure_directory(global_mat_dir); const int precision = inp.out_dmr[1]; - // File-name selection and append/truncate are independent decisions: - // mode 0 keeps the fixed filename and overwrites it at every write, - // mode 1 keeps the fixed filename and appends later steps, and mode 2 - // embeds the ionic step in the g{step} filename. - ModuleIO::write_dmr(dm->get_dmr_vec(), &ucell, precision, pv, out_app_flag, - ucell.get_iat2iwt(), ucell.nat, global_mat_dir, istep, ofs_running); + ModuleIO::write_dmr(dm->get_dmr_vec(), &ucell, precision, pv.para_2d(), out_app_flag, + ucell.get_iat2iwt(), ucell.nat, global_mat_dir, istep, ofs_running); } //------------------------------------------------------------------ @@ -211,7 +207,8 @@ void ModuleIO::ctrl_scf_lcao(UnitCell& ucell, } const int precision = inp.out_dmk[1]; - ModuleIO::write_dmk(dm->get_dmk_vec(), kv, precision, efermis, &(ucell), pv, global_mat_dir, istep, out_app_flag, ofs_running); + ModuleIO::write_dmk(dm->get_dmk_vec(), kv, precision, efermis, &(ucell), pv.para_2d(), + global_mat_dir, istep, out_app_flag, ofs_running); } //------------------------------------------------------------------ @@ -298,7 +295,7 @@ void ModuleIO::ctrl_scf_lcao(UnitCell& ucell, std::vector*> hr_vec = p_hamilt->getHR_vector(); const hamilt::HContainer* sr = p_hamilt->getSR(); - ModuleIO::write_hsr(hr_vec, sr, &ucell, inp.out_hsr[0], precision, pv, + ModuleIO::write_hsr(hr_vec, sr, &ucell, inp.out_hsr[0], precision, pv.para_2d(), out_app_flag, gamma_only, ucell.get_iat2iwt(), ucell.nat, istep, global_mat_dir, pelec->eferm, ofs_running); @@ -582,7 +579,7 @@ void ModuleIO::ctrl_scf_lcao(UnitCell& ucell, 1, // true for upper triangle matrix (inp.out_app_flag == 1), t_fn, - pv, + pv.para_2d(), GlobalV::DRANK, PARAM.inp.ks_solver); diff --git a/source/source_io/module_dhs/write_dh.cpp b/source/source_io/module_dhs/write_dh.cpp index 30bb8dcf20..09b37863fb 100644 --- a/source/source_io/module_dhs/write_dh.cpp +++ b/source/source_io/module_dhs/write_dh.cpp @@ -143,7 +143,7 @@ void write_dh_perI(WriteDHParams& params, false, append_fk, fk, - pv, + pv.para_2d(), GlobalV::DRANK, PARAM.inp.ks_solver); } diff --git a/source/source_io/module_dm/test/test_write_dmr.cpp b/source/source_io/module_dm/test/test_write_dmr.cpp index bfdd8bbb86..62317d99bd 100644 --- a/source/source_io/module_dm/test/test_write_dmr.cpp +++ b/source/source_io/module_dm/test/test_write_dmr.cpp @@ -57,8 +57,6 @@ void init_serial_orbitals(Parallel_Orbitals& pv) pv.atom_begin_row[1] = 2; pv.atom_begin_col[0] = 0; pv.atom_begin_col[1] = 2; - pv.para_2d().nrow = 2; - pv.para_2d().ncol = 2; pv.para_2d().set_serial(2, 2); } diff --git a/source/source_io/module_hs/hsk_writer.cpp b/source/source_io/module_hs/hsk_writer.cpp index 38e8ec7ec5..b603c6ddca 100644 --- a/source/source_io/module_hs/hsk_writer.cpp +++ b/source/source_io/module_hs/hsk_writer.cpp @@ -62,7 +62,7 @@ void ModuleIO::write_hsk( 1, append_mat, h_fn, - pv, + pv.para_2d(), drank, ks_solver); @@ -89,7 +89,7 @@ void ModuleIO::write_hsk( 1, append_mat, s_fn, - pv, + pv.para_2d(), drank, ks_solver); } // end ik diff --git a/source/source_io/module_hs/hterm_writer.cpp b/source/source_io/module_hs/hterm_writer.cpp index 3576aa65f3..296cf597b3 100644 --- a/source/source_io/module_hs/hterm_writer.cpp +++ b/source/source_io/module_hs/hterm_writer.cpp @@ -161,7 +161,7 @@ static void write_hk_common(hamilt::HContainer& hR, false, append, fname, - pv, + pv.para_2d(), drank, ks_solver); } diff --git a/source/source_io/module_hs/unittests/CMakeLists.txt b/source/source_io/module_hs/unittests/CMakeLists.txt index 56c2dcd879..2bcf5290e0 100644 --- a/source/source_io/module_hs/unittests/CMakeLists.txt +++ b/source/source_io/module_hs/unittests/CMakeLists.txt @@ -14,6 +14,9 @@ addtest( LIBS parameter SOURCES test_lat_r_csr.cpp ../lat_r_csr.cpp + ../../../source_basis/module_ao/parallel_orbitals.cpp + ../../../source_base/parallel_2d.cpp + ../../../source_base/tool_title.cpp ../../../source_base/global_variable.cpp ../../../source_base/parallel_reduce.cpp ../../../source_base/parallel_common.cpp diff --git a/source/source_io/module_hs/unittests/csr_test_helpers.h b/source/source_io/module_hs/unittests/csr_test_helpers.h index f6c956510e..d537226791 100644 --- a/source/source_io/module_hs/unittests/csr_test_helpers.h +++ b/source/source_io/module_hs/unittests/csr_test_helpers.h @@ -160,8 +160,6 @@ void init_serial_orbitals(Parallel_Orbitals& pv) pv.atom_begin_row[1] = 2; pv.atom_begin_col[0] = 0; pv.atom_begin_col[1] = 2; - pv.para_2d().nrow = 2; - pv.para_2d().ncol = 2; pv.para_2d().set_serial(2, 2); } diff --git a/source/source_io/module_hs/unittests/test_hs_dense_io.cpp b/source/source_io/module_hs/unittests/test_hs_dense_io.cpp index 680390a072..b06a7838fa 100644 --- a/source/source_io/module_hs/unittests/test_hs_dense_io.cpp +++ b/source/source_io/module_hs/unittests/test_hs_dense_io.cpp @@ -279,7 +279,7 @@ TEST(HsDenseIoSaveMat, BinaryGammaUpperTriangle) const std::vector local = distribute_matrix(pv, global, dim); ModuleIO::save_mat(0, local.data(), dim, true, 8, true, false, filename, - pv, test_rank, "cg"); + pv.para_2d(), test_rank, "cg"); if (test_rank == 0) { @@ -312,7 +312,7 @@ TEST(HsDenseIoSaveMat, BinaryComplexUpperTriangleMpiReduction) = distribute_matrix(pv, global, dim); ModuleIO::save_mat(0, local.data(), dim, true, 8, true, false, filename, - pv, test_rank, "cg"); + pv.para_2d(), test_rank, "cg"); if (test_rank == 0) { @@ -348,9 +348,9 @@ TEST(HsDenseIoSaveMat, BinaryAppendCompleteRecordsAndOverwrite) = distribute_matrix(pv, replacement, dim); ModuleIO::save_mat(0, first_local.data(), dim, true, 8, true, true, - filename, pv, test_rank, "cg"); + filename, pv.para_2d(), test_rank, "cg"); ModuleIO::save_mat(1, second_local.data(), dim, true, 8, true, true, - filename, pv, test_rank, "cg"); + filename, pv.para_2d(), test_rank, "cg"); if (test_rank == 0) { @@ -365,7 +365,7 @@ TEST(HsDenseIoSaveMat, BinaryAppendCompleteRecordsAndOverwrite) MPI_Barrier(DIAG_WORLD); #endif ModuleIO::save_mat(2, replacement_local.data(), dim, true, 8, true, false, - filename, pv, test_rank, "cg"); + filename, pv.para_2d(), test_rank, "cg"); if (test_rank == 0) { @@ -397,12 +397,12 @@ TEST(HsDenseIoSaveMat, AppendFalseOverwritesSameIstep) // Step 1: append=true writes the full record for istep=0. ModuleIO::save_mat(0, first_local.data(), dim, true, 8, true, true, - filename, pv, test_rank, "cg"); + filename, pv.para_2d(), test_rank, "cg"); // Step 2: append=false on the same istep must truncate the file, not // append a second copy. ModuleIO::save_mat(0, replacement_local.data(), dim, true, 8, true, false, - filename, pv, test_rank, "cg"); + filename, pv.para_2d(), test_rank, "cg"); if (test_rank == 0) { diff --git a/source/source_io/module_hs/unittests/test_hsr_writer.cpp b/source/source_io/module_hs/unittests/test_hsr_writer.cpp index 402457154b..f13ee49b67 100644 --- a/source/source_io/module_hs/unittests/test_hsr_writer.cpp +++ b/source/source_io/module_hs/unittests/test_hsr_writer.cpp @@ -463,7 +463,7 @@ TEST(HsrWriterIo, HContainerBinaryMpiGatherWritesCompleteFiles) eferm.ef = 0.4; // Ry; used only for the H(R) header std::ofstream ofs_running_null; // not opened; test does not inspect the running log ModuleIO::write_hsr( - hr_vec, &sr_parallel, &ucell, 2, 8, parallel_pv, 1, true, iat2iwt, 1, 0, "./", eferm, + hr_vec, &sr_parallel, &ucell, 2, 8, parallel_pv.para_2d(), 1, true, iat2iwt, 1, 0, "./", eferm, ofs_running_null); MPI_Barrier(MPI_COMM_WORLD); @@ -551,7 +551,7 @@ TEST(HsrWriterIo, WriteHsrTextCsrCarriesPerSpinFermiFromEfermi) std::ofstream ofs_running_null; // not opened; test does not inspect the running log ModuleIO::write_hsr( - hr_vec, &sr, &ucell, 1, 8, pv, 1, true, iat2iwt_ptr, 1, 0, "./", eferm, + hr_vec, &sr, &ucell, 1, 8, pv.para_2d(), 1, true, iat2iwt_ptr, 1, 0, "./", eferm, ofs_running_null); const std::string output_up = read_file(hr_up_filename); diff --git a/source/source_io/module_hs/unittests/test_lat_r_csr.cpp b/source/source_io/module_hs/unittests/test_lat_r_csr.cpp index 5258ace65f..d956716179 100644 --- a/source/source_io/module_hs/unittests/test_lat_r_csr.cpp +++ b/source/source_io/module_hs/unittests/test_lat_r_csr.cpp @@ -16,36 +16,6 @@ * - ModuleIO::save_lat_r * - write one lattice-R block in CSR format */ -Parallel_Orbitals::Parallel_Orbitals() -{ -} - -Parallel_Orbitals::~Parallel_Orbitals() -{ -} - -Parallel_2D::~Parallel_2D() -{ -} - -void Parallel_2D::set_serial(const int M_A, const int N_A) -{ - this->nrow = M_A; - this->ncol = N_A; - this->is_serial = true; - this->global2local_row_.resize(M_A); - this->global2local_row_[0] = 0; - this->global2local_row_[1] = 1; - this->global2local_row_[2] = -1; - this->global2local_row_[3] = 2; - this->global2local_row_[4] = -1; //Some rows have global2local_row_ < 0 -} - -int Parallel_2D::get_global_row_size() const -{ - return this->nrow; -} - TEST(ModuleIOTest, SaveLatR) { // Create temporary output file diff --git a/source/source_io/unittests/CMakeLists.txt b/source/source_io/unittests/CMakeLists.txt index 501831fae5..7f7b475480 100644 --- a/source/source_io/unittests/CMakeLists.txt +++ b/source/source_io/unittests/CMakeLists.txt @@ -320,6 +320,7 @@ if(ENABLE_LIBRI) LIBS parameter base device SOURCES test_restart_exx_csr.cpp ../module_restart/restart_exx_csr.cpp ../module_hs/hs_sparse_io.cpp ../module_hs/lat_r_csr.cpp + ../../source_basis/module_ao/parallel_orbitals.cpp tmp_mocks.cpp) endif() diff --git a/source/source_io/unittests/test_write_wfc_nao.cpp b/source/source_io/unittests/test_write_wfc_nao.cpp index 755ca074d6..686596dfd9 100644 --- a/source/source_io/unittests/test_write_wfc_nao.cpp +++ b/source/source_io/unittests/test_write_wfc_nao.cpp @@ -195,10 +195,7 @@ class WriteWfcLcaoTest : public testing::Test } #ifdef __MPI pv.para_2d().init(nbasis, nbands, 1, MPI_COMM_WORLD); - for (int i = 0; i < 9; i++) - { - pv.desc_wfc[i] = pv.para_2d().get_desc()[i]; - } + pv.set_desc_wfc_Eij(nbasis, nbands, pv.para_2d().get_row_size()); // create a global PV, and distribute the psi_init to local Parallel_2D pv_glb; diff --git a/source/source_io/unittests/tmp_mocks.cpp b/source/source_io/unittests/tmp_mocks.cpp index bcb2060a70..23091e6f68 100644 --- a/source/source_io/unittests/tmp_mocks.cpp +++ b/source/source_io/unittests/tmp_mocks.cpp @@ -1,6 +1,5 @@ #include "source_cell/unitcell.h" -#include "source_basis/module_ao/parallel_orbitals.h" // constructor of Atom Atom::Atom() @@ -50,10 +49,3 @@ UnitCell::UnitCell() UnitCell::~UnitCell() { } - -Parallel_Orbitals::Parallel_Orbitals() -{ -} -Parallel_Orbitals::~Parallel_Orbitals() -{ -} diff --git a/source/source_lcao/module_bse/hamilt_bse.cpp b/source/source_lcao/module_bse/hamilt_bse.cpp index 8321544c05..f55888dedb 100644 --- a/source/source_lcao/module_bse/hamilt_bse.cpp +++ b/source/source_lcao/module_bse/hamilt_bse.cpp @@ -525,7 +525,7 @@ void HamiltBSE::cal_V_by_grid(bool is_A) // for (int ik1 = 0;ik1 < nk;++ik1) // LR_Util::print_tensor(v_k_2d[ik1], "V(k)[ik=" + std::to_string(ik1) + "]", &this->pmat); #ifdef __MPI - LR::ao_to_mo_pblas(v_k_2d, this->pmat, psi_is, this->pc, this->naos, + LR::ao_to_mo_pblas(v_k_2d, this->pmat.para_2d(), psi_is, this->pc, this->naos, nocc[is], nvirt[is], pV_col, V_col_local.data(), false, LR_Util::MO_TYPE::VO, T(1.0)); for (int ik1 = 0; ik1 < this->nk; ++ik1) { diff --git a/source/source_lcao/module_deltaspin/test/spin_constrain_test.cpp b/source/source_lcao/module_deltaspin/test/spin_constrain_test.cpp index 7ff102f2ea..64b4cfd5f0 100644 --- a/source/source_lcao/module_deltaspin/test/spin_constrain_test.cpp +++ b/source/source_lcao/module_deltaspin/test/spin_constrain_test.cpp @@ -157,8 +157,8 @@ TYPED_TEST(SpinConstrainTest, SetSolverParameters) TYPED_TEST(SpinConstrainTest, SetParaV) { Parallel_Orbitals paraV; + // A default-constructed distribution has no local elements (nloc == 0). // warning 1 - paraV.para_2d().nloc = 0; testing::internal::CaptureStdout(); EXPECT_EXIT(this->sc.set_ParaV(¶V), ::testing::ExitedWithCode(1), ""); std::string output = testing::internal::GetCapturedStdout(); diff --git a/source/source_lcao/module_dftu/dftu_nao_occ.cpp b/source/source_lcao/module_dftu/dftu_nao_occ.cpp index c176c3dc93..57b0ca40db 100644 --- a/source/source_lcao/module_dftu/dftu_nao_occ.cpp +++ b/source/source_lcao/module_dftu/dftu_nao_occ.cpp @@ -145,21 +145,21 @@ void acc_over_kstar(OccupationMatrix& occmat, std::vector> srho_rot; if (isym < nsym) { // unitary space-group operation (isym==0 is the identity) - srho_rot = dftu_occ_symrot.rot_matrix_ao(srho_ibz, ik_ibz, kstar_size, isym, pv); + srho_rot = dftu_occ_symrot.rot_matrix_ao(srho_ibz, ik_ibz, kstar_size, isym, pv.para_2d()); } else { // antiunitary element: TRS * (spatial operation), see restore_dm const int isym_M = ucell.symm.magnetic_nspin4 ? isym : (isym - nsym); if (nspin == 4) { - if (sigma_y.empty()) { sigma_y = dftu_occ_symrot.set_sigma_y_2d(pv); } + if (sigma_y.empty()) { sigma_y = dftu_occ_symrot.set_sigma_y_2d(pv.para_2d()); } srho_rot = dftu_occ_symrot.trs_spin_rotate( - dftu_occ_symrot.rot_matrix_ao(srho_ibz, ik_ibz, kstar_size, isym_M, pv, false), - sigma_y, pv, 1.0); + dftu_occ_symrot.rot_matrix_ao(srho_ibz, ik_ibz, kstar_size, isym_M, pv.para_2d(), false), + sigma_y, pv.para_2d(), 1.0); } else { - srho_rot = dftu_occ_symrot.rot_matrix_ao(srho_ibz, ik_ibz, kstar_size, isym_M, pv, true); + srho_rot = dftu_occ_symrot.rot_matrix_ao(srho_ibz, ik_ibz, kstar_size, isym_M, pv.para_2d(), true); } } acc_for_ik(occmat, ucell, pv, srho_rot.data(), spin, l_channel); @@ -367,7 +367,7 @@ void cal_occ_mat_k(const Parallel_Orbitals* pv, ModuleSymmetry::Symmetry_rotation_k::get_bvk_cells(period), period, ucell.lat); dftu_occ_symrot.cal_Ms(kv, ucell.symm, ucell.atoms, ucell.latvec, ucell.lmax, ucell.nat, ucell.iat2it, ucell.iat2ia, - *pv, nspin); + pv->para_2d(), nspin); dftu_occ_symrot_built = true; } diff --git a/source/source_lcao/module_dftu/dftu_nao_op.cpp b/source/source_lcao/module_dftu/dftu_nao_op.cpp index 5dedc370d3..378a838a8c 100644 --- a/source/source_lcao/module_dftu/dftu_nao_op.cpp +++ b/source/source_lcao/module_dftu/dftu_nao_op.cpp @@ -94,7 +94,7 @@ void hamilt::DFTU_onsite>::ensure_symmetry_rotation this->ucell->nat, this->ucell->iat2it, this->ucell->iat2ia, - pv, + pv.para_2d(), this->nspin); this->symrot_built_ = true; } @@ -133,7 +133,7 @@ void hamilt::DFTU_onsite>::contribute_from_symmetri module_dm::DensityMatrix dmr_sym(&pv, nspin0, kvec_d_full, nks_full); dmr_sym.init_dmr(*this->dm_->get_dmr_ptr(1)); dmr_sym.get_dmk_vec() - = this->symrot_.restore_dm(*this->kv_, this->dm_->get_dmk_vec(), pv); + = this->symrot_.restore_dm(*this->kv_, this->dm_->get_dmk_vec(), pv.para_2d()); dmr_sym.cal_dmr(-1); const hamilt::HContainer* dmr = dmr_sym.get_dmr_ptr(this->current_spin + 1); diff --git a/source/source_lcao/module_dftu/unittests/CMakeLists.txt b/source/source_lcao/module_dftu/unittests/CMakeLists.txt index 1902c710ba..fb68297283 100644 --- a/source/source_lcao/module_dftu/unittests/CMakeLists.txt +++ b/source/source_lcao/module_dftu/unittests/CMakeLists.txt @@ -58,7 +58,12 @@ if(ENABLE_MPI) addtest( TARGET MODULE_DFTU_fs_k LIBS base device - SOURCES test_dftu_nao_fs_k.cpp) + SOURCES + test_dftu_nao_fs_k.cpp + ../../../source_basis/module_ao/parallel_orbitals.cpp + # The test constructs Parallel_Orbitals normally and links the MPI-built + # base library, so it must keep __MPI to match the Parallel_2D layout. + KEEP_FEATURE_DEFINITIONS __MPI) addtest( TARGET MODULE_DFTU_fs_accum diff --git a/source/source_lcao/module_dftu/unittests/test_dftu_nao_folding.cpp b/source/source_lcao/module_dftu/unittests/test_dftu_nao_folding.cpp index 3df7024be5..d78c970ffe 100644 --- a/source/source_lcao/module_dftu/unittests/test_dftu_nao_folding.cpp +++ b/source/source_lcao/module_dftu/unittests/test_dftu_nao_folding.cpp @@ -42,8 +42,7 @@ namespace DFTU_LCAO TEST(GetLinearIndexTest, RowMajor) { Parallel_Orbitals pv; - pv.para_2d().nrow = 3; - pv.para_2d().ncol = 4; + pv.para_2d().set_serial(3, 4); // row-major: mu * ncol + nu ("cg" is not column-major) EXPECT_EQ(get_linear_index("cg", 0, 0, pv), 0); @@ -56,8 +55,7 @@ TEST(GetLinearIndexTest, RowMajor) TEST(GetLinearIndexTest, ColumnMajor) { Parallel_Orbitals pv; - pv.para_2d().nrow = 3; - pv.para_2d().ncol = 4; + pv.para_2d().set_serial(3, 4); // column-major: mu + nu * nrow EXPECT_EQ(get_linear_index("scalapack_gvx", 0, 