Vars U_R(Vars Ui, Vars Uipo, Vars Uipt){//does U_L for the right state
Vars UR;
UR.mass = Uipo.mass - 0.5*minmod(thet*(Uipo.mass-Ui.mass),0.5*(Uipt.mass-Ui.mass),thet*(Uipt.mass-Uipo.mass));
UR.xvelocity = Uipo.xvelocity - 0.5*minmod(thet*(Uipo.xvelocity-Ui.xvelocity),.5*(Uipt.xvelocity-Ui.xvelocity),thet*(Uipt.xvelocity-Uipo.xvelocity));
UR.yvelocity = Uipo.yvelocity - 0.5*minmod(thet*(Uipo.yvelocity-Ui.yvelocity),.5*(Uipt.yvelocity-Ui.yvelocity),thet*(Uipt.yvelocity-Uipo.yvelocity));
UR.press = Uipo.press - 0.5*minmod(thet*(Uipo.press-Ui.press), 0.5*(Uipt.press-Ui.press), thet*(Uipt.press-Uipo.press));
UR.energy = Et(UR.mass, UR.xvelocity, UR.yvelocity, UR.press);
return(UR);
}
Vars U_L(Vars Ui, Vars Uimo, Vars Uipo){//interpolates the conserved variables at interface for the left state
Vars UL;
UL.mass = Ui.mass + 0.5*minmod(thet*(Ui.mass - Uimo.mass), 0.5*(Uipo.mass - Uimo.mass),thet*(Uipo.mass-Ui.mass));
UL.xvelocity = Ui.xvelocity + 0.5*minmod(thet*(Ui.xvelocity - Uimo.xvelocity), 0.5*(Uipo.xvelocity - Uimo.xvelocity),thet*(Uipo.xvelocity-Ui.xvelocity));
UL.yvelocity = Ui.yvelocity + 0.5*minmod(thet*(Ui.yvelocity - Uimo.yvelocity), 0.5*(Uipo.yvelocity - Uimo.yvelocity),thet*(Uipo.yvelocity-Ui.yvelocity));
UL.press = Ui.press + 0.5*minmod(thet*(Ui.press - Uimo.press), 0.5*(Uipo.press - Uimo.press),thet*(Uipo.press-Ui.press));
UL.energy = Et(UL.mass, UL.xvelocity, UL.yvelocity, UL.press);
return(UL);
}
void plm( struct domain * theDomain ){
struct cell * theCells = theDomain->theCells;
int Nr = theDomain->Nr;
double PLM = theDomain->theParList.PLM;
int i,q;
for( i=0 ; i<Nr ; ++i ){
int im = i-1;
int ip = i+1;
if( i==0 ) im = 0;
if( i==Nr-1 ) ip = Nr-1;
struct cell * c = theCells+i;
struct cell * cL = theCells+im;
struct cell * cR = theCells+ip;
double drL = cL->dr;
double drC = c->dr;
double drR = cR->dr;
for( q=0 ; q<NUM_Q ; ++q ){
double pL = cL->prim[q];
double pC = c->prim[q];
double pR = cR->prim[q];
double sL = pC - pL;
sL /= .5*( drC + drL );
double sR = pR - pC;
sR /= .5*( drR + drC );
double sC = pR - pL;
sC /= .5*( drL + drR ) + drC;
c->grad[q] = minmod( PLM*sL , sC , PLM*sR );
}
}
}
void Project(CELL * cell)
{
REAL minmod(REAL, REAL, REAL);
REAL minmod2(REAL, REAL, REAL, REAL, REAL);
UINT i, j, k;
REAL u[NVAR], ux[NVAR], uxb[NVAR], dul[NVAR], dur[NVAR], R[NVAR][NVAR],
Ri[NVAR][NVAR], fact;
fact = sqrt(3.0);
for(i = 1; i < NC - 1; i++) {
for(j = 0; j < NVAR; j++) {
dul[j] = cell[i].Un[j][0] - cell[i - 1].Un[j][0];
dur[j] = cell[i + 1].Un[j][0] - cell[i].Un[j][0];
u[j] = cell[i].Un[j][0];
ux[j] = fact * cell[i].Un[j][1];
}
EigMat(u, R, Ri);
Multi(Ri, ux);
Multi(Ri, dul);
Multi(Ri, dur);
for(j = 0; j < NVAR; j++) {
if(fabs(ux[j]) <= Mfact * cell[i].h * cell[i].h)
uxb[j] = ux[j];
else
uxb[j] = minmod(ux[j], dul[j], dur[j]);
}
Multi(R, uxb);
for(j = 0; j < NVAR; j++) {
uxb[j] = uxb[j] / fact;
if(cell[i].Un[j][1] != uxb[j]) {
cell[i].Un[j][1] = uxb[j];
for(k = 2; k < cell[i].p; k++)
cell[i].Un[j][k] = 0.0;
}
}
}
}
inline __device__ float derivative(float& left, float& center, float& right) {
return minmod(KPSIMULATOR_MINMOD_THETA*(center-left),
0.5f*(right-left),
KPSIMULATOR_MINMOD_THETA*(right-center));
}
void evolve(double *G, double *h, double *u, double g, double dx, double dt, int nBC, int n,double *nG, double *nh, double theta)
{
//modifies nh and nG to give the new values of h and G after a single time step
double idx = 1.0 / dx;
double ithree = 1.0 / 3.0;
int i = nBC - 1;
//calculate gradients
double dhib = (h[i] - h[i-1]);
double dhim = 0.5*(h[i+1] - h[i-1]);
double dhif = (h[i+1] - h[i]);
//double duib = (u[i] - u[i-1]);
//double dui = 0.5*(u[i+1] - u[i-1]);
//double duif = (u[i+1] - u[i]);
double dGib = (G[i] - G[i-1]);
