The syntax for creating a HYPRE Numeric Solver object follows the basic syntax used for creating every plugin based object. In addition, the HYPRE solver supports several optional attributes that can be used to configure the solver and preconditioner types that will be used, together with tolerances and several other solver specific options.

Example 1

NumericalSolver 
{
  id           = 'solver',
  typeName     = 'HypreSolver',
  description  = 'Iterative matrix solver',
 
  solverMethod = 'pcg',
  precond      = 'boomeramg',
  -- Default options for boomeramg
  precondOptions = {strongThreshold = 0.7, relaxationType = "sym-l1-gauss-seidel", coarsenType = "pmis", 
                      interpType = "extended-i-gpu", aggressiveCoarsenLevels = 2, coarsenPaths = 2}
}

Example 2

NumericalSolver 
{
  id           = 'solver',
  typeName     = 'HypreSolver',
  description  = 'Iterative matrix solver',
 
  solverMethod = 'ilu',
  -- Default options for ilu
  solverOptions = {type = "ilut", dropThreshold = 0.06, triSolver = "direct", fillRate = 5, fillLevel = 0}
}

Different from other GeMA Solver plugins, HYPRE treats solvers and preconditioners as equal objects. In this fashion, all solvers can work as preconditioners, with especial cases for ILU and BoomerAMG that when used as main solvers do not support preconditioning. For best results (stability and speed) we recommend using the PCG as the main solver and BoomerAMG as preconditioner and a small number of "maxIter" (i.e. 200) for large systems. Other methods can easily break/fail solving systems.

IMPORTANT: For now BoomerAMG is not recommend for solving 'FULL' coupling / strong coupled. Matrix entries are handled individually instead of per node, which may cause such cases to completely fail.

Fields

Field	Description	Type	Required	Def. Value
id	The numeric solver name.	String	Yes
typeName	The plugin name. Should be equal to "HypreSolver".	String	Yes
description	A description of the numeric solver purpose.	String	No
Solver and preconditioner definition options:
solverMethod	Defines the type of linear solver to be used. Allowed values are: `"pcg", "boomeramg", "bicgstab", "ilu", "gmres", "flexgmres", "lgmres", "cogmres" and "hybrid"`.	String	No	`"pcg"`
solverOptions	A table with solver options. Used by `"ilu"` and `"boomeramg"` solver methods. (ILU / BoomerAMG options table)	Table	No	`nil`
precond	Defines the type of preconditioner to be used. Allowed values are: `"none", "pcg", "boomeramg", "bicgstab", "ilu", "gmres", "flexgmres", "lgmres", "cogmres" and "hybrid"`.	String	No	`"none"`
precondOptions	A table with preconditioner options. Used by `"ilu"` and `"boomeramg"` preconditioners.	Table	No	`nil`
General options:
maxIter	Defines the maximum number of iterations for the solver.	Integer	No	100
tol	The convergence tolerance.	Number	No	1.0e-5
printIterations	A boolean flag that when set to `true` prints the number of iterations.	Boolean	No	false
printResidual	A boolean flag that when set to `true` prints the residual.	Boolean	No	false
printConsoleMessages	A boolean flag that when set to `true` enables the printing the internal HYPRE messages. Note: Those messages appear only in the console, not in the GeMA log.	Boolean	No	false

Field	Description	Type	Required	Def. Value
ILU options:
type	The ILU type. Available options are: `"iluk", "ilut", "gmres-iluk", "gmres-ilut", "nsh-iluk", "nsh-ilut", "ras-iluk", "ras-ilut", "ddpq-gmres-iluk", "ddpq-gmres-ilut", "gmres-rap-ilu0"`.	String	No	"ilut"
fillLevel	The fill level used by `"iluk"` variants.	Integer	No	0
dropThreshold	The drop threshold of matrix entries used by `"ilut"` variants.	Real	No	1e-2
triSolver	The triangular solver type. Available values are `"iterative", "direct"`.	String	No	"direct"
fillRate	The fill rate used by `"ilut"` variants. Represents the maximum allowed non-zeros per matrix row.	Integer	No	5

Field	Description	Type	Required	Def. Value
BoomerAMG options:
strongThreshold	The strong threshold affecting the coarsening and interpolation. Accpets values in range (0,1).	Real	No	0.9
relaxationType	The smoother used for multigrid relaxation. (Defaults to symetric l1 coarsening). Available types are `"jacobi-1", "jacobi-2", "gauss-seidel", "gauss-seidel-sor-fw", "gauss-seidel-sor-bw", "sym-gauss-seidel-sor", "gauss-elimination", "two-gauss-seidel-1", "two-gauss-seidel-2", "l1-gauss-seidel-fw", "l1-gauss-seidel-bw", "chebyshev", "kaczmarz", "sym-l1-gauss-seidel"`.	String	No	"sym-l1-gauss-seidel"
coarsenType	The coarsening strategy. Available types are `"falgout", "pmis", "hmis", "cgc", "cgc-e"`.	String	No	"pmis"
interpType	The interpolation type used. Available types are `"classic", "ls", "classic-hyp", "direct-gpu", "multipass-1", "multipass-2", "extended-i-gpu", "extended-i", "standard-1", "standard-2", "ff", "ff1", "extended-gpu", "adaptive-gpu", "extended-mat", "extended-i-mat", "extended-e-mat"`.	String	No	"extended-i-gpu"
aggressiveCoarsenLevels	The number of levels to apply aggressive coarsening.	Integer	No	4
coarsenPaths	The number of paths checked during coarsening. (More paths less aggressive)	Integer	No	1