0, pv), 0); diff --git a/source/source_lcao/module_dftu/unittests/test_dftu_nao_fs_k.cpp b/source/source_lcao/module_dftu/unittests/test_dftu_nao_fs_k.cpp index 7b690e6d11..ead6786d71 100644 --- a/source/source_lcao/module_dftu/unittests/test_dftu_nao_fs_k.cpp +++ b/source/source_lcao/module_dftu/unittests/test_dftu_nao_fs_k.cpp @@ -32,10 +32,10 @@ UnitCell::~UnitCell() {} * see the same objects the caller passed in. * * Note: we allocate the dependency objects on the heap and never destroy - * them, so the linker never sees references to their (heavy) constructors - * or destructors. This keeps the test target free of BLACS / neighbor / - * two-center linkage while still exercising the reference semantics of - * DftuFsEnv. + * them, so the linker never sees references to their (heavy) destructors. + * The Parallel_Orbitals object is constructed normally: its default + * constructor initializes the distribution member, which a raw + * ::operator new allocation would skip. ***********************************************************************/ namespace DFTU_LCAO @@ -54,7 +54,7 @@ class DftuFsKTest : public ::testing::Test dftu = static_cast(::operator new(sizeof(Plus_U_Base))); fsr = static_cast(::operator new(sizeof(ForceStressArrays))); gd = static_cast(::operator new(sizeof(Grid_Driver))); - pv = static_cast(::operator new(sizeof(Parallel_Orbitals))); + pv = new Parallel_Orbitals(); } Plus_U_Base* dftu = nullptr; diff --git a/source/source_lcao/module_lr/cal_edm.h b/source/source_lcao/module_lr/cal_edm.h index 3c99f9d72c..aaaf5744e8 100644 --- a/source/source_lcao/module_lr/cal_edm.h +++ b/source/source_lcao/module_lr/cal_edm.h @@ -83,7 +83,7 @@ namespace LR { eig_ks_plus_ext[occupied] = eig_ks[idx_eig_ks + occupied] + eig_ext_istate; } - edm[ik] = cal_edm_single_kpoint(X_ao_occ.data(), px_ao_occ, eig_ks_plus_ext.data(), pmat); + edm[ik] = cal_edm_single_kpoint(X_ao_occ.data(), px_ao_occ, eig_ks_plus_ext.data(), pmat.para_2d()); } return edm; } @@ -113,7 +113,7 @@ namespace LR // 1. c * W * c #ifdef __MPI - const std::vector cWc = cal_dm_trans_pblas(W, p_occ_occ, c, pc, naos, nocc, nvirt, pmat, (T)1., LR_Util::MO_TYPE::OO); + const std::vector cWc = cal_dm_trans_pblas(W, p_occ_occ, c, pc, naos, nocc, nvirt, pmat.para_2d(), (T)1., LR_Util::MO_TYPE::OO); #else const std::vector cWc = cal_dm_trans_blas(W, c, nocc, nvirt, (T)1., LR_Util::MO_TYPE::OO); #endif @@ -126,10 +126,10 @@ namespace LR px, epsi_Z.data() + ik * px.get_local_size()); } #ifdef __MPI - std::vector cZc = cal_dm_trans_pblas(epsi_Z.data(), px, c, pc, naos, nocc, nvirt, pmat); + std::vector cZc = cal_dm_trans_pblas(epsi_Z.data(), px, c, pc, naos, nocc, nvirt, pmat.para_2d()); for (ct::Tensor& block : cZc) { - LR_Util::matsym(block.data(), naos, pmat); + LR_Util::matsym(block.data(), naos, pmat.para_2d()); } #else std::vector cZc = cal_dm_trans_blas(epsi_Z.data(), c, nocc, nvirt); @@ -144,10 +144,10 @@ namespace LR // `K_cvcx` already carries the factor 2 of $W^X_{ij}=2K_{ij}[D^X]$ (see `op_K_cvcx` above), // `matsym` then supplies the 1/2 that turns $2\,X_\kappa K c_\lambda$ into the symmetric pair above. #ifdef __MPI - std::vector cKc = cal_dm_trans_pblas(K_cvcx, px, c, pc, naos, nocc, nvirt, pmat, (T)1.0); + std::vector cKc = cal_dm_trans_pblas(K_cvcx, px, c, pc, naos, nocc, nvirt, pmat.para_2d(), (T)1.0); for (ct::Tensor& block : cKc) { - LR_Util::matsym(block.data(), naos, pmat); + LR_Util::matsym(block.data(), naos, pmat.para_2d()); } #else std::vector cKc = cal_dm_trans_blas(K_cvcx, c, nocc, nvirt, (T)1.0); @@ -348,10 +348,10 @@ namespace LR auto set_dm_trans = [&dmx_buf, X, &off_x, &px, &psi_ks_spin, &pc, naos, &nocc, &nvirt, &pmat, nk, &DM_trans](const int is)->void { #ifdef __MPI - dmx_buf = cal_dm_trans_pblas(X + off_x[is], px[is], psi_ks_spin[is], pc, naos, nocc[is], nvirt[is], pmat); + dmx_buf = cal_dm_trans_pblas(X + off_x[is], px[is], psi_ks_spin[is], pc, naos, nocc[is], nvirt[is], pmat.para_2d()); for (auto& t: dmx_buf) { - LR_Util::mattrans(t.data(), naos, pmat); + LR_Util::mattrans(t.data(), naos, pmat.para_2d()); } #else dmx_buf = cal_dm_trans_blas(X + off_x[is], psi_ks_spin[is], nocc[is], nvirt[is]); diff --git a/source/source_lcao/module_lr/cal_w_from_z.h b/source/source_lcao/module_lr/cal_w_from_z.h index 05ea300e26..ac4f160592 100644 --- a/source/source_lcao/module_lr/cal_w_from_z.h +++ b/source/source_lcao/module_lr/cal_w_from_z.h @@ -141,9 +141,9 @@ namespace LR { const auto psi_ks_is = LR_Util::get_psi_spin(psi_ks, is, nk); #ifdef __MPI - dm_trans_2d = cal_dm_trans_pblas(x_ptr, px[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat); + dm_trans_2d = cal_dm_trans_pblas(x_ptr, px[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat.para_2d()); for (auto& t: dm_trans_2d) - LR_Util::matsym(t.data(), naos, pmat); + LR_Util::matsym(t.data(), naos, pmat.para_2d()); #else dm_trans_2d = cal_dm_trans_blas(x_ptr, psi_ks_is, nocc[is], nvirt[is]); for (auto& t: dm_trans_2d) @@ -158,9 +158,9 @@ namespace LR { const auto psi_ks_is = LR_Util::get_psi_spin(psi_ks, is, nk); #ifdef __MPI - std::vector z_2d = cal_dm_trans_pblas(z_ptr, px[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat); + std::vector z_2d = cal_dm_trans_pblas(z_ptr, px[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat.para_2d()); for (auto& t: z_2d) - LR_Util::matsym(t.data(), naos, pmat); + LR_Util::matsym(t.data(), naos, pmat.para_2d()); #else std::vector z_2d = cal_dm_trans_blas(z_ptr, psi_ks_is, nocc[is], nvirt[is]); for (auto& t: z_2d) @@ -168,7 +168,7 @@ namespace LR #endif #ifdef __MPI - dm_diff_2d = cal_dm_diff_pblas(x_ptr, px[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat); + dm_diff_2d = cal_dm_diff_pblas(x_ptr, px[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat.para_2d()); #else dm_diff_2d = cal_dm_diff_blas(x_ptr, psi_ks_is, naos, nocc[is], nvirt[is]); #endif @@ -300,12 +300,12 @@ namespace LR const T* const x_ptr = X + off_x[is]; const T* const z_ptr = Z + off_x[is]; #ifdef __MPI - std::vector z_2d = cal_dm_trans_pblas(z_ptr, px[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat); + std::vector z_2d = cal_dm_trans_pblas(z_ptr, px[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat.para_2d()); for (auto& t: z_2d) { - LR_Util::matsym(t.data(), naos, pmat); + LR_Util::matsym(t.data(), naos, pmat.para_2d()); } - dm_buf = cal_dm_diff_pblas(x_ptr, px[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat); + dm_buf = cal_dm_diff_pblas(x_ptr, px[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat.para_2d()); #else std::vector z_2d = cal_dm_trans_blas(z_ptr, psi_ks_is, nocc[is], nvirt[is]); for (auto& t: z_2d) diff --git a/source/source_lcao/module_lr/hamilt_casida.h b/source/source_lcao/module_lr/hamilt_casida.h index a9e7aa5c6f..db84baa876 100644 --- a/source/source_lcao/module_lr/hamilt_casida.h +++ b/source/source_lcao/module_lr/hamilt_casida.h @@ -138,8 +138,8 @@ namespace LR { const auto psi_ks_is = LR_Util::get_psi_spin(psi_ks_in, is, nk); #ifdef __MPI - std::vector dm_trans_2d = cal_dm_trans_pblas(X, pX[is], psi_ks_is, pc_in, naos, nocc[is], nvirt[is], pmat_in, (T)1.0 / (T)nk); - if (this->tdm_sym) for (auto& t : dm_trans_2d) LR_Util::matsym(t.data(), naos, pmat_in); + std::vector dm_trans_2d = cal_dm_trans_pblas(X, pX[is], psi_ks_is, pc_in, naos, nocc[is], nvirt[is], pmat_in.para_2d(), (T)1.0 / (T)nk); + if (this->tdm_sym) for (auto& t : dm_trans_2d) LR_Util::matsym(t.data(), naos, pmat_in.para_2d()); #else std::vector dm_trans_2d = cal_dm_trans_blas(X, psi_ks_is, nocc[is], nvirt[is], (T)1.0 / (T)nk); if (this->tdm_sym) for (auto& t : dm_trans_2d) LR_Util::matsym(t.data(), naos); diff --git a/source/source_lcao/module_lr/hamilt_ulr.cpp b/source/source_lcao/module_lr/hamilt_ulr.cpp index 5775b29825..e951eb383e 100644 --- a/source/source_lcao/module_lr/hamilt_ulr.cpp +++ b/source/source_lcao/module_lr/hamilt_ulr.cpp @@ -76,10 +76,10 @@ HamiltULR::HamiltULR(const std::string& xc_kernel, const auto psi_ks_is = LR_Util::get_psi_spin(psi_ks_in, is, nk); // LR_Util::print_value(X, pX_in[is].get_local_size()); #ifdef __MPI - std::vector dm_trans_2d = cal_dm_trans_pblas(X, pX[is], psi_ks_is, pc_in, naos, nocc[is], nvirt[is], pmat_in, (T)1.0 / (T)nk); + std::vector