double dGim = 0.5*(G[i+1] - G[i-1]);
double dGif = (G[i+1] - G[i]);
//calculate values at right of i cell
double dhi = minmod(theta*dhib, dhim, theta*dhif);
//double dui = minmod(theta*duib, duim, theta*duif);
double dGi = minmod(theta*dGib, dGim, theta*dGif);
double ue = 0.5*(u[i+1]+u[i]);
double hir = h[i] + 0.5*dhi;
double Gir = G[i] + 0.5*dGi;
double uir = ue;
//calculate values at left of i+1 cell
//calculate gradients
double dhip1b = (h[i+1] - h[i]);
double dhip1m = 0.5*(h[i+2] - h[i]);
double dhip1f = (h[i+2] - h[i+1]);
//double duip1b = (u[i+1] - u[i]);
//double duip1m = 0.5*(u[i+2] - u[i]);
//double duip1f = (u[i+2] - u[i+1]);
double dGip1b = (G[i+1] - G[i]);
double dGip1m = 0.5*(G[i+2] - G[i]);
double dGip1f = (G[i+2] - G[i+1]);
double dhip1 = minmod(theta*dhip1b, dhip1m, theta*dhip1f);
//double duip1 = minmod(theta*duip1b, duip1m, theta*duip1f);
double dGip1 = minmod(theta*dGip1b, dGip1m, theta*dGip1f);
double hip1l = h[i+1] - 0.5*dhip1;
double Gip1l = G[i+1] - 0.5*dGip1;
double uip1l = ue;
//right force
double duer = idx*(u[i+1] - u[i]);
double sqrtghel = sqrt(g*hir);
double sqrtgher = sqrt(g*hip1l);
double sl = fmin(0,fmin(uir - sqrtghel, uip1l - sqrtgher));
double sr = fmax(0,fmax(uir + sqrtghel, uip1l + sqrtgher));
double felh = uir*hir;
double felG = Gir*uir + 0.5*g*hir*hir - 2*ithree*hir*hir*hir*duer*duer;
double ferh = uip1l*hip1l;
double ferG = Gip1l*uip1l + 0.5*g*hip1l*hip1l -2*ithree*hip1l*hip1l*hip1l*duer*duer;
double isrmsl = 0.0;
if(sr != sl) isrmsl = 1.0 / (sr - sl);
double foh = isrmsl*(sr*felh - sl*ferh + sl*sr*(hip1l - hir));
double foG = isrmsl*(sr*felG - sl*ferG + sl*sr*(Gip1l - Gir));
double fih = foh;
double fiG = foG;
dhi = dhip1;
//dui = duip1;
dGi = dGip1;
for (i = nBC ; i < n +nBC;i++)
{
ue = 0.5*(u[i+1]+u[i]);
//i right
hir = h[i] + 0.5*dhi;
Gir = G[i] + 0.5*dGi;
uir = ue;
//calculate gradients
dhip1b = (h[i+1] - h[i]);
dhip1m = 0.5*(h[i+2] - h[i]);
dhip1f = (h[i+2] - h[i+1]);
//duip1b = (u[i+1] - u[i]);
//duip1m = 0.5*(u[i+2] - u[i]);
//duip1f = (u[i+2] - u[i+1]);
dGip1b = (G[i+1] - G[i]);
dGip1m = 0.5*(G[i+2] - G[i]);
dGip1f = (G[i+2] - G[i+1]);
dhip1 = minmod(theta*dhip1b, dhip1m, theta*dhip1f);
//duip1 = minmod(theta*duip1b, duip1m, theta*duip1f);
dGip1 = minmod(theta*dGip1b, dGip1m, theta*dGip1f);
hip1l = h[i+1] - 0.5*dhip1;
Gip1l = G[i+1] - 0.5*dGip1;
uip1l = ue;
duer = idx*(u[i+1] - u[i]);
sqrtghel = sqrt(g*hir);
sqrtgher = sqrt(g*hip1l);
sl = fmin(0,fmin(uir - sqrtghel, uip1l - sqrtgher));
sr = fmax(0,fmax(uir + sqrtghel, uip1l + sqrtgher));
felh = uir*hir;
felG = Gir*uir + 0.5*g*hir*hir - 2*ithree*hir*hir*hir*duer*duer;
ferh = uip1l*hip1l;
ferG = Gip1l*uip1l + 0.5*g*hip1l*hip1l -2*ithree*hip1l*hip1l*hip1l*duer*duer;
isrmsl = 0.0;
if(sr != sl) isrmsl = 1.0 / (sr - sl);
foh = isrmsl*(sr*felh - sl*ferh + sl*sr*(hip1l - hir));
foG = isrmsl*(sr*felG - sl*ferG + sl*sr*(Gip1l - Gir));
//source term
nh[i -nBC] = h[i] -dt*idx*(foh - fih);
nG[i -nBC] = G[i] -dt*idx*(foG -fiG);
fih = foh;
fiG = foG;
dhi = dhip1;
//dui = duip1;
dGi = dGip1;
}
}
// moved here from apps/plasma/lib/2d/ApplyLimiter_divfree.cpp
// (which I copied and modified from lib/2d/cart; I changed
// the coefficients to limiter more aggressively.)
//
void ApplyLimiterKrivodonova_modified(dTensorBC4& aux, dTensorBC4& q,
void (*ProjectRightEig)(int,const dTensor1&,
const dTensor1&,const dTensor2&,
dTensor2&),
void (*ProjectLeftEig)(int,const dTensor1&,
const dTensor1&,const dTensor2&,
dTensor2&))
{
// -------------------------------------------------------------
// Limiter based on a modification of the following paper:
// L. Krivodonova. "Limiters for high-order discontinuous
// Galerkin methods." J. Comp. Phys., Vol. 226, pg 879-896.
