Run, Parallel and Grid Setup
This page explains the controls used before the governing equations are solved. Some settings change the mathematical problem, such as 2D/3D and planar/axisymmetric geometry. Others change only where and how the same problem is executed.
Theory reference. Each option entry includes a link to the corresponding equation, physical model, closure, or numerical method in the Theory Manual. Defaults, availability, simulation effects, and practical guidance are documented directly with the GUI option.
Most parallel settings do not change the intended physical solution. They change decomposition, communication and wall-clock cost. Small roundoff-level differences are normal because parallel reductions and iteration ordering can differ.
Simulation execution
Run controls determine where the same CMPS case is executed.
| GUI option | Meaning / formulation | Effect on the simulation | Practical guidance | Theory reference |
|---|---|---|---|---|
| Simulation NameDefault: NewSimulation_0 | Names the case and associated run context. | Reporting / display No direct effect on equations. It controls case identification and generated output context. | Use a short, unique name that identifies geometry and operating condition. | Parallel/execution process model |
| Host Name | Selects or records the execution host for the solver process. | Cost Execution environment only. | For local runs, keep the local host. For remote execution, use the configured compute host. | Parallel/execution process model |
| Working Directory | Directory from which the case is run and where run-side files are read/written. | Reporting / display No physics effect; incorrect paths can prevent a run or redirect outputs. | Use a directory with sufficient storage and write permission. | Parallel/execution process model |
| Solver Path | Location of the CMPS solver executable selected by the GUI. | Cost No mathematical effect. It determines which installed solver build is launched. | Use the solver build supplied for the intended platform and acceleration mode. | Parallel/execution process model |
| Dimension: 2D / 3DDefault: 3D for a new general case | Sets the spatial dimension of the finite-volume problem. In 2D, the vector state has two velocity components; in 3D it has three. | PhysicsCost Changes equation size, geometric operators and memory/time cost. A genuinely three-dimensional flow cannot be represented by a 2D case. | Use 2D only when geometry, boundary data and expected physics are invariant in the omitted direction. | Finite-volume grid/discretization |
| Run Option: SerialDefault: Selected initially | Runs one solver process. | Cost Same governing equations; avoids inter-process communication but uses one process. | Useful for small cases, debugging case setup and reference comparisons. | Parallel/execution process model |
| Run Option: Single Machine | Runs multiple solver processes on one machine. The mesh is decomposed into partitions. | Cost Same target solution; communication occurs between local processes. | Use when the machine has multiple CPU cores and the problem is large enough to amortize communication. | Parallel/execution process model |
| Run Option: Cluster | Runs partitions across configured compute nodes. | Cost Same target solution; network communication becomes part of the cost. | Use for large meshes. Good partition balance and fast interconnects matter. | Parallel/execution process model |
| Partitioning: K-wayDefault: Selected initially | Creates the requested number of subdomains in a direct partitioning pass. | Cost Changes load balance and number of inter-partition faces, therefore parallel efficiency; not the continuum model. | Good general-purpose choice for most parallel cases. | Domain decomposition and partitioning |
| Partitioning: Recursive | Recursively subdivides the mesh into partitions. | Cost Can produce a different balance/interface pattern than K-way, affecting communication and convergence ordering slightly. | Compare only when K-way produces poor balance or excessive interface area. | Domain decomposition and partitioning |
| Number Of CoresDefault: 2 | Sets the number of solver partitions/processes for parallel execution. | Cost Higher count can reduce wall-clock time until communication and memory bandwidth dominate. It does not increase spatial resolution. | Increase cores after checking that cells per partition remain large enough for efficient work. | Domain decomposition and partitioning |
| Launcher: Auto / ssh / fork / rshDefault: Auto | Selects how parallel processes are started. Auto lets the runtime choose from the configured environment. | Cost Execution only; the numerical formulation is unchanged. | Prefer Auto unless your cluster administrator requires a specific launch mechanism. | Parallel/execution process model |
| Workload Manager: Auto / Slurm / PBS / LoadLeveler / LSF / SGE / CobaltDefault: Auto | Selects the cluster workload-manager interface used when CMPS starts a distributed run. Auto delegates the selection to the configured runtime/environment. | Cost Controls process launch and allocation only; it does not change the governing equations or mesh. A wrong choice normally causes launch/allocation failure rather than a different CFD solution. | Use Auto on a normally configured system. Select the named manager only when the compute cluster is actually managed by it. | Parallel/execution process model |
| Hosts File: Spawn Multiple / ScriptDefault: Script | Chooses how host allocation information is interpreted for cluster launching. | Cost Execution only. A wrong host allocation can cause launch failures or poor placement. | Use the mode recommended by the cluster configuration. | Parallel/execution process model |
| Host entries | A host line assigns a process count to a host, conceptually host : process-count. | Cost Controls process placement and communication topology. | Keep total assigned processes consistent with the requested core count. | Parallel/execution process model |
| Read / Save Hosts File | Loads or saves the host-allocation list. | Reporting / display No equation effect. | Useful for repeatable cluster configurations. | Parallel/execution process model |
Grid coordinate system and checks
These controls determine how CMPS interprets the mesh geometry.