dm_trans_2d = cal_dm_trans_pblas(X, pX[is], psi_ks_is, pc_in, naos, nocc[is], nvirt[is], pmat_in.para_2d(), (T)1.0 / (T)nk); if (this->tdm_sym) for (auto& t: dm_trans_2d) - LR_Util::matsym(t.data(), naos, pmat_in); + LR_Util::matsym(t.data(), naos, pmat_in.para_2d()); #else std::vector dm_trans_2d = cal_dm_trans_blas(X, psi_ks_is, nocc[is], nvirt[is], (T)1.0 / (T)nk); if (this->tdm_sym) diff --git a/source/source_lcao/module_lr/hamilt_zeq_l.h b/source/source_lcao/module_lr/hamilt_zeq_l.h index 057b0bea85..635e79c689 100644 --- a/source/source_lcao/module_lr/hamilt_zeq_l.h +++ b/source/source_lcao/module_lr/hamilt_zeq_l.h @@ -74,9 +74,9 @@ namespace LR { const auto psi_ks_is = LR_Util::get_psi_spin(psi_ks, is, this->nk); #ifdef __MPI - std::vector dm_trans_2d = cal_dm_trans_pblas(X, this->pX[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat); + std::vector dm_trans_2d = cal_dm_trans_pblas(X, this->pX[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat.para_2d()); for (auto& t: dm_trans_2d) - LR_Util::matsym(t.data(), naos, pmat); + LR_Util::matsym(t.data(), naos, pmat.para_2d()); #else std::vector dm_trans_2d = cal_dm_trans_blas(X, psi_ks_is, nocc[is], nvirt[is]); for (auto& t: dm_trans_2d) @@ -184,10 +184,10 @@ namespace LR const auto psi_ks_is = LR_Util::get_psi_spin(this->psi_ks_, is, this->nk); #ifdef __MPI this->dm_buf_ = cal_dm_trans_pblas(X, this->pX[is], psi_ks_is, this->pc_, - this->naos_, this->nocc[is], this->nvirt[is], this->pmat_); + this->naos_, this->nocc[is], this->nvirt[is], this->pmat_.para_2d()); for (auto& t: this->dm_buf_) { - LR_Util::matsym(t.template data(), this->naos_, this->pmat_); + LR_Util::matsym(t.template data(), this->naos_, this->pmat_.para_2d()); } #else this->dm_buf_ = cal_dm_trans_blas(X, psi_ks_is, this->nocc[is], this->nvirt[is]); diff --git a/source/source_lcao/module_lr/hamilt_zeq_r.h b/source/source_lcao/module_lr/hamilt_zeq_r.h index 9722762fd7..0ec0ace05d 100644 --- a/source/source_lcao/module_lr/hamilt_zeq_r.h +++ b/source/source_lcao/module_lr/hamilt_zeq_r.h @@ -127,10 +127,10 @@ namespace LR { const auto psi_ks_is = LR_Util::get_psi_spin(psi_ks, is, this->nk); #ifdef __MPI - std::vector dm_trans_2d = cal_dm_trans_pblas(X, this->pX[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat); + std::vector dm_trans_2d = cal_dm_trans_pblas(X, this->pX[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat.para_2d()); for (auto& t: dm_trans_2d) { - LR_Util::mattrans(t.data(), naos, pmat); + LR_Util::mattrans(t.data(), naos, pmat.para_2d()); } #else std::vector dm_trans_2d = cal_dm_trans_blas(X, psi_ks_is, nocc[is], nvirt[is]); @@ -156,9 +156,9 @@ namespace LR { const auto psi_ks_is = LR_Util::get_psi_spin(psi_ks, is, this->nk); #ifdef __MPI - std::vector dm_diff_2d = cal_dm_diff_pblas(X, this->pX[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat); + std::vector dm_diff_2d = cal_dm_diff_pblas(X, this->pX[is], psi_ks_is, pc, naos, nocc[is], nvirt[is], pmat.para_2d()); for (auto& t: dm_diff_2d) - LR_Util::matsym(t.data(), naos, pmat); + LR_Util::matsym(t.data(), naos, pmat.para_2d()); #else std::vector dm_diff_2d = cal_dm_diff_blas(X, psi_ks_is, naos, nocc[is], nvirt[is]); for (auto& t: dm_diff_2d) @@ -331,16 +331,16 @@ namespace LR const auto psi_ks_is = LR_Util::get_psi_spin(this->psi_ks_, is, this->nk); #ifdef __MPI this->dmx_buf_ = cal_dm_trans_pblas(X, this->pX[is], psi_ks_is, this->pc_, - this->naos_, this->nocc[is], this->nvirt[is], this->pmat_); + this->naos_, this->nocc[is], this->nvirt[is], this->pmat_.para_2d()); for (auto& t: this->dmx_buf_) { - LR_Util::mattrans(t.template data(), this->naos_, this->pmat_); + LR_Util::mattrans(t.template data(), this->naos_, this->pmat_.para_2d()); } this->dmd_buf_ = cal_dm_diff_pblas(X, this->pX[is], psi_ks_is, this->pc_, - this->naos_, this->nocc[is], this->nvirt[is], this->pmat_); + this->naos_, this->nocc[is], this->nvirt[is], this->pmat_.para_2d()); for (auto& t: this->dmd_buf_) { - LR_Util::matsym(t.template data(), this->naos_, this->pmat_); + LR_Util::matsym(t.template data(), this->naos_, this->pmat_.para_2d()); } #else this->dmx_buf_ = cal_dm_trans_blas(X, psi_ks_is, this->nocc[is], this->nvirt[is]); diff --git a/source/source_lcao/module_lr/lr_density.cpp b/source/source_lcao/module_lr/lr_density.cpp index ecf931298b..bb0a47ede6 100644 --- a/source/source_lcao/module_lr/lr_density.cpp +++ b/source/source_lcao/module_lr/lr_density.cpp @@ -68,7 +68,7 @@ void LR_Density::cal_eh_density_single_state(const T* const X_istate, const i auto c_spin = LR_Util::get_psi_spin(psi_ks_, orbital_spin, nk_); #ifdef __MPI const std::vector dm_diff_k = - cal_dm_diff_pblas(X_istate, pX_[ispin], c_spin, pc_, nao_, nocc_[ispin], nvirt_[ispin], pmat_); + cal_dm_diff_pblas(X_istate, pX_[ispin], c_spin, pc_, nao_, nocc_[ispin], nvirt_[ispin], pmat_.para_2d()); #else const std::vector dm_diff_k = cal_dm_diff_blas(X_istate, c_spin, nao_, nocc_[ispin], nvirt_[ispin]); diff --git a/source/source_lcao/module_lr/lr_grad_cs.cpp b/source/source_lcao/module_lr/lr_grad_cs.cpp index 35e95e2267..b6164da7b4 100644 --- a/source/source_lcao/module_lr/lr_grad_cs.cpp +++ b/source/source_lcao/module_lr/lr_grad_cs.cpp @@ -47,7 +47,7 @@ std::vector evaluate_closed_shell_force( // The imag part will be cancelled in the force calculation, so we use double DM(R) to calculate force. // But complex transition DM(R) is still used in energy density matrix calculation. #ifdef __MPI - const auto& dm_trans_k = cal_dm_trans_pblas(Xz.data() + offset, paraX_g[ispin], c, inputs.pc, inputs.nbasis, inputs.nocc[ispin], nvirt_g[ispin], inputs.pmat); + const auto& dm_trans_k = cal_dm_trans_pblas(Xz.data() + offset, paraX_g[ispin], c, inputs.pc, inputs.nbasis, inputs.nocc[ispin], nvirt_g[ispin], inputs.pmat.para_2d()); #else const auto& dm_trans_k = cal_dm_trans_blas(Xz.data() + offset, c, inputs.nocc[ispin], nvirt_g[ispin]); #endif @@ -74,7 +74,7 @@ std::vector evaluate_closed_shell_force( // LR_Util::print_DMR(dm_trans, "dm_trans of istate " + std::to_string(istate)); // difference density matrix #ifdef __MPI - std::vector dm_diff_k = cal_dm_diff_pblas(Xz.data() + offset, paraX_g[ispin], c, inputs.pc, inputs.nbasis, inputs.nocc[ispin], nvirt_g[ispin], inputs.pmat); + std::vector dm_diff_k = cal_dm_diff_pblas(Xz.data() + offset, paraX_g[ispin], c, inputs.pc, inputs.nbasis, inputs.nocc[ispin], nvirt_g[ispin], inputs.pmat.para_2d()); #else std::vector dm_diff_k = cal_dm_diff_blas(Xz.data() + offset, c, inputs.nbasis, inputs.nocc[ispin], nvirt_g[ispin]); #endif @@ -85,7 +85,7 @@ std::vector evaluate_closed_shell_force( // LR_Util::print_value(dm_diff_k[0].data(), inputs.pmat.get_col_size(), inputs.pmat.get_row_size()); #ifdef __MPI - const std::vector& dm_relaxed_k = cal_dm_trans_pblas(Z.template data() + zoffset, paraX_g[ispin], c, inputs.pc, inputs.nbasis, inputs.nocc[ispin], nvirt_g[ispin], inputs.pmat); + const std::vector& dm_relaxed_k = cal_dm_trans_pblas(Z.template data() + zoffset, paraX_g[ispin], c, inputs.pc, inputs.nbasis, inputs.nocc[ispin], nvirt_g[ispin], inputs.pmat.para_2d()); #else const std::vector& dm_relaxed_k = cal_dm_trans_blas(Z.template data() + zoffset, c, inputs.nocc[ispin], nvirt_g[ispin]); #endif @@ -94,7 +94,7 @@ std::vector evaluate_closed_shell_force( #ifdef __MPI for (auto& d: dm_relaxed_k) { - LR_Util::matsym(d.data(), inputs.nbasis, inputs.pmat); + LR_Util::matsym(d.data(), inputs.nbasis, inputs.pmat.para_2d()); } // symmetrize #else for (auto& d: dm_relaxed_k) @@ -141,7 +141,7 @@ std::vector evaluate_closed_shell_force( if (inputs.test_force && inputs.nocc[0] == 1 && nvirt_g[0] == 1) { #ifdef __MPI - const std::vector& dm_diff = cal_dm_diff_pblas(Xz.data() + offset, paraX_g[0], c, inputs.pc, inputs.nbasis, inputs.nocc[0], nvirt_g[0], inputs.pmat); + const std::vector& dm_diff = cal_dm_diff_pblas(Xz.data() + offset, paraX_g[0], c, inputs.pc, inputs.nbasis, inputs.nocc[0], nvirt_g[0], inputs.pmat.para_2d()); #else const std::vector& dm_diff = cal_dm_diff_blas(Xz.data() + offset, c, inputs.nbasis, inputs.nocc[0], nvirt_g[0]); #endif diff --git a/source/source_lcao/module_lr/lr_grad_os.cpp b/source/source_lcao/module_lr/lr_grad_os.cpp index fec1ae8a44..7cdc138cbf 100644 --- a/source/source_lcao/module_lr/lr_grad_os.cpp +++ b/source/source_lcao/module_lr/lr_grad_os.cpp @@ -53,14 +53,14 @@ std::vector evaluate_open_shell_force( { #ifdef __MPI dmx_k[is] = cal_dm_trans_pblas(X_istate + off_x[is], paraX_g[is], c_spin[is], inputs.pc, - inputs.nbasis, inputs.nocc[is], nvirt_g[is], inputs.pmat); + inputs.nbasis, inputs.nocc[is], nvirt_g[is], inputs.pmat.para_2d()); dmdiff_k[is] = cal_dm_diff_pblas(X_istate + off_x[is], paraX_g[is], c_spin[is], inputs.pc, - inputs.nbasis, inputs.nocc[is], nvirt_g[is], inputs.pmat); + inputs.nbasis, inputs.nocc[is], nvirt_g[is], inputs.pmat.para_2d()); std::vector dmz_k = cal_dm_trans_pblas(Z_istate + off_x[is], paraX_g[is], c_spin[is], - inputs.pc, inputs.nbasis, inputs.nocc[is], nvirt_g[is], inputs.pmat); + inputs.pc, inputs.nbasis, inputs.nocc[is], nvirt_g[is], inputs.pmat.para_2d()); for (auto& d: dmz_k) { - LR_Util::matsym(d.template data(), inputs.nbasis, inputs.pmat); + LR_Util::matsym(d.template data(), inputs.nbasis, inputs.pmat.para_2d()); } #else dmx_k[is] = cal_dm_trans_blas(X_istate + off_x[is], c_spin[is], inputs.nocc[is], nvirt_g[is]); diff --git a/source/source_lcao/module_lr/lr_spectrum.cpp b/source/source_lcao/module_lr/lr_spectrum.cpp index 2cc82c0581..5739d08628 100644 --- a/source/source_lcao/module_lr/lr_spectrum.cpp +++ b/source/source_lcao/module_lr/lr_spectrum.cpp @@ -19,7 +19,7 @@ module_dm::DensityMatrix LR::LR_Spectrum::cal_transition_density_matrix const int offset_x = offset_b + is * nk * this->pX[0].get_local_size(); //1. transition density #ifdef __MPI - std::vector dm_trans_2d = cal_dm_trans_pblas(X + offset_x, this->pX[is], psi_ks_vec[is], this->pc, this->naos, this->nocc[is], this->nvirt[is], this->pmat, (T)1.0 / (T)nk); + std::vector dm_trans_2d = cal_dm_trans_pblas(X + offset_x, this->pX[is], psi_ks_vec[is], this->pc, this->naos, this->nocc[is], this->nvirt[is], this->pmat.para_2d(), (T)1.0 / (T)nk); // if (this->tdm_sym) for (auto& t : dm_trans_2d) LR_Util::matsym(t.data(), naos, pmat); #else std::vector dm_trans_2d = cal_dm_trans_blas(X + offset_x, this->psi_ks_vec[is], this->nocc[is], this->nvirt[is], (T)1.0 / (T)nk); diff --git a/source/source_lcao/module_lr/operator_casida/op_gxc_ulr.h b/source/source_lcao/module_lr/operator_casida/op_gxc_ulr.h index 005cc64153..19ea414fd5 100644 --- a/source/source_lcao/module_lr/operator_casida/op_gxc_ulr.h +++ b/source/source_lcao/module_lr/operator_casida/op_gxc_ulr.h @@ -76,10 +76,10 @@ namespace LR { #ifdef __MPI dmk[is] = cal_dm_trans_pblas(X_full + off_x[is], pX_[is], psi_spin_[is], pc_, - naos_, nocc_[is], nvirt_[is], pmat_); + naos_, nocc_[is], nvirt_[is], pmat_.para_2d()); for (auto& t: dmk[is]) { - LR_Util::matsym(t.template data(), naos_, pmat_); + LR_Util::matsym(t.template data(), naos_, pmat_.para_2d()); } #else dmk[is] = cal_dm_trans_blas(X_full + off_x[is], psi_spin_[is], nocc_[is], nvirt_[is]); @@ -123,7 +123,7 @@ namespace LR folding_HR(*this->hR_, v_2d[ik].template data(), kv_.kvec_d[ik], nrow, 1); } #ifdef __MPI - ao_to_mo_pblas(v_2d, pmat_, psi_spin_[tau], pc_, naos_, nocc_[tau], nvirt_[tau], + ao_to_mo_pblas(v_2d, pmat_.para_2d(), psi_spin_[tau], pc_, naos_, nocc_[tau], nvirt_[tau], pout_[tau], out_full + off_out[tau], /*add_on=*/true, mo_type_, factor_); #else ao_to_mo_blas(v_2d, psi_spin_[tau], nocc_[tau], nvirt_[tau], diff --git a/source/source_lcao/module_lr/operator_casida/operator_lr_exx.cpp b/source/source_lcao/module_lr/operator_casida/operator_lr_exx.cpp index c0dfaab6ba..1379ecb72a 100644 --- a/source/source_lcao/module_lr/operator_casida/operator_lr_exx.cpp +++ b/source/source_lcao/module_lr/operator_casida/operator_lr_exx.cpp @@ -158,7 +158,8 @@ namespace LR ModuleBase::TITLE("OperatorLREXX", "cal_Hs"); ModuleBase::timer::start("OperatorLREXX", "cal_Hs"); // convert parallel info to LibRI interfaces - std::vector, std::set>> judge = RI_2D_Comm::get_2D_judge(ucell,this->pmat); + std::vector, std::set>> judge + = RI_2D_Comm::get_2D_judge(ucell, this->pmat.para_2d()); // suppose Cs,Vs, have already been calculated in the ion-step of ground state // and DM_trans has been calculated in hPsi() outside. @@ -171,7 +172,7 @@ namespace LR for (int ik = 0;ik < nk;++ik) { DMk_trans_pointer[ik] = &DMk_trans_vector[ik]; } // if multi-k, DM_trans(TR=double) -> Ds_trans(TR=T=complex) auto Ds_trans = - RI_2D_Comm::split_m2D_ktoR(ucell,this->kv, DMk_trans_pointer, this->pmat, 1); + RI_2D_Comm::split_m2D_ktoR(ucell, this->kv, DMk_trans_pointer, this->pmat.para_2d(), 1); // 2. cal_Hs auto lri = this->exx_lri.lock(); diff --git a/source/source_lcao/module_lr/operator_casida/operator_lr_hxc.cpp b/source/source_lcao/module_lr/operator_casida/operator_lr_hxc.cpp index b24411fe96..9d79079fa9 100644 --- a/source/source_lcao/module_lr/operator_casida/operator_lr_hxc.cpp +++ b/source/source_lcao/module_lr/operator_casida/operator_lr_hxc.cpp @@ -58,14 +58,14 @@ namespace LR { case MO_TO_AO_TYPE::CC_vo: //[AX]^{Hxc}_{ai}=\sum_{\mu,\nu}c^*_{\mu,a}V^{Hxc}_{\mu,\nu}c_{\nu,i} #ifdef __MPI - ao_to_mo_pblas(v_hxc_2d, this->pmat, psil_ks, this->pc, this->naos, this->nocc[sl], this->nvirt[sl], this->pX[sl], hpsi, /*add_on=*/true, LR_Util::MO_TYPE::VO, this->factor_); + ao_to_mo_pblas(v_hxc_2d, this->pmat.para_2d(), psil_ks, this->pc, this->naos, this->nocc[sl], this->nvirt[sl], this->pX[sl], hpsi, /*add_on=*/true, LR_Util::MO_TYPE::VO, this->factor_); #else ao_to_mo_blas(v_hxc_2d, psil_ks, this->nocc[sl], this->nvirt[sl], hpsi, /*add_on=*/true, LR_Util::MO_TYPE::VO, this->factor_); #endif break; case MO_TO_AO_TYPE::CC_oo: //[AX]^{Hxc}_{ij}=\sum_{\mu,\nu}c^*_{\mu,i}V^{Hxc}_{\mu,\nu}c_{\nu,j} #ifdef __MPI - ao_to_mo_pblas(v_hxc_2d, this->pmat, psil_ks, this->pc, this->naos, this->nocc[sl], this->nvirt[sl], this->pX[sl], hpsi, /*add_on=*/true, LR_Util::MO_TYPE::OO, this->factor_); + ao_to_mo_pblas(v_hxc_2d, this->pmat.para_2d(), psil_ks, this->pc, this->naos, this->nocc[sl], this->nvirt[sl], this->pX[sl], hpsi, /*add_on=*/true, LR_Util::MO_TYPE::OO, this->factor_); #else ao_to_mo_blas(v_hxc_2d, psil_ks, this->nocc[sl], this->nvirt[sl], hpsi, /*add_on=*/true, LR_Util::MO_TYPE::OO, this->factor_); #endif @@ -73,9 +73,9 @@ namespace LR case MO_TO_AO_TYPE::CXC: { #ifdef __MPI - CVCX_virt_pblas(v_hxc_2d, this->pmat, psil_ks, this->pc, psi_in, this->pX[sl], + CVCX_virt_pblas(v_hxc_2d, this->pmat.para_2d(), psil_ks, this->pc, psi_in, this->pX[sl], this->naos, this->nocc[sl], this->nvirt[sl], hpsi, /*add_on=*/true, this->factor_); - CVCX_occ_pblas(v_hxc_2d, this->pmat, psil_ks, this->pc, psi_in, this->pX[sl], + CVCX_occ_pblas(v_hxc_2d, this->pmat.para_2d(), psil_ks, this->pc, psi_in, this->pX[sl], this->naos, this->nocc[sl], this->nvirt[sl], hpsi, /*add_on=*/true, -this->factor_); #else CVCX_virt_blas(v_hxc_2d, psil_ks, psi_in, this->naos, this->nocc[sl], this->nvirt[sl], hpsi, /*add_on=*/true, this->factor_); @@ -85,7 +85,7 @@ namespace LR } case MO_TO_AO_TYPE::CXC_o: #ifdef __MPI - CVCX_occ_pblas(v_hxc_2d, this->pmat, psil_ks, this->pc, psi_in, this->pX[sl], + CVCX_occ_pblas(v_hxc_2d, this->pmat.para_2d(), psil_ks, this->pc, psi_in, this->pX[sl], this->naos, this->nocc[sl], this->nvirt[sl], hpsi, /*add_on=*/true, this->factor_); #else CVCX_occ_blas(v_hxc_2d, psil_ks, psi_in, this->naos, this->nocc[sl], this->nvirt[sl], hpsi, /*add_on=*/true, this->factor_); diff --git a/source/source_lcao/module_lr/state_track.cpp b/source/source_lcao/module_lr/state_track.cpp index 581339d439..37200d16b4 100644 --- a/source/source_lcao/module_lr/state_track.cpp +++ b/source/source_lcao/module_lr/state_track.cpp @@ -90,9 +90,9 @@ void project_local_references(const RootInputs& inputs, const RootBasis& o // Only oo/vv are needed; construct each once and reuse across references. const int occupied_start = 0; const auto occupied = mo_overlap_dist(old.coefficients[spin], typed_ao, - current.coefficients[spin], inputs.pc, inputs.pmat, pi, occupied_start); + current.coefficients[spin], inputs.pc, inputs.pmat.para_2d(), pi, occupied_start); const auto virtuals = mo_overlap_dist(old.coefficients[spin], typed_ao, - current.coefficients[spin], inputs.pc, inputs.pmat, pv, no); + current.coefficients[spin], inputs.pc, inputs.pmat.para_2d(), pv, no); for (auto* reference : references) { std::vector