Bind MPI wrapper routines to the corresponding MPI data type

src/almo_scf.F

@@ -81,8 +81,7 @@ MODULE almo_scf
    USE kinds,                           ONLY: default_path_length,&
                                               dp
    USE mathlib,                         ONLY: binomial
-   USE message_passing,                 ONLY: mp_comm_type,&
-                                              mp_sum
+   USE message_passing,                 ONLY: mp_comm_type
    USE molecule_types,                  ONLY: get_molecule_set_info,&
                                               molecule_type
    USE mscfg_types,                     ONLY: get_matrix_from_submatrices,&
@@ -2036,7 +2035,7 @@ SUBROUTINE nlmo_compactification(qs_env, almo_scf_env, matrix)
                                 nfullrows_total=Nrows, &
                                 nfullcols_total=Ncols)
             CALL para_group%set_handle(para_group_handle)
-            CALL mp_sum(retained(ispin), para_group)
+            CALL para_group%sum(retained(ispin))
 
             !devide by the total no. elements
             occ(ispin) = retained(ispin)/Nrows/Ncols

src/almo_scf_optimizer.F

@@ -93,8 +93,7 @@ MODULE almo_scf_optimizer
    USE kinds,                           ONLY: dp
    USE machine,                         ONLY: m_flush,&
                                               m_walltime
-   USE message_passing,                 ONLY: mp_comm_type,&
-                                              mp_sum
+   USE message_passing,                 ONLY: mp_comm_type
    USE particle_methods,                ONLY: get_particle_set
    USE particle_types,                  ONLY: particle_type
    USE qs_energy_types,                 ONLY: qs_energy_type
@@ -1342,7 +1341,7 @@ SUBROUTINE almo_scf_xalmo_pcg(qs_env, almo_scf_env, optimizer, quench_t, &
 
                         reim_diag = 0.0_dp
                         CALL dbcsr_get_diag(tempOccOcc(ispin), reim_diag)
-                        CALL mp_sum(reim_diag, para_group)
+                        CALL para_group%sum(reim_diag)
                         z2(:) = z2(:) + reim_diag(:)*reim_diag(:)
 
                      END DO
@@ -2433,7 +2432,7 @@ SUBROUTINE almo_scf_construct_nlmos(qs_env, optimizer, &
                                    retain_sparsity=.TRUE.)
                ALLOCATE (diagonal(nocc(ispin)))
                CALL dbcsr_get_diag(m_sig_sqrti_ii(ispin), diagonal)
-               CALL mp_sum(diagonal, para_group)
+               CALL para_group%sum(diagonal)
                ! TODO: works for zero diagonal elements?
                diagonal(:) = 1.0_dp/SQRT(diagonal(:))
                CALL dbcsr_set(m_sig_sqrti_ii(ispin), 0.0_dp)
@@ -6527,7 +6526,7 @@ SUBROUTINE compute_obj_nlmos(localization_obj_function_ispin, penalty_func_ispin
 
             reim_diag = 0.0_dp
             CALL dbcsr_get_diag(tempOccOcc2, reim_diag)
-            CALL mp_sum(reim_diag, para_group)
+            CALL para_group%sum(reim_diag)
             z2(:) = z2(:) + reim_diag(:)*reim_diag(:)
 
          END DO
@@ -6660,7 +6659,7 @@ SUBROUTINE compute_gradient_nlmos(m_grad_out, m_B0, weights, &
             CALL dbcsr_get_diag(m_temp_oo_4, tg_diagonal)
             CALL dbcsr_set(m_temp_oo_4, 0.0_dp)
             CALL dbcsr_set_diag(m_temp_oo_4, tg_diagonal)
-            !CALL mp_sum(tg_diagonal, para_group)
+            !CALL para_group%sum(tg_diagonal)
             z2(:) = z2(:) + tg_diagonal(:)*tg_diagonal(:)
 
             CALL dbcsr_multiply("N", "N", 1.0_dp, &

src/almo_scf_qs.F

@@ -50,8 +50,7 @@ MODULE almo_scf_qs
                                               do_bondparm_covalent,&
                                               do_bondparm_vdw
    USE kinds,                           ONLY: dp
-   USE message_passing,                 ONLY: mp_allgather,&
-                                              mp_comm_type
+   USE message_passing,                 ONLY: mp_comm_type
    USE molecule_types,                  ONLY: get_molecule_set_info,&
                                               molecule_type
    USE particle_types,                  ONLY: particle_type
@@ -980,7 +979,7 @@ SUBROUTINE almo_scf_construct_quencher(qs_env, almo_scf_env)
 
