Transporter class
Transporter drives the time integration of transport equations.
Contents
Constructors, destructors, conversion operators
- Transporter(CkMigrateMessage* m) explicit
- Migrate constructor: returning from a checkpoint.
Public functions
- auto Transporter() -> Transporter_SDAG_CODE explicit
- Constructor.
- void load(std::size_t meshid, std::size_t nelem)
- Reduction target: the mesh has been read from file on all PEs.
- void partitioned(std::size_t meshid)
- Reduction target: a mesh has been partitioned.
- void partition()
- Reduction target: all Solver (PEs) have computed the number of chares they will recieve contributions from during linear solution.
- void distributed(std::size_t meshid)
- Reduction target: all compute nodes have distrbuted their mesh after partitioning.
- void queriedRef(std::size_t meshid)
- Reduction target: all Refiner chares have queried their boundary edges.
- void respondedRef(std::size_t meshid)
- Reduction target: all Refiner chares have setup their boundary edges.
- void refinserted(std::size_t meshid, std::size_t error)
- Reduction target: all compute nodes have created the mesh refiners.
- void discinserted(std::size_t meshid)
- Reduction target: all Discretization chares have been inserted.
- void disccreated(std::size_t summeshid, std::size_t npoin)
- Reduction target: all Discretization constructors have been called.
- void workinserted(std::size_t meshid)
- Reduction target: all worker (derived discretization) chares have been inserted.
- void compatibility(std::size_t meshid)
- Reduction target: all Refiner chares have received a round of edges, and have run their compatibility algorithm.
- void matched(std::size_t summeshid, std::size_t nextra, std::size_t nref, std::size_t nderef, std::size_t sumrefmode)
- Reduction target: all Refiner chares have matched/corrected the tagging of chare-boundary edges, all chares are ready to perform refinement.
-
void bndint(tk::
real sx, tk:: real sy, tk:: real sz, tk:: real cb, tk:: real summeshid) - Compute surface integral across the whole problem and perform leak-test.
- void refined(std::size_t summeshid, std::size_t nelem, std::size_t npoin)
- Reduction target: all chares have refined their mesh.
- void resized(std::size_t meshid)
- Reduction target: all worker chares have resized their own data after AMR or ALE.
- void startEsup(std::size_t meshid)
- Reduction target: all worker chares have generated their own esup.
- void queried(std::size_t meshid)
- Reduction target: all Sorter chares have queried their boundary edges.
- void responded(std::size_t meshid)
- Reduction target: all Sorter chares have setup their boundary edges.
- void pepartitioned()
- Non-reduction target for receiving progress report on partitioning mesh.
- void pedistributed()
- Non-reduction target for receiving progress report on distributing mesh.
- void chbnd()
- Non-reduction target for receiving progress report on finding bnd nodes.
- void chcomm()
- Non-reduction target for receiving progress report on node ID comm map.
- void chmask()
- Non-reduction target for receiving progress report on node ID mask.
- void chreordered()
- Non-reduction target for receiving progress report on reordering mesh.
- void chcreated()
- Non-reduction target for receiving progress report on creating workers.
- void chbndface()
- Non-reduction target for receiving progress report on finding bnd faces.
- void chcomfac()
- Non-reduction target for receiving progress report on face communication.
- void chghost()
- Non-reduction target for receiving progress report on sending ghost data.
- void chadj()
- Non-reduction target for receiving progress report on face adjacency.
- void doneInsertingGhosts(std::size_t meshid)
- void comfinal(std::size_t initial, std::size_t summeshid)
- Reduction target indicating that the communication maps have been setup.
-
void totalvol(tk::
real v, tk:: real initial, tk:: real summeshid) - Reduction target summing total mesh volume.
-
void minstat(tk::
real d0, tk:: real d1, tk:: real d2, tk:: real rmeshid) - Reduction target yielding the minimum mesh statistics across all workers.