local_reference(local_count); diff --git a/source/source_lcao/module_lr/utils/lr_util_hcontainer.h b/source/source_lcao/module_lr/utils/lr_util_hcontainer.h index 03b3dfd817..663c950a9a 100644 --- a/source/source_lcao/module_lr/utils/lr_util_hcontainer.h +++ b/source/source_lcao/module_lr/utils/lr_util_hcontainer.h @@ -233,7 +233,7 @@ namespace LR_Util #ifdef __MPI // dm(k) is 2D-block-cyclic distributed, so the transpose needs the PBLAS routine // `mattrans` (pdtran_/pztranc_) rather than a plain serial swap. - LR_Util::mattrans(dk.data(), pv.para_2d().get_global_row_size(), pv); + LR_Util::mattrans(dk.data(), pv.para_2d().get_global_row_size(), pv.para_2d()); #else // In a serial build the entire square AO matrix is local, in column-major order. const int n = pv.para_2d().get_global_row_size(); @@ -260,7 +260,7 @@ namespace LR_Util for (auto& dk : dm.get_dmk_vec()) { #ifdef __MPI - LR_Util::mattrans(dk.data(), pv.para_2d().get_global_row_size(), pv); + LR_Util::mattrans(dk.data(), pv.para_2d().get_global_row_size(), pv.para_2d()); #else throw std::runtime_error("transpose_DMR requires MPI (PBLAS mattrans) for the 2D-block-cyclic dm(k) transpose."); #endif @@ -288,7 +288,7 @@ namespace LR_Util if (symmetrize) for (int ik = 0; ik < nk; ++ik) #ifdef __MPI - LR_Util::matsym(dmk[ik].data(), pmat.para_2d().get_global_row_size(), pmat); + LR_Util::matsym(dmk[ik].data(), pmat.para_2d().get_global_row_size(), pmat.para_2d()); #else LR_Util::matsym(dmk[ik].data(), pmat.para_2d().get_global_row_size()); #endif @@ -331,7 +331,7 @@ namespace LR_Util for (int ik = 0; ik < nk; ++ik) { #ifdef __MPI - LR_Util::matsym(dmk[is][ik].data(), pmat.para_2d().get_global_row_size(), pmat); + LR_Util::matsym(dmk[is][ik].data(), pmat.para_2d().get_global_row_size(), pmat.para_2d()); #else LR_Util::matsym(dmk[is][ik].data(), pmat.para_2d().get_global_row_size()); #endif @@ -375,7 +375,8 @@ namespace LR_Util { DMk_trans_pointer[ik] = &dm.get_dmk_vec()[ik]; } - return RI_2D_Comm::split_m2D_ktoR(ucell, kv, DMk_trans_pointer, pmat, /*nspin=*/1)[0]; + return RI_2D_Comm::split_m2D_ktoR( + ucell, kv, DMk_trans_pointer, pmat.para_2d(), /*nspin=*/1)[0]; } // return SPIN_multiple*D[0] as implemented in split_m2D_ktoR // SPIN_multiple = map({ {1,0.5}, {2,1}, {4,1} }).at(nspin) @@ -389,7 +390,8 @@ namespace LR_Util std::vector*> DMk_trans_pointer(nks); for (int iks = 0;iks < dm.get_DMK_nks();++iks) DMk_trans_pointer[iks] = &dm.get_dmk_vec()[iks]; - return RI_2D_Comm::split_m2D_ktoR(ucell, kv, DMk_trans_pointer, pmat, nspin); + return RI_2D_Comm::split_m2D_ktoR( + ucell, kv, DMk_trans_pointer, pmat.para_2d(), nspin); } #endif } diff --git a/source/source_lcao/module_lr/utils/spectrum_mo.cpp b/source/source_lcao/module_lr/utils/spectrum_mo.cpp index 10e4873ca6..f30e514f64 100644 --- a/source/source_lcao/module_lr/utils/spectrum_mo.cpp +++ b/source/source_lcao/module_lr/utils/spectrum_mo.cpp @@ -88,7 +88,7 @@ std::vector> cal_velocity_mo(const UnitCell& ucell, } } #ifdef __MPI - LR::ao_to_mo_pblas(vk, pmat, c_psi_ks, pc, nbasis, + LR::ao_to_mo_pblas(vk, pmat.para_2d(), c_psi_ks, pc, nbasis, nocc[0], nvirt[0], pmo, v_mo.data(), false, // add_on LR_Util::MO_TYPE::ALL, unit_factor); diff --git a/source/source_lcao/module_rdmft/rdmft.cpp b/source/source_lcao/module_rdmft/rdmft.cpp index 204ba6e8c1..64340e655e 100644 --- a/source/source_lcao/module_rdmft/rdmft.cpp +++ b/source/source_lcao/module_rdmft/rdmft.cpp @@ -163,7 +163,7 @@ void RDMFT::init(Parallel_Orbitals& ParaV_in, this->symrot_exx.cal_Ms(*kv, ucell->symm, ucell->atoms, ucell->latvec, ucell->lmax, ucell->nat, ucell->iat2it, ucell->iat2ia, - *ParaV, nspin); + ParaV->para_2d(), nspin); } if (this->exx_info_->info_ri.real_number) diff --git a/source/source_lcao/module_rdmft/rdmft_pot.cpp b/source/source_lcao/module_rdmft/rdmft_pot.cpp index 8b70669071..573deaa03b 100644 --- a/source/source_lcao/module_rdmft/rdmft_pot.cpp +++ b/source/source_lcao/module_rdmft/rdmft_pot.cpp @@ -233,7 +233,7 @@ void RDMFT::cal_V_XC(const UnitCell& ucell) // get DM_XC of all k points if( exx_spacegroup_symmetry ) { - DM_XC = symrot_exx.restore_dm(*this->kv, DM_XC, *ParaV); // class vector could be auto resize() + DM_XC = symrot_exx.restore_dm(*this->kv, DM_XC, ParaV->para_2d()); // class vector could be auto resize() } std::vector< const std::vector* > DM_XC_pointer(DM_XC.size()); for(int ik=0; ik::cal_V_XC(const UnitCell& ucell) // transfer the DM_XC to appropriate format std::vector>,RI::Tensor>>> Ds_XC_d = std::is_same::value //gamma_only_local - ? RI_2D_Comm::split_m2D_ktoR(ucell,*kv, DM_XC_pointer, *ParaV, nspin) - : RI_2D_Comm::split_m2D_ktoR(ucell,*kv, DM_XC_pointer, *ParaV, nspin, this->exx_spacegroup_symmetry); + ? RI_2D_Comm::split_m2D_ktoR(ucell,*kv, DM_XC_pointer, ParaV->para_2d(), nspin) + : RI_2D_Comm::split_m2D_ktoR(ucell,*kv, DM_XC_pointer, ParaV->para_2d(), nspin, this->exx_spacegroup_symmetry); // provide the Ds_XC to Vxc_fromRI(V_exx_XC) if (this->exx_spacegroup_symmetry && this->exx_info_->info_ri.exx_symmetry_realspace) { - Vxc_fromRI_d->cal_exx_elec(Ds_XC_d, ucell,*ParaV, &this->symrot_exx); + Vxc_fromRI_d->cal_exx_elec(Ds_XC_d, ucell, *ParaV, &this->symrot_exx); } else { - Vxc_fromRI_d->cal_exx_elec(Ds_XC_d, ucell,*ParaV); + Vxc_fromRI_d->cal_exx_elec(Ds_XC_d, ucell, *ParaV); } // when we doing V_exx_XC.contributeHk(ik), we get HK_XC constructed by the special DM_XC @@ -274,17 +274,17 @@ void RDMFT::cal_V_XC(const UnitCell& ucell) // transfer the DM_XC to appropriate format std::vector>,RI::Tensor>>>> Ds_XC_c = std::is_same::value //gamma_only_local - ? RI_2D_Comm::split_m2D_ktoR>(ucell,*kv, DM_XC_pointer, *ParaV, nspin) - : RI_2D_Comm::split_m2D_ktoR>(ucell,*kv, DM_XC_pointer, *ParaV, nspin, this->exx_spacegroup_symmetry); + ? RI_2D_Comm::split_m2D_ktoR>(ucell,*kv, DM_XC_pointer, ParaV->para_2d(), nspin) + : RI_2D_Comm::split_m2D_ktoR>(ucell,*kv, DM_XC_pointer, ParaV->para_2d(), nspin, this->exx_spacegroup_symmetry); // // provide the Ds_XC to Vxc_fromRI(V_exx_XC) if (this->exx_spacegroup_symmetry && this->exx_info_->info_ri.exx_symmetry_realspace) { - Vxc_fromRI_c->cal_exx_elec(Ds_XC_c, ucell,*ParaV, &this->symrot_exx); + Vxc_fromRI_c->cal_exx_elec(Ds_XC_c, ucell, *ParaV, &this->symrot_exx); } else { - Vxc_fromRI_c->cal_exx_elec(Ds_XC_c, ucell,*ParaV); + Vxc_fromRI_c->cal_exx_elec(Ds_XC_c, ucell, *ParaV); } // when we doing V_exx_XC.contributeHk(ik), we get HK_XC constructed by the special DM_XC diff --git a/source/source_lcao/module_ri/exx_lri.hpp b/source/source_lcao/module_ri/exx_lri.hpp index 926f64c915..0bdd5a621f 100644 --- a/source/source_lcao/module_ri/exx_lri.hpp +++ b/source/source_lcao/module_ri/exx_lri.hpp @@ -856,7 +856,8 @@ void Exx_LRI::cal_exx_elec(const std::vector, std::set>> judge = RI_2D_Comm::get_2D_judge(ucell,pv); + const std::vector, std::set>> judge + = RI_2D_Comm::get_2D_judge(ucell, pv.para_2d()); if(p_symrot) { this->exx_lri.set_symmetry(true, p_symrot->get_irred_sector()); } diff --git a/source/source_lcao/module_ri/exx_lri_interface.hpp b/source/source_lcao/module_ri/exx_lri_interface.hpp index 12b1d2f946..adc53d1d97 100644 --- a/source/source_lcao/module_ri/exx_lri_interface.hpp +++ b/source/source_lcao/module_ri/exx_lri_interface.hpp @@ -232,7 +232,7 @@ void Exx_LRI_Interface::exx_eachiterinit(const int istep, auto cal = [this, &ucell,&kv, &pv, &flag_restart](const module_dm::DensityMatrix& dm_in) { if (this->exx_spacegroup_symmetry) - { this->mix_DMk_2D.mix(symrot_.restore_dm(kv, dm_in.get_dmk_vec(), pv), flag_restart); } + { this->mix_DMk_2D.mix(symrot_.restore_dm(kv, dm_in.get_dmk_vec(), pv.para_2d()), flag_restart); } else { this->mix_DMk_2D.mix(dm_in.get_dmk_vec(), flag_restart); } const std::vector>>> Ds = @@ -240,7 +240,7 @@ void Exx_LRI_Interface::exx_eachiterinit(const int istep, ucell, *this->exx_ptr->p_kv, this->mix_DMk_2D.get_DMk_out(), - pv, + pv.para_2d(), PARAM.inp.nspin, this->exx_spacegroup_symmetry); if(this->exx_spacegroup_symmetry && this->exx_ptr->info.exx_symmetry_realspace) @@ -448,7 +448,7 @@ bool Exx_LRI_Interface::exx_after_converge( const bool flag_restart = (this->two_level_step == 0 && PARAM.inp.init_wfc != "file") ? true : false; if(this->exx_spacegroup_symmetry) - { this->mix_DMk_2D.mix(symrot_.restore_dm(kv, dm.get_dmk_vec(), pv), flag_restart); } + { this->mix_DMk_2D.mix(symrot_.restore_dm(kv, dm.get_dmk_vec(), pv.para_2d()), flag_restart); } else { this->mix_DMk_2D.mix(dm.get_dmk_vec(), flag_restart); } const std::vector>>> Ds = @@ -456,7 +456,7 @@ bool Exx_LRI_Interface::exx_after_converge( ucell, *this->exx_ptr->p_kv, this->mix_DMk_2D.get_DMk_out(), - pv, + pv.para_2d(), nspin, this->exx_spacegroup_symmetry); if(this->exx_spacegroup_symmetry && this->exx_ptr->info.exx_symmetry_realspace) diff --git a/source/source_lcao/module_rpa/rpa_lri.cpp b/source/source_lcao/module_rpa/rpa_lri.cpp index 7c9cde7ba7..4d60adbb10 100644 --- a/source/source_lcao/module_rpa/rpa_lri.cpp +++ b/source/source_lcao/module_rpa/rpa_lri.cpp @@ -208,12 +208,12 @@ void RPA_LRI::cal_postSCF_exx(const module_dm::DensityMatrix ucell.nat, ucell.iat2it, ucell.iat2ia, - parav, + parav.para_2d(), this->runtime.input.nspin); // output Ts (symrot_R.txt) and Ms (symrot_k.txt) ModuleSymmetry::print_symrot_info_R(symrot, ucell.symm, ucell.lmax, Rs); ModuleSymmetry::print_symrot_info_k(symrot, kv, ucell); - mix_DMk_2D.mix(symrot.restore_dm(kv, dm.get_dmk_vec(), parav), true); + mix_DMk_2D.mix(symrot.restore_dm(kv, dm.get_dmk_vec(), parav.para_2d()), true); } else { @@ -224,7 +224,7 @@ void RPA_LRI::cal_postSCF_exx(const module_dm::DensityMatrix = RI_2D_Comm::split_m2D_ktoR(ucell, kv, mix_DMk_2D.get_DMk_out(), - parav, + parav.para_2d(), this->runtime.input.nspin, exx_spacegroup_symmetry); diff --git a/source/source_lcao/module_rt/test/CMakeLists.txt b/source/source_lcao/module_rt/test/CMakeLists.txt index fe73cb269f..adedf2483f 100644 --- a/source/source_lcao/module_rt/test/CMakeLists.txt +++ b/source/source_lcao/module_rt/test/CMakeLists.txt @@ -34,7 +34,8 @@ endif() addtest( TARGET MODULE_LCAO_tddft_middle_hamilt_test LIBS parameter base device tddft_test_lib - SOURCES middle_hamilt_test.cpp ../middle_hamilt.cpp) + SOURCES middle_hamilt_test.cpp ../middle_hamilt.cpp + ../../../source_basis/module_ao/parallel_orbitals.cpp) addtest( TARGET MODULE_LCAO_tddft_band_energy_test @@ -46,12 +47,14 @@ addtest( addtest( TARGET MODULE_LCAO_tddft_norm_psi_test LIBS parameter base device tddft_test_lib - SOURCES norm_psi_test.cpp ../norm_psi.cpp) + SOURCES norm_psi_test.cpp ../norm_psi.cpp + ../../../source_basis/module_ao/parallel_orbitals.cpp) addtest( TARGET MODULE_LCAO_tddft_upsi_test LIBS parameter base device tddft_test_lib - SOURCES upsi_test1.cpp upsi_test2.cpp upsi_test3.cpp ../upsi.cpp) + SOURCES upsi_test1.cpp upsi_test2.cpp upsi_test3.cpp ../upsi.cpp + ../../../source_basis/module_ao/parallel_orbitals.cpp) addtest( TARGET MODULE_LCAO_tddft_propagator_test @@ -62,7 +65,8 @@ addtest( ../propagator.cpp ../propagator_cn2.cpp ../propagator_taylor.cpp - ../propagator_etrs.cpp) + ../propagator_etrs.cpp + ../../../source_basis/module_ao/parallel_orbitals.cpp) addtest( TARGET MODULE_LCAO_tddft_snap_psibeta_half_test diff --git a/source/source_lcao/module_rt/test/middle_hamilt_test.cpp b/source/source_lcao/module_rt/test/middle_hamilt_test.cpp index e7f240c6ad..be089f4533 100644 --- a/source/source_lcao/module_rt/test/middle_hamilt_test.cpp +++ b/source/source_lcao/module_rt/test/middle_hamilt_test.cpp @@ -19,12 +19,6 @@ */ #define doublethreshold 1e-8 -Parallel_Orbitals::Parallel_Orbitals() -{ -} -Parallel_Orbitals::~Parallel_Orbitals() -{ -} TEST(MiddleHamiltTest, testMiddleHamilt) { @@ -37,15 +31,7 @@ TEST(MiddleHamiltTest, testMiddleHamilt) bool print_matrix = false; Parallel_Orbitals* pv; pv = new Parallel_Orbitals(); - pv->para_2d().nloc = nlocal * nlocal; - pv->para_2d().ncol = nlocal; - pv->para_2d().nrow = nlocal; - - // Initialize input matrices - int info; - int mb = 1, nb = 1, lda = nlocal, ldc = nlocal; - int irsrc = 0, icsrc = 0, lld = numroc_(&nlocal, &mb, &myprow, &irsrc, &nprow); - descinit_(pv->para_2d().desc, &nlocal, &nlocal, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); + ASSERT_EQ(pv->para_2d().set(nlocal, nlocal, 1, ictxt), 0); // Initialize data Htmp = new std::complex[nlocal * nlocal]; diff --git a/source/source_lcao/module_rt/test/norm_psi_test.cpp b/source/source_lcao/module_rt/test/norm_psi_test.cpp index 5af72e04cb..e0bee3382c 100644 --- a/source/source_lcao/module_rt/test/norm_psi_test.cpp +++ b/source/source_lcao/module_rt/test/norm_psi_test.cpp @@ -19,12 +19,6 @@ */ #define doublethreshold 1e-8 -Parallel_Orbitals::Parallel_Orbitals() -{ -} -Parallel_Orbitals::~Parallel_Orbitals() -{ -} TEST(NormPsiTest, testNormPsi) { @@ -35,29 +29,10 @@ TEST(NormPsiTest, testNormPsi) bool print_matrix = false; Parallel_Orbitals* pv; pv = new Parallel_Orbitals(); - pv->para_2d().nloc = nlocal * nlocal; + ASSERT_EQ(pv->para_2d().set(nlocal, nlocal, 1, ictxt), 0); pv->nloc_wfc = nlocal * nband; - pv->para_2d().ncol = nlocal; - pv->para_2d().nrow = nlocal; pv->ncol_bands = nband; - pv->para_2d().dim0 = 1; - pv->para_2d().dim1 = 1; - pv->para_2d().nb = 1; - pv->para_2d().blacs_ctxt = 0; - pv->para_2d().set_coord(0, 0); - - int dim[2]; - dim[0] = nprow; - dim[1] = npcol; - - // Initialize input matrices - int info; - int mb = 1, nb = 1, lda = nband, ldc = nlocal; - int irsrc = 0, icsrc = 0, lld = numroc_(&nlocal, &mb, &myprow, &irsrc, &nprow), - lld1 = numroc_(&nband, &mb, &myprow, &irsrc, &nprow); - descinit_(pv->para_2d().desc, &nlocal, &nlocal, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); - descinit_(pv->desc_wfc, &nlocal, &nband, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); - descinit_(pv->desc_Eij, &nband, &nband, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); + pv->set_desc_wfc_Eij(nlocal, nband, pv->para_2d().get_row_size()); // Initialize data Stmp = new std::complex[nlocal * nlocal]; @@ -129,23 +104,10 @@ TEST(NormPsiTest, EndpointMetricDiffersFromPropagationMetric) Parallel_Orbitals pv; const int nlocal = 2; const int nband = 1; - const int block = 1; - const int zero = 0; - pv.para_2d().nloc = 4; + ASSERT_EQ(pv.para_2d().set(nlocal, nlocal, 1, ictxt), 0); pv.nloc_wfc = 2; - pv.para_2d().nrow = pv.para_2d().ncol = 2; pv.ncol_bands = 1; - pv.para_2d().dim0 = pv.para_2d().dim1 = pv.para_2d().nb = 1; - pv.para_2d().blacs_ctxt = ictxt; - pv.para_2d().set_coord(0, 0); - int info = 0; - int lld = 2; - descinit_(pv.para_2d().desc, &nlocal, &nlocal, &block, &block, &zero, &zero, &ictxt, &lld, &info); - ASSERT_EQ(info, 0); - descinit_(pv.desc_wfc, &nlocal, &nband, &block, &block, &zero, &zero, &ictxt, &lld, &info); - ASSERT_EQ(info, 0); - descinit_(pv.desc_Eij, &nband, &nband, &block, &block, &zero, &zero, &ictxt, &lld, &info); - ASSERT_EQ(info, 0); + pv.set_desc_wfc_Eij(nlocal, nband, pv.para_2d().get_row_size()); std::complex endpoint[4] = {2.0, {0.0, -0.4}, {0.0, 0.4}, 1.0}; std::complex midpoint[4] = {1.5, {0.0, -0.2}, {0.0, 0.2}, 1.0}; std::complex right[2] = {1.0, {0.0, 0.5}}; diff --git a/source/source_lcao/module_rt/test/propagator_test1.cpp b/source/source_lcao/module_rt/test/propagator_test1.cpp index ea9d3b3703..cfb3b270c6 100644 --- a/source/source_lcao/module_rt/test/propagator_test1.cpp +++ b/source/source_lcao/module_rt/test/propagator_test1.cpp @@ -17,12 +17,6 @@ * - compute propagator of method Crank-Nicolson. */ #define doublethreshold 1e-8 -Parallel_Orbitals::Parallel_Orbitals() -{ -} -Parallel_Orbitals::~Parallel_Orbitals() -{ -} TEST(PropagatorTest, testPropagatorCN) { @@ -33,17 +27,9 @@ TEST(PropagatorTest, testPropagatorCN) bool print_matrix = false; Parallel_Orbitals* pv; pv = new Parallel_Orbitals(); - pv->para_2d().nloc = nlocal * nlocal; - pv->para_2d().ncol = nlocal; - pv->para_2d().set_coord(0, 0); + ASSERT_EQ(pv->para_2d().set(nlocal, nlocal, 1, ictxt), 0); const double md_dt = 4 * ModuleBase::AU_to_FS; - // Initialize input matrices - int info; - int mb = 1, nb = 1, lda = nlocal, ldc = nlocal; - int irsrc = 0, icsrc = 0, lld = numroc_(&nlocal, &mb, &myprow, &irsrc, &nprow); - descinit_(pv->para_2d().desc, &nlocal, &nlocal, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); - // Initialize data U_operator = new std::complex[nlocal * nlocal]; Stmp = new std::complex[nlocal * nlocal]; diff --git a/source/source_lcao/module_rt/test/propagator_test2.cpp b/source/source_lcao/module_rt/test/propagator_test2.cpp