// -------------------------------------------------------------
const int mx = q.getsize(1);
const int my = q.getsize(2);
const int meqn = q.getsize(3);
const int kmax = q.getsize(4);
const int mbc = q.getmbc();
const int maux = aux.getsize(2);
const int method1 = int((sqrt(1+8*kmax)-1)/2);
const int ksize = (method1==2) ? 1 : 3;
dTensorBC4 wx_cent(mx,my,meqn,ksize,mbc);
dTensorBC4 wy_cent(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_right(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_left(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_up(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_down(mx,my,meqn,ksize,mbc);
void ConvertQtoW(int ixy, const dTensorBC4& aux, const dTensorBC4& qold,
dTensorBC4& dwp, dTensorBC4& dwm, dTensorBC4& w_cent,
void (*ProjectLeftEig)(int,const dTensor1&,
const dTensor1&,const dTensor2&,
dTensor2&));
void ConvertWtoQ(int ixy, const dTensorBC4& aux, const dTensorBC4& qold,
const dTensorBC4& win, dTensorBC4& qout,
void (*ProjectRightEig)(int,const dTensor1&,
const dTensor1&,const dTensor2&,
dTensor2&));
// ----------------------------------------------------------
// Key: storage of characteristic variables
//
// wx_cent(i,j,me,1) = Lx * q(i,j,me,2)
// wx_cent(i,j,me,2) = Lx * q(i,j,me,4)
// wx_cent(i,j,me,3) = Lx * q(i,j,me,5)
//
// dw_right(i,j,me,1) = Lx * (q(i+1,j,me,1)-q(i,j,me,1))
// dw_right(i,j,me,2) = Lx * (q(i+1,j,me,2)-q(i,j,me,2))
// dw_right(i,j,me,3) = Lx * (q(i,j+1,me,2)-q(i,j,me,2))
//
// dw_left(i,j,me,1) = Lx * (q(i,j,me,1)-q(i-1,j,me,1))
// dw_left(i,j,me,2) = Lx * (q(i,j,me,2)-q(i-1,j,me,2))
// dw_left(i,j,me,3) = Lx * (q(i,j,me,2)-q(i,j-1,me,2))
//
// wy_cent(i,j,me,1) = Ly * q(i,j,me,3)
// wy_cent(i,j,me,2) = Ly * q(i,j,me,4)
// wy_cent(i,j,me,3) = Ly * q(i,j,me,6)
//
// dw_up(i,j,me,1) = Ly * (q(i,j+1,me,1)-q(i,j,me,1))
// dw_up(i,j,me,2) = Ly * (q(i+1,j,me,3)-q(i,j,me,3))
// dw_up(i,j,me,3) = Ly * (q(i,j+1,me,3)-q(i,j,me,3))
//
// dw_down(i,j,me,1) = Ly * (q(i,j,me,1)-q(i,j-1,me,1))
// dw_down(i,j,me,2) = Ly * (q(i,j,me,3)-q(i-1,j,me,3))
// dw_down(i,j,me,3) = Ly * (q(i,j,me,3)-q(i,j-1,me,3))
//
// ----------------------------------------------------------
// apply Krivodonova limiter to suppress oscillations
//
// Moment limiter (highest to lowest)
switch ( method1 )
{
default: unsupported_value_error(method1);
case 2: // 2nd order in space
// Convert to characteristic variables in x-direction
ConvertQtoW(1,aux,q,dw_right,dw_left,wx_cent,ProjectLeftEig);
// Convert to characteristic variables in y-direction
ConvertQtoW(2,aux,q,dw_up,dw_down,wy_cent,ProjectLeftEig);
// Limit in both the x-direction and the y-direction
#pragma omp parallel for
for (int i=(3-mbc); i<=(mx+mbc-2); i++)
for (int j=(3-mbc); j<=(my+mbc-2); j++)
for (int m=1; m<=meqn; m++)
{
// x-direction: limit linear term
{
const double dwp = dw_right.get(i,j,m,1);
const double dwm = dw_left.get(i,j,m,1);
const double wx_now = wx_cent.get(i,j,m,1);
const double wx_limited = minmod(wx_now, dwp/sq3, dwm/sq3);
wx_cent.set(i,j,m,1, wx_limited );
}
// y-direction: limit linear term
{
const double dwp = dw_up.get(i,j,m,1);
const double dwm = dw_down.get(i,j,m,1);
const double wy_now = wy_cent.get(i,j,m,1);
const double wy_limited = minmod(wy_now, dwp/sq3, dwm/sq3);
wy_cent.set(i,j,m,1, wy_limited );
}
}
// Convert back to conserved variables in x-direction
ConvertWtoQ(1,aux,q,wx_cent,q,ProjectRightEig);
// Convert back to conserved variables in y-direction
