int MPIR_Ineighbor_alltoallw_sched_allcomm_linear(const void *sendbuf, const int sendcounts[],
                                                  const MPI_Aint sdispls[],
                                                  const MPI_Datatype sendtypes[], void *recvbuf,
                                                  const int recvcounts[], const MPI_Aint rdispls[],
                                                  const MPI_Datatype recvtypes[],
                                                  MPIR_Comm * comm_ptr, MPIR_Sched_t s)
{
    int mpi_errno = MPI_SUCCESS;
    int indegree, outdegree, weighted;
    int k, l;
    int *srcs, *dsts;
    MPIR_CHKLMEM_DECL(2);

    mpi_errno = MPIR_Topo_canon_nhb_count(comm_ptr, &indegree, &outdegree, &weighted);
    if (mpi_errno)
        MPIR_ERR_POP(mpi_errno);
    MPIR_CHKLMEM_MALLOC(srcs, int *, indegree * sizeof(int), mpi_errno, "srcs", MPL_MEM_COMM);
    MPIR_CHKLMEM_MALLOC(dsts, int *, outdegree * sizeof(int), mpi_errno, "dsts", MPL_MEM_COMM);
    mpi_errno = MPIR_Topo_canon_nhb(comm_ptr,
                                    indegree, srcs, MPI_UNWEIGHTED,
                                    outdegree, dsts, MPI_UNWEIGHTED);
    if (mpi_errno)
        MPIR_ERR_POP(mpi_errno);

    for (k = 0; k < outdegree; ++k) {
        char *sb;
        MPIR_Ensure_Aint_fits_in_pointer(MPIR_VOID_PTR_CAST_TO_MPI_AINT sendbuf + sdispls[k]);

        sb = ((char *) sendbuf) + sdispls[k];
        mpi_errno = MPIR_Sched_send(sb, sendcounts[k], sendtypes[k], dsts[k], comm_ptr, s);
        if (mpi_errno)
            MPIR_ERR_POP(mpi_errno);
    }

    for (l = 0; l < indegree; ++l) {
        char *rb;
        MPIR_Ensure_Aint_fits_in_pointer(MPIR_VOID_PTR_CAST_TO_MPI_AINT recvbuf + rdispls[l]);

        rb = ((char *) recvbuf) + rdispls[l];
        mpi_errno = MPIR_Sched_recv(rb, recvcounts[l], recvtypes[l], srcs[l], comm_ptr, s);
        if (mpi_errno)
            MPIR_ERR_POP(mpi_errno);
    }

    MPIR_SCHED_BARRIER(s);

  fn_exit:
    MPIR_CHKLMEM_FREEALL();
    return mpi_errno;
  fn_fail:
    goto fn_exit;
}
int MPIR_Ineighbor_allgatherv_sched_allcomm_linear(const void *sendbuf, int sendcount,
                                                   MPI_Datatype sendtype, void *recvbuf,
                                                   const int recvcounts[], const int displs[],
                                                   MPI_Datatype recvtype, MPIR_Comm * comm_ptr,
                                                   MPIR_Sched_t s)
{
    int mpi_errno = MPI_SUCCESS;
    int indegree, outdegree, weighted;
    int k, l;
    int *srcs, *dsts;
    MPI_Aint recvtype_extent;
    MPIR_CHKLMEM_DECL(2);

    MPIR_Datatype_get_extent_macro(recvtype, recvtype_extent);

    mpi_errno = MPIR_Topo_canon_nhb_count(comm_ptr, &indegree, &outdegree, &weighted);
    if (mpi_errno)
        MPIR_ERR_POP(mpi_errno);
    MPIR_CHKLMEM_MALLOC(srcs, int *, indegree * sizeof(int), mpi_errno, "srcs", MPL_MEM_COMM);
    MPIR_CHKLMEM_MALLOC(dsts, int *, outdegree * sizeof(int), mpi_errno, "dsts", MPL_MEM_COMM);
    mpi_errno = MPIR_Topo_canon_nhb(comm_ptr,
                                    indegree, srcs, MPI_UNWEIGHTED,
                                    outdegree, dsts, MPI_UNWEIGHTED);
    if (mpi_errno)
        MPIR_ERR_POP(mpi_errno);

    for (k = 0; k < outdegree; ++k) {
        mpi_errno = MPIR_Sched_send(sendbuf, sendcount, sendtype, dsts[k], comm_ptr, s);
        if (mpi_errno)
            MPIR_ERR_POP(mpi_errno);
    }

    for (l = 0; l < indegree; ++l) {
        char *rb = ((char *) recvbuf) + displs[l] * recvtype_extent;
        mpi_errno = MPIR_Sched_recv(rb, recvcounts[l], recvtype, srcs[l], comm_ptr, s);
        if (mpi_errno)
            MPIR_ERR_POP(mpi_errno);
    }

    MPIR_SCHED_BARRIER(s);

  fn_exit:
    MPIR_CHKLMEM_FREEALL();
    return mpi_errno;
  fn_fail:
    goto fn_exit;
}
int MPIR_Ineighbor_allgatherv_default(const void *sendbuf, int sendcount, MPI_Datatype sendtype, void *recvbuf, const int recvcounts[], const int displs[], MPI_Datatype recvtype, MPID_Comm *comm_ptr, MPID_Sched_t s)
{
    int mpi_errno = MPI_SUCCESS;
    int indegree, outdegree, weighted;
    int i, k,l;
    int *srcs, *dsts;
    int comm_size;
    MPI_Aint recvtype_extent;
    MPIU_CHKLMEM_DECL(2);

    comm_size = comm_ptr->local_size;

    MPID_Datatype_get_extent_macro(recvtype, recvtype_extent);

    for (i = 0; i < comm_size; ++i) {
        MPID_Ensure_Aint_fits_in_pointer(MPI_VOID_PTR_CAST_TO_MPI_AINT recvbuf +
                                         (displs[i] * recvtype_extent));
    }

    mpi_errno = MPIR_Topo_canon_nhb_count(comm_ptr, &indegree, &outdegree, &weighted);
    if (mpi_errno) MPIU_ERR_POP(mpi_errno);
    MPIU_CHKLMEM_MALLOC(srcs, int *, indegree*sizeof(int), mpi_errno, "srcs");
    MPIU_CHKLMEM_MALLOC(dsts, int *, outdegree*sizeof(int), mpi_errno, "dsts");
    mpi_errno = MPIR_Topo_canon_nhb(comm_ptr,
                                    indegree, srcs, MPI_UNWEIGHTED,
                                    outdegree, dsts, MPI_UNWEIGHTED);
    if (mpi_errno) MPIU_ERR_POP(mpi_errno);

    for (k = 0; k < outdegree; ++k) {
        mpi_errno = MPID_Sched_send(sendbuf, sendcount, sendtype, dsts[k], comm_ptr, s);
        if (mpi_errno) MPIU_ERR_POP(mpi_errno);
    }

    for (l = 0; l < indegree; ++l) {
        char *rb = ((char *)recvbuf) + displs[l] * recvtype_extent;
        mpi_errno = MPID_Sched_recv(rb, recvcounts[l], recvtype, srcs[l], comm_ptr, s);
        if (mpi_errno) MPIU_ERR_POP(mpi_errno);
    }

    MPID_SCHED_BARRIER(s);

fn_exit:
    MPIU_CHKLMEM_FREEALL();
    return mpi_errno;
fn_fail:
    goto fn_exit;
}