-
void maxstat(tk::
real d0, tk:: real d1, tk:: real d2, tk:: real rmeshid) - Reduction target yielding the maximum mesh statistics across all workers.
-
void sumstat(tk::
real d0, tk:: real d1, tk:: real d2, tk:: real d3, tk:: real d4, tk:: real d5, tk:: real summeshid) - Reduction target yielding the sum of mesh statistics across all workers.
- void pdfstat(CkReductionMsg* msg)
- Reduction target yielding PDF of mesh statistics across all workers.
-
void boxvol(tk::
real* meshdata, int n) - Reduction target computing total volume of IC box.
- void solutionTransferred()
- Reduction target broadcasting to Schemes after mesh transfer.
-
void minDtAcrossMeshes(tk::
real* reducndata, int n) - Reduction target that computes minimum timestep across meshes.
- void diagnostics(CkReductionMsg* msg)
- Reduction target optionally collecting diagnostics, e.g., residuals, from all worker chares.
- void resume()
- Resume execution from checkpoint/restart files.
- void checkpoint(std::size_t finished, std::size_t meshid)
- Save checkpoint/restart files.
- void finish(std::size_t meshid = 0)
- Normal finish of time stepping.
Charm++ pack/unpack serializer member functions
- enum class TimerTag { MESH_READ =0 }
- Timer tags.
- std::vector<std::string> m_input
- List of mesh files to be used for potentially multiple solvers.
- std::vector<int> m_nchare
- Number of worker chares (one per mesh)
- std::unordered_map<std::size_t, std::size_t> m_meshid
- Sum of mesh ids (across all chares, key) for each meshid (value)
- std::vector<std::size_t> m_ncit
- Number of mesh ref corr iter (one per mesh)
- std::size_t m_nload
- Number of meshes loaded.
- std::size_t m_ntrans
- Number of meshes that have transferred solution.
- std::size_t m_ndtmsh
- Number of meshes that have computed their dt.
- std::vector<tk::
real> m_dtmsh - Minimum dt on each mesh (sized at each time step and then emptied)
- std::size_t m_npart
- Number of meshes partitioned.
- std::size_t m_nstat
- Number of mesh statistics computed.
- std::size_t m_ndisc
- Number of Discretization arrays created.
- std::size_t m_nchk
- Number of worker arrays checkpointed.
- std::size_t m_ncom
- Number of worker arrays have finished setting up their comm maps.
- std::vector<std::size_t> m_nt0refit
- Number of t0ref mesh ref iters (one per mesh)
- std::vector<std::size_t> m_ndtrefit
- Number of dtref mesh ref iters (one per mesh)
- std::vector<std::size_t> m_noutrefit
- Number of outref mesh ref iters (one per mesh)
- std::vector<std::size_t> m_noutderefit
- Number of outderef mesh ref iters (one per mesh)
- std::vector<Scheme> m_scheme
- Discretization scheme (one per mesh)
- std::vector<CProxy_Partitioner> m_partitioner
- Partitioner nodegroup proxies (one per mesh)
- std::vector<CProxy_Refiner> m_refiner
- Mesh refiner array proxies (one per mesh)
- std::vector<tk::CProxy_MeshWriter> m_meshwriter
- Mesh writer nodegroup proxies (one per mesh)
- std::vector<CProxy_Sorter> m_sorter
- Mesh sorter array proxy (one per mesh)
- std::vector<std::size_t> m_nelem
- Number of mesh points (per mesh)
- std::vector<std::size_t> m_npoin
- Nonzero if finished with timestepping (one per mesh)
- std::vector<tk::
real> m_meshvol - Total mesh volume (one per mesh)
- std::vector<std::array<tk::
real, 3>> m_minstat - Minimum mesh statistics (one per mesh)
- std::vector<std::array<tk::
real, 3>> m_maxstat - Maximum mesh statistics (one per mesh)
- std::vector<std::array<tk::
real, 3>> m_avgstat - Average mesh statistics (one per mesh)
- std::map<TimerTag, tk::
Timer> m_timer - Timers.