ConvertWtoQ(2,aux,q,wy_cent,q,ProjectRightEig);
break;
case 3: // 3rd order in space
// Convert to characteristic variables in x-direction
ConvertQtoW(1,aux,q,dw_right,dw_left,wx_cent,ProjectLeftEig);
// Convert to characteristic variables in y-direction
ConvertQtoW(2,aux,q,dw_up,dw_down,wy_cent,ProjectLeftEig);
// Limit in the x-direction
#pragma omp parallel for
for (int i=(3-mbc); i<=(mx+mbc-2); i++)
for (int j=(3-mbc); j<=(my+mbc-2); j++)
for (int m=1; m<=meqn; m++)
{
bool limited_x2=false;
// x-direction: limit quadratic term
{
const double dwp = dw_right.get(i,j,m,2);
const double dwm = dw_left.get(i,j,m,2);
const double wx_now = wx_cent.get(i,j,m,3);
// const double wx_limited = minmod(wx_now, 0.5*(sq3/sq5)*dwp,
// 0.5*(sq3/sq5)*dwm);
const double wx_limited = minmod(wx_now, (0.5/sq3)*dwp,
(0.5/sq3)*dwm);
if(wx_limited!=wx_now)
{
limited_x2=true;
wx_cent.set(i,j,m,3, wx_limited );
}
}
//if(limited_x2)
// x-direction: limit mixed term
{
const double dwp = dw_right.get(i,j,m,3);
const double dwm = dw_left.get(i,j,m,3);
const double wx_now = wx_cent.get(i,j,m,2);
const double wx_limited = minmod(wx_now, dwp/sq3, dwm/sq3);
if(wx_limited!=wx_now)
{
limited_x2=true;
wx_cent.set(i,j,m,2, wx_limited );
}
}
if(limited_x2)
// x-direction: limit linear term
{
const double dwp = dw_right.get(i,j,m,1);
const double dwm = dw_left.get(i,j,m,1);
const double wx_now = wx_cent.get(i,j,m,1);
const double wx_limited = minmod(wx_now, dwp/sq3, dwm/sq3);
if(wx_limited!=wx_now)
{
wx_cent.set(i,j,m,1, wx_limited );
}
}
bool limited_y2=false;
// y-direction: limit quadratic term
{
const double dwp = dw_up.get(i,j,m,3);
const double dwm = dw_down.get(i,j,m,3);
const double wy_now = wy_cent.get(i,j,m,3);
// const double wy_limited = minmod(wy_now, 0.5*(sq3/sq5)*dwp,
// 0.5*(sq3/sq5)*dwm);
const double wy_limited = minmod(wy_now, (0.5/sq3)*dwp,
(0.5/sq3)*dwm);
if(wy_limited!=wy_now)
{
limited_y2=true;
wy_cent.set(i,j,m,3, wy_limited );
}
}
//if(limited_y2)
// y-direction: limit mixed term
{
const double dwp = dw_up.get(i,j,m,2);
const double dwm = dw_down.get(i,j,m,2);
const double wy_now = wy_cent.get(i,j,m,2);
const double wy_limited = minmod(wy_now, dwp/sq3, dwm/sq3);
if(wy_limited!=wy_now)
{
limited_y2=true;
wy_cent.set(i,j,m,2, wy_limited);
}
}
if(limited_y2)
// y-direction: limit linear term
{
const double dwp = dw_up.get(i,j,m,1);
const double dwm = dw_down.get(i,j,m,1);
const double wy_now = wy_cent.get(i,j,m,1);
const double wy_limited = minmod(wy_now, dwp/sq3, dwm/sq3);
if(wy_limited!=wy_now)
{
wy_cent.set(i,j,m,1, wy_limited );
}
}
}
// Convert back to conserved variables in x-direction
ConvertWtoQ(1,aux,q,wx_cent,q,ProjectRightEig);
// Convert back to conserved variables in y-direction
ConvertWtoQ(2,aux,q,wy_cent,q,ProjectRightEig);
break;
}
}
/*
Computes the undivided differences with limiter.
*/
static double slopefit(double left, double center, double right, double theta)
{
return minmod(theta*(right-center),
minmod(0.5*(right-left), theta*(center-left)) );
}
void ApplyLimiter(dTensorBC4& aux, dTensorBC4& q,
void (*ProjectRightEig)(int,const dTensor1&,
const dTensor1&,const dTensor2&,
dTensor2&),
void (*ProjectLeftEig)(int,const dTensor1&,
const dTensor1&,const dTensor2&,
dTensor2&))
{
// -------------------------------------------------------------
// Limiter based on a modification of the following paper:
// L. Krivodonova. "Limiters for high-order discontinuous
// Galerkin methods." J. Comp. Phys., Vol. 226, pg 879-896.
