Fluid Dynamics Library
Public Member Functions | Protected Member Functions | Static Private Member Functions | Private Attributes
fdl::FluidSolver Class Reference
Inheritance diagram for fdl::FluidSolver:
fdl::XmlInterface fdl::Drawable

List of all members.

Public Member Functions

 FluidSolver (fdl::Grid *grid)
 FluidSolver (TiXmlElement &element)
void step (float dt=0)
void project (float dt)
void applyForces (float dt)
void advect (float dt)
void diffuse (float dt, float rate=0.25f)
void diffuseHeat (float dt, float rate)
void applyViscosity (float dt)
void setCGTolerance (float tol)
void setCGMaxIter (unsigned N)
virtual void fromXML (TiXmlElement &element)
virtual TiXmlElement toXML ()
virtual void draw () const
const Vector & getCurlMagnitude () const
const Vector & getDivergence () const
const Vector & getPressure () const
fdl::Grid * getGrid () const

Protected Member Functions

float computeMaxTimeStep () const
void axpy_prod (const Vector &x, Vector &y) const
void solvePreconditioner (const Vector &b, Vector &x)
void constructMatrix (float dx, float dt, float rho=0.25f)
void constructPreconditioner (float rho=0.25f, float tau=0.97f)
float cgSolve (const Vector &b, Vector &x)
float pcgSolve (const Vector &b, Vector &x)

Static Private Member Functions

template<class T >
static void convolve (const std::vector< double > &kernel, const std::vector< T > &signal, std::vector< T > &output)
static std::vector< double > * gaussianHeatKernel (const int size, double diffusion, float dt)

Private Attributes

std::vector< fdl::Vector3 > curl_
std::vector< fdl::Vector3 > vorticityConfinementForce_
Vector curlMagnitude_
fdl::Grid * grid_
Vector divergence_
Vector pressure_
Vector tempP_
Vector tempW_
Vector tempZ_
Vector tempR_
Vector tempQ_
Vector ADiag_
Vector APlusX_
Vector APlusY_
Vector APlusZ_
Vector precond_
float tol_cg
unsigned maxiter_cg
Vector3 gravity_
int gridX_
int gridY_
int gridZ_
int numPoints_
int slice_
int velSlice_
float time_
float dx_

Detailed Description

Definition at line 39 of file fluidsolver.h.

Constructor & Destructor Documentation

fdl::FluidSolver::FluidSolver ( fdl::Grid *  grid)

FluidSolver constructor collect information about the grid in order to allocate all the data for the linear solve(s) and forces.

Parameters:
griddomain to be solved

Definition at line 25 of file fluidsolver.cpp.

Member Function Documentation

void fdl::FluidSolver::advect ( float  dt)

Performs a semi-lagrangian particle back-trace from each cell center.

See: Jos Stam. Stable Fluids. SIGGRAPH, pages 121–128, 1999.

Parameters:
dtdelta time value to step forward

Definition at line 676 of file fluidsolver.cpp.

void fdl::FluidSolver::applyForces ( float  dt)

Computes and applies all body forces to the system. Currently this includes vorticity confinement and bouyancy. More to come...

Parameters:
dtdelta time value to step forward

Definition at line 296 of file fluidsolver.cpp.

void fdl::FluidSolver::applyViscosity ( float  dt)

NOT IMPLEMENTED YET

Parameters:
dtdelta time value to step forward

Definition at line 630 of file fluidsolver.cpp.

void fdl::FluidSolver::axpy_prod ( const Vector &  x,
Vector &  y 
) const [protected]

axpy_prod computes y = Ax. This function provides an alternative to the BLAS function axpy_prod which uses the sparse matrix A represented by vectors only.
See the uBlas axpy_prod method for more info.

Parameters:
xvector to be multiplied by A
yresult of Ax

Definition at line 1046 of file fluidsolver.cpp.

float fdl::FluidSolver::cgSolve ( const Vector &  b,
Vector &  x 
) [protected]

cgSolve performs a conjugate gradient solve for the pressure matrix given two vectors x and b.