       ! third, communicate local length to the other nodes
       ALLOCATE (list_length_cpu(nNodes), list_offset_cpu(nNodes))
-      CALL mp_allgather(2*local_list_length, list_length_cpu, group)
+      CALL group%allgather(2*local_list_length, list_length_cpu)
 
       ! fourth, create a global list
       list_offset_cpu(1) = 0
@@ -992,8 +991,8 @@ SUBROUTINE almo_scf_construct_quencher(qs_env, almo_scf_env)
 
       ! fifth, communicate all list data
       ALLOCATE (global_list(global_list_length))
-      CALL mp_allgather(local_list, global_list, &
-                        list_length_cpu, list_offset_cpu, group)
+      CALL group%allgatherv(local_list, global_list, &
+                            list_length_cpu, list_offset_cpu)
       DEALLOCATE (list_length_cpu, list_offset_cpu)
       DEALLOCATE (local_list)
 
@@ -1396,7 +1395,7 @@ SUBROUTINE almo_scf_construct_quencher(qs_env, almo_scf_env)
 
       ! first, communicate map sizes on the other nodes
       ALLOCATE (domain_entries_cpu(nNodes), offset_for_cpu(nNodes))
-      CALL mp_allgather(2*domain_map_local_entries, domain_entries_cpu, group)
+      CALL group%allgather(2*domain_map_local_entries, domain_entries_cpu)
 
       ! second, create
       offset_for_cpu(1) = 0
@@ -1413,8 +1412,8 @@ SUBROUTINE almo_scf_construct_quencher(qs_env, almo_scf_env)
          domain_map_local(2*(ientry - 1) + 1) = almo_scf_env%domain_map(ispin)%pairs(ientry, 1)
          domain_map_local(2*ientry) = almo_scf_env%domain_map(ispin)%pairs(ientry, 2)
       END DO
-      CALL mp_allgather(domain_map_local, domain_map_global, &
-                        domain_entries_cpu, offset_for_cpu, group)
+      CALL group%allgatherv(domain_map_local, domain_map_global, &
+                            domain_entries_cpu, offset_for_cpu)
       DEALLOCATE (domain_entries_cpu, offset_for_cpu)
       DEALLOCATE (domain_map_local)
 

src/arnoldi/arnoldi_api.F

@@ -41,8 +41,7 @@ MODULE arnoldi_api
                                               create_replicated_row_vec_from_matrix,&
                                               dbcsr_matrix_colvec_multiply
    USE kinds,                           ONLY: dp
-   USE message_passing,                 ONLY: mp_comm_type,&
-                                              mp_sum
+   USE message_passing,                 ONLY: mp_comm_type
 #include "../base/base_uses.f90"
 
    IMPLICIT NONE
@@ -369,7 +368,7 @@ SUBROUTINE arnoldi_conjugate_gradient(matrix_a, vec_x, matrix_p, converged, thre
       END DO
       CALL dbcsr_iterator_stop(dbcsr_iter)
       control => get_control(my_arnoldi)
-      CALL mp_sum(vec_x, control%mp_group)
+      CALL control%mp_group%sum(vec_x)
       ! Deallocated the work vectors
       CALL dbcsr_release(x)
       CALL dbcsr_release(vectors%input_vec)
@@ -526,7 +525,7 @@ FUNCTION vec_dot_vec(avec, bvec, mpgrp) RESULT(adotb)
          END IF
       END DO
       CALL dbcsr_iterator_stop(dbcsr_iter)
-      CALL mp_sum(adotb, mpgrp)
+      CALL mpgrp%sum(adotb)
 
    END FUNCTION vec_dot_vec
 

src/arnoldi/arnoldi_methods.F

@@ -28,8 +28,7 @@ MODULE arnoldi_methods
    USE dbcsr_vector, ONLY: dbcsr_matrix_colvec_multiply
    USE kinds, ONLY: real_4, &
                     real_8
-   USE message_passing, ONLY: mp_bcast, &
-                              mp_sum, mp_comm_type
+   USE message_passing, ONLY: mp_comm_type
 #include "../base/base_uses.f90"
 
    IMPLICIT NONE
@@ -417,7 +416,7 @@ SUBROUTINE arnoldi_init_${nametype1}$ (matrix, vectors, arnoldi_data)
 