// -------------------------------------------------------------
const int mx = q.getsize(1);
const int my = q.getsize(2);
const int meqn = q.getsize(3);
const int kmax = q.getsize(4);
const int mbc = q.getmbc();
const int maux = aux.getsize(2);
const int method1 = int((sqrt(1+8*kmax)-1)/2);
//const int ksize = (method1==2)?1:3;
int ksize;
if (method1==2)
{ ksize = 1; }
else
{ ksize = 3; }
dTensorBC4 wx_cent(mx,my,meqn,ksize,mbc);
dTensorBC4 wy_cent(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_right(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_left(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_up(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_down(mx,my,meqn,ksize,mbc);
void ConvertQtoW(int ixy, const dTensorBC4& aux, const dTensorBC4& qold,
dTensorBC4& dwp, dTensorBC4& dwm, dTensorBC4& w_cent,
void (*ProjectLeftEig)(int,const dTensor1&,
const dTensor1&,const dTensor2&,
dTensor2&));
void ConvertWtoQ(int ixy, const dTensorBC4& aux, const dTensorBC4& qold,
const dTensorBC4& win, dTensorBC4& qout,
void (*ProjectRightEig)(int,const dTensor1&,
const dTensor1&,const dTensor2&,
dTensor2&));
// ----------------------------------------------------------
// Key: storage of characteristic variables
//
// wx_cent(i,j,me,1) = Lx * q(i,j,me,2)
// wx_cent(i,j,me,2) = Lx * q(i,j,me,4)
// wx_cent(i,j,me,3) = Lx * q(i,j,me,5)
//
// dw_right(i,j,me,1) = Lx * (q(i+1,j,me,1)-q(i,j,me,1))
// dw_right(i,j,me,2) = Lx * (q(i+1,j,me,2)-q(i,j,me,2))
// dw_right(i,j,me,3) = Lx * (q(i,j+1,me,2)-q(i,j,me,2))
//
// dw_left(i,j,me,1) = Lx * (q(i,j,me,1)-q(i-1,j,me,1))
// dw_left(i,j,me,2) = Lx * (q(i,j,me,2)-q(i-1,j,me,2))
// dw_left(i,j,me,3) = Lx * (q(i,j,me,2)-q(i,j-1,me,2))
//
// wy_cent(i,j,me,1) = Ly * q(i,j,me,3)
// wy_cent(i,j,me,2) = Ly * q(i,j,me,4)
// wy_cent(i,j,me,3) = Ly * q(i,j,me,6)
//
// dw_up(i,j,me,1) = Ly * (q(i,j+1,me,1)-q(i,j,me,1))
// dw_up(i,j,me,2) = Ly * (q(i+1,j,me,3)-q(i,j,me,3))
// dw_up(i,j,me,3) = Ly * (q(i,j+1,me,3)-q(i,j,me,3))
//
// dw_down(i,j,me,1) = Ly * (q(i,j,me,1)-q(i,j-1,me,1))
// dw_down(i,j,me,2) = Ly * (q(i,j,me,3)-q(i-1,j,me,3))
// dw_down(i,j,me,3) = Ly * (q(i,j,me,3)-q(i,j-1,me,3))
//
// ----------------------------------------------------------
// Moment limiter (highest to lowest)
switch ( method1 )
{
case 2: // 2nd order in space
// Convert to characteristic variables in x-direction
ConvertQtoW(1,aux,q,dw_right,dw_left,wx_cent,ProjectLeftEig);
// Convert to characteristic variables in y-direction
ConvertQtoW(2,aux,q,dw_up,dw_down,wy_cent,ProjectLeftEig);
// Limit in both the x-direction and the y-direction
#pragma omp parallel for
for (int i=(3-mbc); i<=(mx+mbc-2); i++)
for (int j=(3-mbc); j<=(my+mbc-2); j++)
for (int m=1; m<=meqn; m++)
{
// x-direction: limit linear term
double dwp = dw_right.get(i,j,m,1);
double dwm = dw_left.get(i,j,m,1);
double wx_now = wx_cent.get(i,j,m,1);
double wx_limited = minmod(wx_now, dwp/sq3, dwm/sq3);
wx_cent.set(i,j,m,1, wx_limited );
// y-direction: limit linear term
dwp = dw_up.get(i,j,m,1);
dwm = dw_down.get(i,j,m,1);
double wy_now = wy_cent.get(i,j,m,1);
double wy_limited = minmod(wy_now, dwp/sq3, dwm/sq3);
wy_cent.set(i,j,m,1, wy_limited );
}
// Convert back to conserved variables in x-direction
ConvertWtoQ(1,aux,q,wx_cent,q,ProjectRightEig);
// Convert back to conserved variables in y-direction
ConvertWtoQ(2,aux,q,wy_cent,q,ProjectRightEig);
break;
case 3: // 3rd order in space
// for this limiter there is actually no reason to
// transform the higher-order terms. This should
// be changed to accelerate the code.
//
// Convert to characteristic variables in x-direction
ConvertQtoW(1,aux,q,dw_right,dw_left,wx_cent,ProjectLeftEig);
//
// Convert to characteristic variables in y-direction
ConvertQtoW(2,aux,q,dw_up,dw_down,wy_cent,ProjectLeftEig);
#undef report_stats