Parameters:
bthe vector representing the negative divergence
xthe resulting pressure gradient after the linear solve
Returns:
float value representing the residual after the iterative solve

Definition at line 946 of file fluidsolver.cpp.

float fdl::FluidSolver::computeMaxTimeStep ( ) const [protected]

Uses the formulation of the maximum timestep from Foster and Fedkiw '01.
See: Nick Foster Ronald Fedkiw. Practical Animation of Liquids. SIGGRAPH, pages 23-30, 2001.

Returns:
the max timestep float for deltaT.

Definition at line 206 of file fluidsolver.cpp.

void fdl::FluidSolver::constructMatrix ( float  dx,
float  dt,
float  rho = 0.25f 
) [protected]

Populates vectors that contain the coefficients of the sparse linear system that we intend to solve. The approach is described in Bridson Chapter 4.

Parameters:
dxcell width of the grid
dtdelta time value to step forward
rhothe global scaling factor accounting for density

Definition at line 748 of file fluidsolver.cpp.

void fdl::FluidSolver::constructPreconditioner ( float  rho = 0.25f,
float  tau = 0.97f 
) [protected]

constructPreconditioner currently makes the modified incomplete cholesky preconditioner for a preconditioned conjugate gradient solve of the positive semi-definite pressure matrix.

Parameters:
rhothe global scaling factor accounting for density
tauprecondiitoner "tuning parameter"

Definition at line 887 of file fluidsolver.cpp.

void fdl::FluidSolver::diffuse ( float  dt,
float  rate = 0.25f 
)

Appy diffusion by scaling the density values by an arbitrary constant rate. The calculation used here is d = e^(-dt*rate).

Parameters:
dtdelta time value to step forward
ratethe rate by which we will scale down the density

Definition at line 543 of file fluidsolver.cpp.

void fdl::FluidSolver::fromXML ( TiXmlElement &  element) [virtual]

Initializes the object purely from XML inputs.

Structure should resemble: 

 	<FluidSolver>
 		<Grid x="50" y="50" z="50" dx="0.1">
 		<maxiter_cg>100</maxiter_cg>
 		<tol_cg>0.0001</tol_cg>
 	</FluidSolver>

Implements fdl::XmlInterface.

Definition at line 139 of file fluidsolver.cpp.

float fdl::FluidSolver::pcgSolve ( const Vector &  b,
Vector &  x 
) [protected]

pcgSolve performs a preconditioned conjugate gradient solve for the pressure matrix given two vectors x and b.

Parameters:
bthe vector representing the negative divergence
xthe resulting pressure gradient after the linear solve
Returns:
float value representing the residual after the iterative solve

Definition at line 988 of file fluidsolver.cpp.

void fdl::FluidSolver::project ( float  dt)

Takes the fluid field update and projects it down to a divergence-free field using a linear solve on the new pressure values. Afterwards, the solved pressures will be applied to the grid_->

Parameters:
dtdelta time value to step forward

Definition at line 386 of file fluidsolver.cpp.

void fdl::FluidSolver::setCGMaxIter ( unsigned  N)

Sets the maximum number of iterations for the conjugate gradients solver(s).

Definition at line 198 of file fluidsolver.cpp.

void fdl::FluidSolver::setCGTolerance ( float  tol)

Sets the tolerance for the residual at step n, $|r_n|_2 \leq \mathrm{tol}$, in the Euclidean 2-norm.

Definition at line 193 of file fluidsolver.cpp.

void fdl::FluidSolver::solvePreconditioner ( const Vector &  b,
Vector &  x 
) [protected]

solvePreconditioner

Parameters:
b
x

Definition at line 1103 of file fluidsolver.cpp.

void fdl::FluidSolver::step ( float  dt = 0)

This is the main function for updating the fluid system. Input of zero or less (which is the default) will trigger a call to compute the maximum time step using computeMaxTimeStep.

Parameters:
dtdelta time value to step forward

Definition at line 230 of file fluidsolver.cpp.

TiXmlElement fdl::FluidSolver::toXML ( ) [virtual]

Structures all member data in an XML formatted node.

Implements fdl::XmlInterface.

Definition at line 164 of file fluidsolver.cpp.

The documentation for this class was generated from the following files:
Generated on Fri Jul 29 2011 22:25:03 for Fluid Dynamics Library by  doxygen 1.7.4