          ! Put the projection into the Hessenberg matrix, and make the vectors orthonormal
          ar_data%Hessenberg(1, 1) = DOT_PRODUCT(v_vec, w_vec)
-         CALL mp_sum(ar_data%Hessenberg(1, 1), pcol_group)
+         CALL pcol_group%sum(ar_data%Hessenberg(1, 1))
          ar_data%f_vec = w_vec - v_vec*ar_data%Hessenberg(1, 1)
 
          ar_data%local_history(:, 1) = v_vec(:)
@@ -563,7 +562,7 @@ SUBROUTINE build_subspace_${nametype1}$ (matrix, vectors, arnoldi_data)
 
             ! check convergence and inform everybody about it, a bit annoying to talk to everybody because of that
             IF (control%myproc == 0) control%converged = rnorm .LT. REAL(control%threshold, ${type_prec}$)
-            CALL mp_bcast(control%converged, 0, control%mp_group)
+            CALL control%mp_group%bcast(control%converged, 0)
             IF (control%converged) EXIT
 
             ! transfer normalized residdum to history and its norm to the Hessenberg matrix
@@ -603,7 +602,7 @@ SUBROUTINE build_subspace_${nametype1}$ (matrix, vectors, arnoldi_data)
          ar_data%Hessenberg(control%current_step + 1, control%current_step) = norm
 
          ! broadcast the Hessenberg matrix so we don't need to care later on
-         CALL mp_bcast(ar_data%Hessenberg, 0, control%mp_group)
+         CALL control%mp_group%bcast(ar_data%Hessenberg, 0)
 
          DEALLOCATE (v_vec, w_vec, h_vec, s_vec)
 
@@ -677,7 +676,7 @@ SUBROUTINE Gram_Schmidt_ortho_${nametype1}$ (h_vec, f_vec, s_vec, w_vec, nrow_lo
          h_vec = ${nametype_zero}$; f_vec = ${nametype_zero}$; s_vec = ${nametype_zero}$
          IF (local_comp) CALL ${nametype1}$gemv('T', nrow_local, j, ${nametype_one}$, local_history, &
                                                 nrow_local, w_vec, 1, ${nametype_zero}$, h_vec, 1)
-         CALL mp_sum(h_vec(1:j), pcol_group)
+         CALL pcol_group%sum(h_vec(1:j))
 
          IF (local_comp) CALL ${nametype1}$gemv('N', nrow_local, j, ${nametype_one}$, reorth_mat, &
                                                 nrow_local, h_vec, 1, ${nametype_zero}$, f_vec, 1)
@@ -714,7 +713,7 @@ SUBROUTINE DGKS_ortho_${nametype1}$ (h_vec, f_vec, s_vec, nrow_local, j, &
 
          IF (local_comp) CALL ${nametype1}$gemv('T', nrow_local, j, ${nametype_one}$, local_history, &
                                                 nrow_local, f_vec, 1, ${nametype_zero}$, s_vec, 1)
-         CALL mp_sum(s_vec(1:j), pcol_group)
+         CALL pcol_group%sum(s_vec(1:j))
          IF (local_comp) CALL ${nametype1}$gemv('N', nrow_local, j, ${nametype_negone}$, reorth_mat, &
                                                 nrow_local, s_vec, 1, ${nametype_one}$, f_vec, 1)
          h_vec(1:j) = h_vec(1:j) + s_vec(1:j)
@@ -739,7 +738,7 @@ SUBROUTINE compute_norms_${nametype1}$ (vec, norm, rnorm, pcol_group)
 
          ! the norm is computed along the processor column
          norm = DOT_PRODUCT(vec, vec)
-         CALL mp_sum(norm, pcol_group)
+         CALL pcol_group%sum(norm)
          rnorm = SQRT(REAL(norm, ${type_prec}$))
          ${rnorm_to_norm}$
 
@@ -816,7 +815,7 @@ SUBROUTINE gev_arnoldi_init_${nametype1}$ (matrix, matrix_arnoldi, vectors, arno
 
          ar_data%rho_scale = ${nametype_zero}$
          ar_data%rho_scale = DOT_PRODUCT(v_vec, w_vec)
-         CALL mp_sum(ar_data%rho_scale, pcol_group)
+         CALL pcol_group%sum(ar_data%rho_scale)
 