// so that we don't waste computation time if we won't report
#ifdef report_stats
#define increment(a) a++
#else
#define increment(a) (void)0
#endif
const double dx=dogParamsCart2.get_dx();
const double dy=dogParamsCart2.get_dy();
// this should be made a configurable parameter
const double M=0.; //.01; //M=.000050;
const double Mdx2=M*dx*dx;
const double Mdy2=M*dy*dy;
#ifdef report_stats
// the accumulation of these variables is not parallelized
int nx_steep = 0;
int nxL1 = 0;
int nxL_mxd = 0;
int nxL_quad = 0;
int nxL_1and2 = 0;
int n = 0;
int ny_steep = 0;
int nyL1 = 0;
int nyL_mxd = 0;
int nyL_quad = 0;
int nyL_1and2 = 0;
#endif
#pragma omp parallel for
for (int i=(3-mbc); i<=(mx+mbc-2); i++)
for (int j=(3-mbc); j<=(my+mbc-2); j++)
for (int m=1; m<=meqn; m++)
{
increment(n);
// limit terms in x direction
bool limited_linear = false;
if(1)
{
// x-direction: limit linear term
bool limited_linear_x=false;
if(1)
{
double wx_now = wx_cent.get(i,j,m,1);
if(fabs(wx_now) > Mdx2) // so the peaks aren't clipped
{
increment(nx_steep);
double dwp = dw_right.get(i,j,m,1);
double dwm = dw_left.get(i,j,m,1);
double wx_limited = minmod(wx_now, dwp/sq3, dwm/sq3);
if(wx_now!=wx_limited)
{
increment(nxL1);
limited_linear_x=true;
limited_linear=true;
wx_cent.set(i,j,m,1, wx_limited );
// zero out higher-order terms
//{
// wx_cent.set(i,j,m,2, 0. );
// wx_cent.set(i,j,m,3, 0. );
//}
}
}
}
}
// limit terms in y direction
if(1)
{
// y-direction: limit linear term
bool limited_linear_y = false;
if(1)
{
double wy_now = wy_cent.get(i,j,m,1);
if(fabs(wy_now) > Mdy2) // so the peaks aren't clipped
{
increment(ny_steep);
double dwp = dw_up.get(i,j,m,1);
double dwm = dw_down.get(i,j,m,1);
double wy_limited = minmod(wy_now, dwp/sq3, dwm/sq3);
if(wy_now!=wy_limited)
{
increment(nyL1);
limited_linear_y = true;
limited_linear = true;
wy_cent.set(i,j,m,1, wy_limited );
// zero out higher-order terms
//{
// wy_cent.set(i,j,m,2, 0. );
// wy_cent.set(i,j,m,3, 0. );
//}
}
}
}
if(limited_linear)
{
wx_cent.set(i,j,m,2, 0. );
wx_cent.set(i,j,m,3, 0. );
wy_cent.set(i,j,m,2, 0. );
wy_cent.set(i,j,m,3, 0. );
}
}
}
#ifdef report_stats
if(nx_steep > 0)
{
printf("n=%4d, nx_steep=%4d, nxL1=%4d"
"\n",
n, nx_steep, nxL1);
}
if(ny_steep > 0)
{
printf("n=%4d, ny_steep=%4d, nyL1=%4d"
"\n",
n, ny_steep, nyL1);
}
#endif
// Convert back to conserved variables in x-direction
ConvertWtoQ(1,aux,q,wx_cent,q,ProjectRightEig);
// Convert back to conserved variables in y-direction
ConvertWtoQ(2,aux,q,wy_cent,q,ProjectRightEig);
break;
}
}
void evolve(double *h,double *G, double *hblank,double *Gblank,double *bed,double g,double *u0,double *u1,double *h0,double *h1,double *b0,double *b1,double theta, int n, int nBC,double dx,double dt)
{
//get averages
double idx = 1.0 / dx;
double *u = malloc(n*sizeof(double));
int i,j;
int nBCn = 3;
double th,hx,bx,bxx,D,ai,bi,ci,gb1,gb2,gb3,ge1,ge2,ge3;
//calculate u at time step
//getufromG(h,G,bed, u, u0[nBC - 1], u1[0], h0[nBC - 1], h1[0], b0[nBC - 1], b1[0], n,dx);
// u seems to be the problem here
getufromG(h,G,bed,u0[nBC - 1], u1[0], h0[nBC - 1], h1[0],b0[nBC - 1], b1[0],dx,n,u);
//boundaries
//beginning
//i=-1
i = -1;
j = nBC-1;
th = h0[j];
hx = 0.5*idx*(h[i+1] - h0[j-1]);
bx = 0.5*idx*(bed[i+1] - b0[j-1]);
bxx =idx*idx*(bed[i+1] -2*b0[j] + b0[j-1]);
D = th + th*hx*bx + 0.5*th*th*bxx + th*bx*bx;
ai = th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
bi = D + 2.0*i3*th*th*th*(idx*idx);
ci = -1.0*th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
gb3 = ai*u0[j-1] + bi*u0[j] +ci*u[i+1];
//printf("uim1 : %f | ui : %f | uip1 : %f\n",u0[j-1],u0[j],u[i+1]);
//printf("uip2 : %f | uip3 : %f | uip4 : %f\n",u[i+2],u[i+3], u[i+4]);
//i=-2
i = -2;
j = nBC -2;
th = h0[j];
hx = 0.5*idx*(h0[j+1] - h0[j-1]);
bx = 0.5*idx*(b0[j+1] - b0[j-1]);
bxx = idx*idx*(b0[j+1] - 2*b0[j] + b0[j+1]);
D = th + th*hx*bx + 0.5*th*th*bxx + th*bx*bx;