- tk::
Progress<7> m_progMesh - Progress object for preparing mesh.
- tk::
Progress<5> m_progWork - Progress object for preparing workers.
- void pup(PUP::er& p) override
- Pack/Unpack serialize member function.
- void operator|(PUP::er& p, Transporter& t)
- Pack/Unpack serialize operator|.
Function documentation
void inciter::
Reduction target: the mesh has been read from file on all PEs.
| Parameters | |
|---|---|
| meshid in | Mesh id (summed accross all compute nodes) |
| nelem in | Number of mesh elements per mesh (summed across all compute nodes) |
void inciter::
Reduction target: a mesh has been partitioned.
| Parameters | |
|---|---|
| meshid in | Mesh id |
void inciter::
Reduction target: all compute nodes have distrbuted their mesh after partitioning.
| Parameters | |
|---|---|
| meshid in | Mesh id |
void inciter::
Reduction target: all Refiner chares have queried their boundary edges.
| Parameters | |
|---|---|
| meshid in | Mesh id |
void inciter::
Reduction target: all Refiner chares have setup their boundary edges.
| Parameters | |
|---|---|
| meshid in | Mesh id |
void inciter::
Reduction target: all compute nodes have created the mesh refiners.
| Parameters | |
|---|---|
| meshid in | Mesh id (aggregated across all compute nodes with operator max) |
| error in | Error code (aggregated across all compute nodes with operator max) |
void inciter::
Reduction target: all Discretization chares have been inserted.
| Parameters | |
|---|---|
| meshid in | Mesh id |
void inciter::
Reduction target: all Discretization constructors have been called.
| Parameters | |
|---|---|
| summeshid in | Mesh id (summed accross all chares) |
| npoin in | Total number of mesh points (summed across all chares) Note that as opposed to npoin in refined(), this npoin is not multi-counted, and thus should be correct in parallel. |
void inciter::
Reduction target: all worker (derived discretization) chares have been inserted.
| Parameters | |
|---|---|
| meshid in | Mesh id |
void inciter::
Reduction target: all Refiner chares have received a round of edges, and have run their compatibility algorithm.
| Parameters | |
|---|---|
| meshid in | Mesh id (aggregated across all chares using operator max) |
This is called iteratively, until convergence by Refiner. At this point all Refiner chares have received a round of edge data (tags whether an edge needs to be refined, etc.), and applied the compatibility algorithm independent of other Refiner chares. We keep going until the mesh is no longer modified by the compatibility algorithm, based on a new round of edge data communication started in Refiner::
void inciter::
Reduction target: all Refiner chares have matched/corrected the tagging of chare-boundary edges, all chares are ready to perform refinement.
| Parameters | |
|---|---|
| summeshid in | Mesh id (summed across all chares) |
| nextra in | Sum (across all chares) of the number of edges on each chare that need correction along chare boundaries |
| nref in | Sum of number of refined tetrahedra across all chares. |
| nderef in | Sum of number of derefined tetrahedra across all chares. |
| sumrefmode in | Sum of contributions from all chares, encoding refinement mode of operation. |
void inciter::
Compute surface integral across the whole problem and perform leak-test.
| Parameters | |
|---|---|
| sx in | X component of vector summed |
| sy in | Y component of vector summed |
| sz in | Z component of vector summed |
| cb in | Invoke callback if positive |
| summeshid in | Mesh id (summed accross all chares) |
This function aggregates partial surface integrals across the boundary faces of the whole problem. After this global sum a non-zero vector result indicates a leak, e.g., a hole in the boundary, which indicates an error in the boundary face data structures used to compute the partial surface integrals.
void inciter::
Reduction target: all chares have refined their mesh.
| Parameters | |
|---|---|
| summeshid in | Mesh id (summed accross all Refiner chares) |
| nelem in | Total number of elements in mesh summed across the distributed mesh |
| npoin in | Total number of mesh points summed across the distributed mesh. Note that in parallel this is larger than the number of points in the mesh, because the boundary nodes are multi-counted. But we only need an equal or larger than npoin for Sorter:: |
void inciter::
Reduction target: all worker chares have generated their own esup.