          CALL dbcsr_matrix_colvec_multiply(matrix(2)%matrix, vectors%input_vec, vectors%result_vec, ${nametype_one}$, &
                                            ${nametype_zero}$, vectors%rep_row_vec, vectors%rep_col_vec)
@@ -825,9 +824,9 @@ SUBROUTINE gev_arnoldi_init_${nametype1}$ (matrix, matrix_arnoldi, vectors, arno
 
          denom = ${nametype_zero}$
          denom = DOT_PRODUCT(v_vec, w_vec)
-         CALL mp_sum(denom, pcol_group)
+         CALL pcol_group%sum(denom)
          IF (control%myproc == 0) ar_data%rho_scale = ar_data%rho_scale/denom
-         CALL mp_bcast(ar_data%rho_scale, 0, control%mp_group)
+         CALL control%mp_group%bcast(ar_data%rho_scale, 0)
 
          ! if the maximum ev is requested we need to optimize with -A-rho*B
          CALL dbcsr_copy(matrix_arnoldi(1)%matrix, matrix(1)%matrix)
@@ -879,7 +878,7 @@ SUBROUTINE gev_build_subspace_${nametype1}$ (matrix, vectors, arnoldi_data)
 
          norm = ${nametype_zero}$
          norm = DOT_PRODUCT(ar_data%x_vec, BZmat(:, 1))
-         CALL mp_sum(norm, pcol_group)
+         CALL pcol_group%sum(norm)
          IF (control%local_comp) THEN
             Zmat(:, 1) = ar_data%x_vec/SQRT(norm); BZmat(:, 1) = BZmat(:, 1)/SQRT(norm)
          END IF
@@ -907,10 +906,10 @@ SUBROUTINE gev_build_subspace_${nametype1}$ (matrix, vectors, arnoldi_data)
             CALL transfer_dbcsr_to_local_array_${nametype1}$ (vectors%result_vec, v_vec, nrow_local, control%local_comp)
             norm = ${nametype_zero}$
             norm = DOT_PRODUCT(ar_data%f_vec, v_vec)
-            CALL mp_sum(norm, pcol_group)
+            CALL pcol_group%sum(norm)
 
             IF (control%myproc == 0) control%converged = REAL(norm, ${type_prec}$) .LT. EPSILON(REAL(1.0, ${type_prec}$))
-            CALL mp_bcast(control%converged, 0, control%mp_group)
+            CALL control%mp_group%bcast(control%converged, 0)
             IF (control%converged) EXIT
             IF (j == control%max_iter - 1) EXIT
 
@@ -927,7 +926,7 @@ SUBROUTINE gev_build_subspace_${nametype1}$ (matrix, vectors, arnoldi_data)
             ar_data%Hessenberg(1:control%current_step, 1:control%current_step) = &
                MATMUL(TRANSPOSE(CZmat(:, 1:control%current_step)), Zmat(:, 1:control%current_step))
          END IF
-         CALL mp_sum(ar_data%Hessenberg, control%mp_group)
+         CALL control%mp_group%sum(ar_data%Hessenberg)
 
          ar_data%local_history = Zmat
          ! broadcast the Hessenberg matrix so we don't need to care later on
@@ -1002,9 +1001,9 @@ SUBROUTINE gev_update_data_${nametype1}$ (matrix, matrix_arnoldi, vectors, arnol
             DEALLOCATE (v_vec)
          END IF
          ! and broadcast the real eigenvalue
-         CALL mp_bcast(control%converged, 0, control%mp_group)
+         CALL control%mp_group%bcast(control%converged, 0)
          ind = control%selected_ind(1)
-         CALL mp_bcast(ar_data%rho_scale, 0, control%mp_group)
+         CALL control%mp_group%bcast(ar_data%rho_scale, 0)
 
 ! Again the maximum value request is done on -A therefore the eigenvalue needs the opposite sign
          ar_data%evals(ind) = ar_data%rho_scale

src/bse.F

@@ -24,9 +24,7 @@ MODULE bse
    USE group_dist_types,                ONLY: get_group_dist,&
                                               group_dist_d1_type
    USE kinds,                           ONLY: dp
-   USE message_passing,                 ONLY: mp_alltoall,&
-                                              mp_request_type,&
-                                              mp_sum
+   USE message_passing,                 ONLY: mp_request_type
    USE mp2_types,                       ONLY: integ_mat_buffer_type
    USE parallel_gemm_api,               ONLY: parallel_gemm
    USE rpa_communication,               ONLY: communicate_buffer
@@ -164,8 +162,8 @@ SUBROUTINE compute_BZ(BZ, Z_vectors, B_iaQ_bse_local, B_bar_iaQ_bse_local, &
 
       END DO
 
-      ! we make the mp_sum to sum over all RI basis functions
-      CALL mp_sum(BZ, para_env%group)
+      ! we make the sum over all RI basis functions
+      CALL para_env%group%sum(BZ)
 