ai = th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
bi = D + 2.0*i3*th*th*th*(idx*idx);
ci = -1.0*th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
gb2 = ai*u0[j-1] + bi*u0[j] +ci*u0[j+1];
//i=-3
i = -3;
j = nBC -3;
th = h0[j];
hx = 0.5*idx*(h0[j+1] - h0[j-1]);
bx = 0.5*idx*(b0[j+1] - b0[j-1]);
bxx = idx*idx*(b0[j+1] - 2*b0[j] + b0[j+1]);
D = th + th*hx*bx + 0.5*th*th*bxx + th*bx*bx;
ai = th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
bi = D + 2.0*i3*th*th*th*(idx*idx);
ci = -1.0*th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
gb1 = ai*u0[j-1] + bi*u0[j] +ci*u0[j+1];
//i = n
i = n;
j = 0;
th = h1[j];
hx = 0.5*idx*(h1[j+1] - h[i-1]);
bx = 0.5*idx*(b1[j+1] - bed[i-1]);
bxx = idx*idx*(b1[j+1] - 2*b1[j] + bed[i-1]);
D = th + th*hx*bx + 0.5*th*th*bxx + th*bx*bx;
ai = th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
bi = D + 2.0*i3*th*th*th*(idx*idx);
ci = -1.0*th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
ge1 = ai*u[i-1] + bi*u1[j] + ci*u1[j+1];
//i = n+1
j = 1;
th = h1[j];
hx = 0.5*idx*(h1[j+1] - h1[j-1]);
bx = 0.5*idx*(b1[j+1] - b1[j-1]);
bxx = idx*idx*(b1[j+1] - 2*b1[j] + b1[j-1]);
D = th + th*hx*bx + 0.5*th*th*bxx + th*bx*bx;
ai = th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
bi = D + 2.0*i3*th*th*th*(idx*idx);
ci = -1.0*th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
ge2 = ai*u1[j-1] + bi*u1[j] +ci*u1[j+1];
//i = n+2
j = 2;
th = h1[j];
hx = 0.5*idx*(h1[j+1] - h1[j-1]);
bx = 0.5*idx*(b1[j+1] - b1[j-1]);
bxx = idx*idx*(b1[j+1] - 2*b1[j] + b1[j-1]);
D = th + th*hx*bx + 0.5*th*th*bxx + th*bx*bx;
ai = th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
bi = D + 2.0*i3*th*th*th*(idx*idx);
ci = -1.0*th*th*hx*(0.5*idx) - i3*th*th*th*(idx*idx);
ge3 = ai*u1[j-1] + bi*u1[j] +ci*u1[j+1];
double *ubeg = malloc(nBCn*sizeof(double));
double *uend = malloc(nBCn*sizeof(double));
double *bbeg = malloc(nBCn*sizeof(double));
double *bend = malloc(nBCn*sizeof(double));
double *hbeg = malloc(nBCn*sizeof(double));
double *hend = malloc(nBCn*sizeof(double));
double *gbeg = malloc(nBCn*sizeof(double));
double *gend = malloc(nBCn*sizeof(double));
ubeg[0] = u0[1];
ubeg[1] = u0[2];
ubeg[2] = u0[3];
hbeg[0] = h0[1];
hbeg[1] = h0[2];
hbeg[2] = h0[3];
bbeg[0] = b0[1];
bbeg[1] = b0[2];
bbeg[2] = b0[3];
uend[0] = u1[0];
uend[1] = u1[1];
uend[2] = u1[2];
hend[0] = h1[0];
hend[1] = h1[1];
hend[2] = h1[2];
bend[0] = b1[0];
bend[1] = b1[1];
bend[2] = b1[2];
gbeg[0] = gb1;
gbeg[1] = gb2;
gbeg[2] = gb3;
gend[0] = ge1;
gend[1] = ge2;
gend[2] = ge3;
//printf("gb1 : %f | gb2 : %f | gb3 : %f \n",gb1,gb2,gb3);
//printf("ge1 : %f | ge2 : %f | ge3 : %f \n",ge1,ge2,ge3);
double *Gbc = malloc((n + 2*nBCn)*sizeof(double));
double *hbc = malloc((n + 2*nBCn)*sizeof(double));
double *ubc = malloc((n + 2*nBCn)*sizeof(double));
double *bedbc = malloc((n + 2*nBCn)*sizeof(double));
conc(gbeg,G,gend,nBCn,n,nBCn,Gbc);
conc(hbeg,h,hend,nBCn,n,nBCn,hbc);
conc(bbeg,bed,bend,nBCn,n,nBCn,bedbc);
conc(ubeg,u,uend,nBCn,n,nBCn,ubc);
//do normal stuff
double wi,wip1,wip2,wip3,wim1,wim2,dwib,dwif,dwim,dhib,dhif,dhim,dGib,dGif,dGim;
double duib,duif,duim, dwi,dhi, dGi, dui;
double hir,wir,Gir,uir,bir,hil,wil,Gil,uil,bil;
//i = 2
i = nBCn -1;
//define the stage
wi = hbc[i] + bedbc[i];
wip1 = hbc[i+1] + bedbc[i+1];
wip2 = hbc[i+2] + bedbc[i+2];
wip3 = hbc[i+3] + bedbc[i+3];
wim1 = hbc[i-1] + bedbc[i-1];
wim2 = hbc[i-2] + bedbc[i-2];
//reconstruct common values first
//i left and right
//gradients
dwib = (wi - wim1);
dwif = (wip1 - wi);
dwim = 0.5*(wip1 - wim1);
dhib = (hbc[i] - hbc[i-1]);
dhif = (hbc[i+1] - hbc[i]);
dhim = 0.5*(hbc[i+1] - hbc[i-1]);
dGib = (Gbc[i] - Gbc[i-1]);
dGif = (Gbc[i+1] - Gbc[i]);
dGim = 0.5*(Gbc[i+1] - Gbc[i-1]);
duib = (ubc[i] - ubc[i-1]);
duif = (ubc[i+1] - ubc[i]);
duim = 0.5*(ubc[i+1] - ubc[i-1]);
//limiting
dwi = minmod(theta*dwib,theta*dwif,dwim);
