HavoqGT
Todo List
Member havoqgt::detail::hash32 (uint32_t a)

requires documentation!

requires testing!

Class havoqgt::mpi::delegate_partitioned_graph< SegementManager >

Test using simple deterministic patterns.

Add edge_data

Make vertex_iterator a random access iterator

Add invalid bit or state to vertex_locator

Verify low-degree CSR creation: ipp line 167

Boostify controller locator

Member havoqgt::mpi::delegate_partitioned_graph< SegementManager >::create_edge_data (SegManagerOther *segment_manager_o, const char *obj_name) const
remove; make like vertex_data
Member havoqgt::mpi::delegate_partitioned_graph< SegementManager >::create_edge_data (const T &init, SegManagerOther *, const char *obj_name=nullptr) const
remove; make like vertex_data
Member havoqgt::mpi::delegate_partitioned_graph< SegementManager >::delegate_partitioned_graph (const SegmentAllocator< void > &seg_allocator, MPI_Comm mpi_comm, Container &edges, uint64_t max_vertex, uint64_t delegate_degree_threshold, ConstructionState stop_after=GraphReady)

make env var

marke env var

Member havoqgt::mpi::delegate_partitioned_graph< SegementManager >::partition_low_degree (Container &unsorted_edges)
this was tripping, is this old? Could be left over from before when targets was vector based.
Member havoqgt::mpi::mpi_all_to_all (std::vector< std::vector< T > > &in_p_vec, std::vector< std::vector< T > > &out_p_vec, MPI_Comm mpi_comm)
Add ability to optimize for sparse sends. Put MPI_Request objects into a resizing vecor to be skipped for 0 byte send/recvs
Class havoqgt::old_environment

delete this

add environment variable documentation

Member havoqgt::parallel_edge_list_reader::max_vertex_id ()
implement
Member havoqgt::parallel_edge_list_reader::parallel_edge_list_reader (const std::vector< std::string > &filenames)
Add undirected flag