    END SUBROUTINE
 
@@ -225,8 +223,8 @@ SUBROUTINE compute_AZ(AZ, Z_vectors, B_iaQ_bse_local, B_bar_ijQ_bse_local, B_abQ
 
       END DO
 
-      ! we make the mp_sum to sum over all RI basis functions
-      CALL mp_sum(AZ, para_env%group)
+      ! we make the sum over all RI basis functions
+      CALL para_env%group%sum(AZ)
 
       ! add (e_a-e_i)*Z_ia
       DO i_occ = 1, homo
@@ -550,7 +548,7 @@ SUBROUTINE allocate_and_fill_local_array(B_local, fm_mat_S, gd_array, &
 
          END DO
 
-         CALL mp_alltoall(num_entries_send, num_entries_rec, 1, para_env%group)
+         CALL para_env%group%alltoall(num_entries_send, num_entries_rec, 1)
 
          ALLOCATE (buffer_rec(0:para_env%num_pe - 1))
          ALLOCATE (buffer_send(0:para_env%num_pe - 1))

src/colvar_methods.F

@@ -81,8 +81,6 @@ MODULE colvar_methods
                                               twopi
    USE mathlib,                         ONLY: vector_product
    USE memory_utilities,                ONLY: reallocate
-   USE message_passing,                 ONLY: mp_sum,&
-                                              mp_sync
    USE mixed_energy_types,              ONLY: mixed_force_type
    USE mixed_environment_utils,         ONLY: get_subsys_map_index
    USE molecule_kind_types,             ONLY: fixd_constraint_type
@@ -2212,8 +2210,8 @@ SUBROUTINE u_colvar(colvar, force_env)
             END DO
 
             ! Handling Parallel execution
-            CALL mp_sync(force_env%para_env%group)
-            CALL mp_sum(glob_natoms, force_env%para_env%group)
+            CALL force_env%para_env%group%sync()
+            CALL force_env%para_env%group%sum(glob_natoms)
 
             ! Transfer forces
             DO iforce_eval = 1, nforce_eval
@@ -2228,7 +2226,7 @@ SUBROUTINE u_colvar(colvar, force_env)
                      END DO
                   END IF
                END IF
-               CALL mp_sum(global_forces(iforce_eval)%forces, force_env%para_env%group)
+               CALL force_env%para_env%group%sum(global_forces(iforce_eval)%forces)
             END DO
 
             wrk_section => colvar%u_param%mixed_energy_section

src/colvar_utils.F

@@ -28,7 +28,6 @@ MODULE colvar_utils
    USE kinds,                           ONLY: dp
    USE mathlib,                         ONLY: invert_matrix
    USE memory_utilities,                ONLY: reallocate
-   USE message_passing,                 ONLY: mp_sum
    USE molecule_kind_list_types,        ONLY: molecule_kind_list_type
    USE molecule_kind_types,             ONLY: colvar_constraint_type,&
                                               fixd_constraint_type,&
@@ -262,9 +261,9 @@ SUBROUTINE eval_colvar(force_env, coords, cvalues, Bmatrix, MassI, Amatrix)
                END IF
             END DO
          END DO MOL
-         CALL mp_sum(n_tot, force_env%para_env%group)
-         CALL mp_sum(cvalues, force_env%para_env%group)
-         IF (PRESENT(Bmatrix)) CALL mp_sum(Bmatrix, force_env%para_env%group)
+         CALL force_env%para_env%group%sum(n_tot)
+         CALL force_env%para_env%group%sum(cvalues)
+         IF (PRESENT(Bmatrix)) CALL force_env%para_env%group%sum(Bmatrix)
       END IF
       offset = n_tot
       ! Intermolecular Colvars

src/common/cp_log_handling.F

@@ -52,7 +52,6 @@ MODULE cp_log_handling
    USE machine,                         ONLY: default_output_unit,&
                                               m_getpid,&
                                               m_hostnm
-   USE message_passing,                 ONLY: mp_comm_free
    USE string_utilities,                ONLY: compress
    USE timings,                         ONLY: print_stack
 #include "../base/base_uses.f90"
@@ -689,7 +688,7 @@ SUBROUTINE my_cp_para_env_release(para_env)
          para_env%ref_count = para_env%ref_count - 1
          IF (para_env%ref_count < 1) THEN
             IF (para_env%owns_group) THEN
-               CALL mp_comm_free(para_env%group)
+               CALL para_env%group%free()
             END IF
             DEALLOCATE (para_env)
          END IF