| GUI option | Meaning / formulation | Effect on the simulation | Practical guidance | Theory reference |
|---|---|---|---|---|
| Coordinate System: Planar | Uses the Cartesian/planar finite-volume form without cylindrical geometric source terms. | Physics Correct for ordinary 2D Cartesian geometry. Using it for a body of revolution omits axisymmetric geometric effects. | Choose for planar 2D meshes. | Planar / axisymmetric coordinates |
| Coordinate System: Axis-symmetric | Interprets a 2D meridional mesh as a body of revolution and adds cylindrical metric/source contributions. A representative conservation form is \(\partial_t(rU)+\partial_x(rF)+\partial_r(rG)=rS+S_{axi}\). | PhysicsAccuracy Changes the governing equations and geometric weighting. Cell/face contributions scale with radius. | Use only for rotationally symmetric geometry and boundary conditions. Ensure the axis is correctly identified. | Planar / axisymmetric coordinates |
| Check Grid | Runs grid consistency/quality checks exposed by the GUI. | Stability / convergence Does not alter the solution by itself; it identifies geometry/connectivity problems that can harm accuracy or convergence. | Run after import, transformation or major mesh modification. | Finite-volume grid/discretization |
| List Face Zones | Prints/opens the available boundary face-zone list. | Reporting / display No solution effect. | Use before assigning boundary conditions to ensure all zones are accounted for. | Finite-volume grid/discretization |
| Translate and Correct Axisymmetric Grid | Applies the GUI-supported transformation/correction used to place an axisymmetric grid consistently with the solver axis convention. | PhysicsInitialization Can change mesh coordinates, therefore geometry. | Use before solving when an imported axisymmetric mesh is not positioned on the expected axis. Re-run grid checking afterwards. | Planar / axisymmetric coordinates |
How parallel decomposition affects a simulation
For a fixed mesh and setup, decomposition does not define a different physical model. However, the linear and nonlinear iteration sequence can change slightly because partition interfaces introduce communication and floating-point reductions occur in a different order. These differences should remain small for a well-converged case. If results depend materially on partition count, examine convergence, tolerances, conservation and mesh quality.
Case, mesh and solver-process actions
| GUI action | Meaning | Effect on simulation | Guidance | Theory reference |
|---|---|---|---|---|
| Read Mesh / Read New Mesh | Loads a supported mesh into the case. | Initialization Replaces/defines simulation geometry and zones. | Run Check Grid and verify zone names afterwards. | Finite-volume grid/discretization |
| Read grid-data & case / Write grid-data & case | Loads or saves the complete grid/case configuration in the supported case workflow. | Can replace or preserve the entire case definition. | Save before major setup changes. | Finite-volume grid/discretization |
| Read setup / Write setup | Loads or saves the GUI setup/configuration separately where supported. | Changes/preserves user settings, not mesh resolution. | Use for repeatable option sets. | Finite-volume grid/discretization |
| Read / Write supported polyhedral grid | Imports or exports the supported polyhedral grid representation. | Changes geometry on read; export is file output only. | Check the grid after import/conversion. | Finite-volume grid/discretization |
| Read / Write solution data | Loads a restart/solution state or writes the current state. | Read changes initialization/restart state; write does not change solution. | Use a compatible grid, physics and variable layout. | Restart state |
| Read / Write User Material Database | Loads or saves user-defined material entries. | Changes available property data when read. | Verify units/property ranges before assigning the materials. | Material-property admissibility |
| Import external unstructured grid | Converts/imports an external unstructured mesh through the supported GUI path. | Defines geometry/zones. | Always run grid checks and confirm boundary classifications. | Finite-volume grid/discretization |
| Load / Update User Script | Loads or refreshes a user-supplied function/profile script used by supported boundary/time-dependent inputs. | Physics Can change boundary forcing with time/position. | Validate units, returned ranges and behavior on a small case before production use. | Boundary/time profiles |
| Spawn | Starts the configured solver process(es). | No equation change; begins execution using the current case settings. | Apply/save all desired settings first. | Parallel/execution process model |
| Connect / Disconnect | Attaches/detaches the GUI from an available solver process. | Does not by itself change the running solver equations. | Disconnect only when you understand whether the remote solver will continue running. | Parallel/execution process model |
| Info | Requests solver/process status information. | Reporting only. | Use to verify process state and run configuration. | Parallel/execution process model |
| End | Requests an orderly end of the solver run. | Stops future iterations/time steps. | Prefer orderly ending so current data can be saved cleanly. | Parallel/execution process model |
| Kill | Force-terminates the solver process. | Stops the run immediately; latest unsaved state can be lost. | Use only when orderly stop/end is unavailable. | Parallel/execution process model |