| Parameters | |
|---|---|
| meshid in | Mesh id |
void inciter::
| Parameters | |
|---|---|
| meshid in | Mesh id |
void inciter::
Reduction target indicating that the communication maps have been setup.
| Parameters | |
|---|---|
| initial in | Sum of contributions from all chares. If larger than zero, we are during time stepping and if zero we are during setup. |
| summeshid in | Mesh id (summed accross the distributed mesh) |
void inciter::
Reduction target summing total mesh volume.
| Parameters | |
|---|---|
| v in | Mesh volume summed across the distributed mesh |
| initial in | Sum of contributions from all chares. If larger than zero, we are during setup, if zero, during time stepping. |
| summeshid in | Mesh id (summed accross the distributed mesh) |
void inciter::
Reduction target yielding the minimum mesh statistics across all workers.
| Parameters | |
|---|---|
| d0 in | Minimum mesh statistics collected over all chares |
| d1 in | Minimum mesh statistics collected over all chares |
| d2 in | Minimum mesh statistics collected over all chares |
| rmeshid in | Mesh id as a real |
void inciter::
Reduction target yielding the maximum mesh statistics across all workers.
| Parameters | |
|---|---|
| d0 in | Maximum mesh statistics collected over all chares |
| d1 in | Maximum mesh statistics collected over all chares |
| d2 in | Maximum mesh statistics collected over all chares |
| rmeshid in | Mesh id as a real |
void inciter::
Reduction target yielding the sum of mesh statistics across all workers.
| Parameters | |
|---|---|
| d0 in | Sum mesh statistics collected over all chares |
| d1 in | Sum mesh statistics collected over all chares |
| d2 in | Sum mesh statistics collected over all chares |
| d3 in | Sum mesh statistics collected over all chares |
| d4 in | Sum mesh statistics collected over all chares |
| d5 in | Sum mesh statistics collected over all chares |
| summeshid in | Mesh id (summed accross the distributed mesh) |
void inciter::
Reduction target yielding PDF of mesh statistics across all workers.
| Parameters | |
|---|---|
| msg in | Serialized PDF |
void inciter::
Reduction target computing total volume of IC box.
| Parameters | |
|---|---|
| meshdata in | Vector containing volumes of all IC mesh blocks, volume of IC box, and mesh id as a real summed across the distributed mesh |
| n in | Size of vector, automatically computed by Charm |
void inciter::
Reduction target that computes minimum timestep across meshes.
| Parameters | |
|---|---|
| reducndata in | Vector containing minimum values of dt and mesh-moved flags, collected across all meshes |
| n in | Size of vector, automatically computed by Charm |
void inciter::
Reduction target optionally collecting diagnostics, e.g., residuals, from all worker chares.
| Parameters | |
|---|---|
| msg in | Serialized diagnostics vector aggregated across all PEs |
void inciter::
Resume execution from checkpoint/restart files.
This is invoked by Charm++ after the checkpoint is done, as well as when the restart (returning from a checkpoint) is complete
void inciter::
Save checkpoint/restart files.
| Parameters | |
|---|---|
| finished in | Nonzero if finished with time stepping |
| meshid in | Mesh id |
void inciter::
Normal finish of time stepping.
| Parameters | |
|---|---|
| meshid in | Mesh id |
void inciter::
Pack/Unpack serialize member function.
| Parameters | |
|---|---|
| p in/out | Charm++'s PUP::er serializer object reference |
void inciter::
Pack/Unpack serialize operator|.
| Parameters | |
|---|---|
| p in/out | Charm++'s PUP::er serializer object reference |
| t in/out | Transporter object reference |
Variable documentation
std::vector<CProxy_Sorter> inciter::
Mesh sorter array proxy (one per mesh)
Number of mesh elements (per mesh)