index 78035faba1..e352bee3a6 100644 --- a/source/source_lcao/module_rt/test/propagator_test2.cpp +++ b/source/source_lcao/module_rt/test/propagator_test2.cpp @@ -28,27 +28,10 @@ TEST(PropagatorTest, testPropagatorTaylor) bool print_matrix = false; Parallel_Orbitals* pv; pv = new Parallel_Orbitals(); - pv->para_2d().nloc = nlocal * nlocal; - pv->para_2d().ncol = nlocal; - pv->para_2d().nrow = nlocal; - pv->para_2d().dim0 = 1; - pv->para_2d().dim1 = 1; - pv->para_2d().nb = 1; - pv->para_2d().blacs_ctxt = 0; - pv->para_2d().set_coord(0, 0); - - int dim[2]; - dim[0] = nprow; - dim[1] = npcol; + ASSERT_EQ(pv->para_2d().set(nlocal, nlocal, 1, ictxt), 0); const double md_dt = 4 * ModuleBase::AU_to_FS; - // Initialize input matrices - int info; - int mb = 1, nb = 1, lda = nlocal, ldc = nlocal; - int irsrc = 0, icsrc = 0, lld = numroc_(&nlocal, &mb, &myprow, &irsrc, &nprow); - descinit_(pv->para_2d().desc, &nlocal, &nlocal, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); - // Initialize data U_operator = new std::complex[nlocal * nlocal]; Stmp = new std::complex[nlocal * nlocal]; diff --git a/source/source_lcao/module_rt/test/propagator_test3.cpp b/source/source_lcao/module_rt/test/propagator_test3.cpp index 1bb00cdd19..7306b81e64 100644 --- a/source/source_lcao/module_rt/test/propagator_test3.cpp +++ b/source/source_lcao/module_rt/test/propagator_test3.cpp @@ -29,27 +29,10 @@ TEST(PropagatorTest, testPropagatorETRS) bool print_matrix = false; Parallel_Orbitals* pv; pv = new Parallel_Orbitals(); - pv->para_2d().nloc = nlocal * nlocal; - pv->para_2d().ncol = nlocal; - pv->para_2d().nrow = nlocal; - pv->para_2d().dim0 = 1; - pv->para_2d().dim1 = 1; - pv->para_2d().nb = 1; - pv->para_2d().blacs_ctxt = 0; - pv->para_2d().set_coord(0, 0); - - int dim[2]; - dim[0] = nprow; - dim[1] = npcol; + ASSERT_EQ(pv->para_2d().set(nlocal, nlocal, 1, ictxt), 0); const double md_dt = 4 * ModuleBase::AU_to_FS; - // Initialize input matrices - int info; - int mb = 1, nb = 1, lda = nlocal, ldc = nlocal; - int irsrc = 0, icsrc = 0, lld = numroc_(&nlocal, &mb, &myprow, &irsrc, &nprow); - descinit_(pv->para_2d().desc, &nlocal, &nlocal, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); - // Initialize data U_operator = new std::complex[nlocal * nlocal]; Stmp = new std::complex[nlocal * nlocal]; diff --git a/source/source_lcao/module_rt/test/test_evolve_psi.cpp b/source/source_lcao/module_rt/test/test_evolve_psi.cpp index 85b2f8f12f..0ef8cfb781 100644 --- a/source/source_lcao/module_rt/test/test_evolve_psi.cpp +++ b/source/source_lcao/module_rt/test/test_evolve_psi.cpp @@ -42,23 +42,10 @@ class EvolvePsiTest : public testing::Test { const int nlocal = 2; const int nband = 1; - const int block = 1; - const int zero = 0; - const int lld = 2; - int info = 0; - pv.para_2d().nloc = 4; + ASSERT_EQ(pv.para_2d().set(nlocal, nlocal, 1, ictxt), 0); pv.nloc_wfc = 2; - pv.para_2d().nrow = pv.para_2d().ncol = 2; pv.nrow_bands = pv.ncol_bands = 1; - pv.para_2d().dim0 = pv.para_2d().dim1 = pv.para_2d().nb = 1; - pv.para_2d().blacs_ctxt = ictxt; - pv.para_2d().set_coord(0, 0); - descinit_(pv.para_2d().desc, &nlocal, &nlocal, &block, &block, &zero, &zero, &ictxt, &lld, &info); - ASSERT_EQ(info, 0); - descinit_(pv.desc_wfc, &nlocal, &nband, &block, &block, &zero, &zero, &ictxt, &lld, &info); - ASSERT_EQ(info, 0); - descinit_(pv.desc_Eij, &nband, &nband, &block, &block, &zero, &zero, &ictxt, &lld, &info); - ASSERT_EQ(info, 0); + pv.set_desc_wfc_Eij(nlocal, nband, pv.para_2d().get_row_size()); } void expect_endpoint_norm(const Complex* psi) diff --git a/source/source_lcao/module_rt/test/upsi_test1.cpp b/source/source_lcao/module_rt/test/upsi_test1.cpp index a1e42720d0..30e835f59e 100644 --- a/source/source_lcao/module_rt/test/upsi_test1.cpp +++ b/source/source_lcao/module_rt/test/upsi_test1.cpp @@ -18,13 +18,6 @@ * - upsi * - apply U_operator to the wave function of the previous step for new wave function. */ -Parallel_Orbitals::Parallel_Orbitals() -{ -} -Parallel_Orbitals::~Parallel_Orbitals() -{ -} - TEST(UpsiTest, testUpsi1) { std::complex* U_operator; @@ -35,16 +28,9 @@ TEST(UpsiTest, testUpsi1) bool print_matrix = false; Parallel_Orbitals* pv; pv = new Parallel_Orbitals(); - pv->para_2d().nloc = nlocal * nband; - pv->para_2d().ncol = nlocal; + ASSERT_EQ(pv->para_2d().set(nlocal, nlocal, 1, ictxt), 0); pv->ncol_bands = nband; - - // Initialize input matrices - int info; - int mb = 1, nb = 1, lda = nband, ldc = nlocal; - int irsrc = 0, icsrc = 0, lld = numroc_(&nlocal, &mb, &myprow, &irsrc, &nprow); - descinit_(pv->para_2d().desc, &nlocal, &nlocal, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); - descinit_(pv->desc_wfc, &nlocal, &nband, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); + pv->set_desc_wfc_Eij(nlocal, nband, pv->para_2d().get_row_size()); // Initialize data U_operator = new std::complex[nlocal * nlocal]; diff --git a/source/source_lcao/module_rt/test/upsi_test2.cpp b/source/source_lcao/module_rt/test/upsi_test2.cpp index 6aba967885..705a2e8a84 100644 --- a/source/source_lcao/module_rt/test/upsi_test2.cpp +++ b/source/source_lcao/module_rt/test/upsi_test2.cpp @@ -29,16 +29,9 @@ TEST(UpsiTest, testUpsi2) bool print_matrix = false; Parallel_Orbitals* pv; pv = new Parallel_Orbitals(); - pv->para_2d().nloc = nlocal * nband; - pv->para_2d().ncol = nlocal; + ASSERT_EQ(pv->para_2d().set(nlocal, nlocal, 1, ictxt), 0); pv->ncol_bands = nband; - - // Initialize input matrices - int info; - int mb = 1, nb = 1, lda = nband, ldc = nlocal; - int irsrc = 0, icsrc = 0, lld = numroc_(&nlocal, &mb, &myprow, &irsrc, &nprow); - descinit_(pv->para_2d().desc, &nlocal, &nlocal, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); - descinit_(pv->desc_wfc, &nlocal, &nband, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); + pv->set_desc_wfc_Eij(nlocal, nband, pv->para_2d().get_row_size()); // Initialize data U_operator = new std::complex[nlocal * nlocal]; diff --git a/source/source_lcao/module_rt/test/upsi_test3.cpp b/source/source_lcao/module_rt/test/upsi_test3.cpp index ce6d14c384..f89d124cda 100644 --- a/source/source_lcao/module_rt/test/upsi_test3.cpp +++ b/source/source_lcao/module_rt/test/upsi_test3.cpp @@ -29,16 +29,9 @@ TEST(UpsiTest, testUpsi3) bool print_matrix = false; Parallel_Orbitals* pv; pv = new Parallel_Orbitals(); - pv->para_2d().nloc = nlocal * nband; - pv->para_2d().ncol = nlocal; + ASSERT_EQ(pv->para_2d().set(nlocal, nlocal, 1, ictxt), 0); pv->ncol_bands = nband; - - // Initialize input matrices - int info; - int mb = 1, nb = 1, lda = nband, ldc = nlocal; - int irsrc = 0, icsrc = 0, lld = numroc_(&nlocal, &mb, &myprow, &irsrc, &nprow); - descinit_(pv->para_2d().desc, &nlocal, &nlocal, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); - descinit_(pv->desc_wfc, &nlocal, &nband, &mb, &nb, &irsrc, &icsrc, &ictxt, &lld, &info); + pv->set_desc_wfc_Eij(nlocal, nband, pv->para_2d().get_row_size()); // Initialize data U_operator = new std::complex[nlocal * nlocal]; diff --git a/source/source_lcao/setup_exx.cpp b/source/source_lcao/setup_exx.cpp index ab547da946..dfaadb3cec 100644 --- a/source/source_lcao/setup_exx.cpp +++ b/source/source_lcao/setup_exx.cpp @@ -74,7 +74,7 @@ void Exx_NAO::before_runner( kv, orb, use_gpu_tables); - this->exd->exx_before_all_runners(kv, ucell, pv); + this->exd->exx_before_all_runners(kv, ucell, pv.para_2d()); } else { @@ -84,7 +84,7 @@ void Exx_NAO::before_runner( kv, orb, use_gpu_tables); - this->exc->exx_before_all_runners(kv, ucell, pv); + this->exc->exx_before_all_runners(kv, ucell, pv.para_2d()); } } } @@ -147,7 +147,7 @@ void Exx_NAO::before_scf( *p_chgmix, ucell, orb, - pv, + pv.para_2d(), update_symmetry, use_gpu_tables); } @@ -159,7 +159,7 @@ void Exx_NAO::before_scf( *p_chgmix, ucell, orb, - pv, + pv.para_2d(), update_symmetry, use_gpu_tables); } diff --git a/tools/01_NAO_generation/Generate_Orbital_AllInOne.sh b/tools/01_NAO_generation/Generate_Orbital_AllInOne.sh index 4165739f0c..44e36929d4 100755 --- a/tools/01_NAO_generation/Generate_Orbital_AllInOne.sh +++ b/tools/01_NAO_generation/Generate_Orbital_AllInOne.sh @@ -900,7 +900,7 @@ EOF EOF for ((i_BL = 1; i_BL < ${BL_number[iSTRU]}; i_BL++)); do cat >>INPUT <>INPUT <