dhi = minmod(theta*dhib,theta*dhif,dhim);
dGi = minmod(theta*dGib,theta*dGif,dGim);
dui = minmod(theta*duib,theta*duif,duim);
//reconstruct right
hir = hbc[i]+ 0.5*dhi;
wir = wi + 0.5*dwi;
Gir = Gbc[i] + 0.5*dGi;
uir = ubc[i] + 0.5*dui;
bir = wir - hir;
//reconstruct left
hil = hbc[i] - 0.5*dhi;
wil = wi - 0.5*dwi;
Gil = Gbc[i] - 0.5*dGi;
uil = ubc[i] - 0.5*dui;
bil = wil - hil;
//only left of i+1 common but do both
double dwip1b,dwip1f,dwip1m,dhip1b,dhip1f,dhip1m,dGip1b,dGip1f,dGip1m;
double duip1b,duip1f,duip1m, dwip1,dhip1, dGip1, duip1;
double hip1r,wip1r,Gip1r,uip1r,bip1r,hip1l,wip1l,Gip1l,uip1l,bip1l;
//gradients
dwip1b = (wip1 - wi);
dwip1f = (wip2 - wip1);
dwip1m = 0.5*(wip2 - wi);
dhip1b = (hbc[i+1] - hbc[i]);
dhip1f = (hbc[i+2] - hbc[i+1]);
dhip1m = 0.5*(hbc[i+2] - hbc[i]);
dGip1b = (Gbc[i+1] - Gbc[i]);
dGip1f = (Gbc[i+2] - Gbc[i+1]);
dGip1m = 0.5*(Gbc[i+2] - Gbc[i]);
duip1b = (ubc[i+1] - ubc[i]);
duip1f = (ubc[i+2] - ubc[i+1]);
duip1m = 0.5*(ubc[i+2] - ubc[i]);
//limiting
dwip1 = minmod(theta*dwip1b,theta*dwip1f,dwip1m);
dhip1 = minmod(theta*dhip1b,theta*dhip1f,dhip1m);
dGip1 = minmod(theta*dGip1b,theta*dGip1f,dGip1m);
duip1 = minmod(theta*duip1b,theta*duip1f,duip1m);
//reconstruct right
hip1r = hbc[i+1] + 0.5*dhip1;
wip1r = wip1 + 0.5*dwip1;
Gip1r = Gbc[i+1] + 0.5*dGip1;
uip1r = ubc[i+1] + 0.5*duip1;
bip1r = wip1r - hip1r;
//reconstruct left
hip1l = hbc[i+1] - 0.5*dhip1;
wip1l = wip1 - 0.5*dwip1;
Gip1l = Gbc[i+1] - 0.5*dGip1;
uip1l = ubc[i+1] - 0.5*duip1;
bip1l = wip1l - hip1l;
//only right of i-1
//i-1 right
double dwim1b,dwim1f,dwim1m,dhim1b,dhim1f,dhim1m,dGim1b,dGim1f,dGim1m;
double duim1b,duim1f,duim1m, dwim1,dhim1, dGim1, duim1;
double him1r,wim1r,Gim1r,uim1r,bim1r,him1l,wim1l,Gim1l,uim1l,bim1l;
//gradients
dwim1b = (wim1 - wim2);
dwim1f = (wi - wim1);
dwim1m = 0.5*(wi - wim2);
dhim1b = (hbc[i-1] - hbc[i-2]);
dhim1f = (hbc[i] - hbc[i-1]);
dhim1m = 0.5*(hbc[i] - hbc[i-2]);
dGim1b = (Gbc[i-1] - Gbc[i-2]);
dGim1f = (Gbc[i] - Gbc[i-1]);
dGim1m = 0.5*(Gbc[i] - Gbc[i-2]);
duim1b = (ubc[i-1] - ubc[i-2]);
duim1f = (ubc[i] - ubc[i-1]);
duim1m = 0.5*(ubc[i] - ubc[i-2]);
//limiting
dwim1 = minmod(theta*dwim1b,theta*dwim1f,dwim1m);
dhim1 = minmod(theta*dhim1b,theta*dhim1f,dhim1m);
dGim1 = minmod(theta*dGim1b,theta*dGim1f,dGim1m);
duim1 = minmod(theta*duim1b,theta*duim1f,duim1m);
//reconstruct right
him1r = hbc[i-1] + 0.5*dhim1;
wim1r = wim1 + 0.5*dwim1;
Gim1r = Gbc[i-1] + 0.5*dGim1;
uim1r = ubc[i-1] + 0.5*duim1;
bim1r = wim1r - him1r;
//reconstruct i+2 left
double dwip2b,dwip2f,dwip2m,dhip2b,dhip2f,dhip2m,dGip2b,dGip2f,dGip2m;
double duip2b,duip2f,duip2m,dwip2,dhip2, dGip2, duip2;
double hip2r,wip2r,Gip2r,uip2r,bip2r,hip2l,wip2l,Gip2l,uip2l,bip2l;
//gradients
dwip2b = (wip2 - wip1);
dwip2f = (wip3 - wip2);
dwip2m = 0.5*(wip3 - wip1);
dhip2b = (hbc[i+2] - hbc[i+1]);
dhip2f = (hbc[i+3] - hbc[i+2]);
dhip2m = 0.5*(hbc[i+3] - hbc[i+1]);
dGip2b = (Gbc[i+2] - Gbc[i+1]);
dGip2f = (Gbc[i+3] - Gbc[i+2]);
dGip2m = 0.5*(Gbc[i+3] - Gbc[i+1]);
duip2b = (ubc[i+2] - ubc[i+1]);
duip2f = (ubc[i+3] - ubc[i+2]);
duip2m = 0.5*(ubc[i+3] - ubc[i+1]);
//limiting
dwip2 = minmod(theta*dwip2b,theta*dwip2f,dwip2m);
dhip2 = minmod(theta*dhip2b,theta*dhip2f,dhip2m);
dGip2 = minmod(theta*dGip2b,theta*dGip2f,dGip2m);
duip2 = minmod(theta*duip2b,theta*duip2f,duip2m);
//reconstruct left
hip2l = hbc[i+2] - 0.5*dhip2;
wip2l = wip2 - 0.5*dwip2;
Gip2l = Gbc[i+2] - 0.5*dGip2;
uip2l = ubc[i+2] - 0.5*duip2;
bip2l = wip2l - hip2l;
//calculate forces
double nbi,hihm,hihp,her,Ger,uer,hel,Gel,uel, himhp;
double duer,duel,dber,dbel,sqrtghel,sqrtgher,sl,sr,felh,felG,ferh,ferG,isrmsl;
double foh,foG,fih,fiG;
//right force i
nbi = fmax(bip1l,bir);
hihm = fmax(0,wir-nbi);
hihp = fmax(0,wip1l-nbi);
her = hihp;
Ger = Gip1l;
uer = uip1l;
hel = hihm;
Gel = Gir;
uel = uir;
//do u like o2fix
double ue = 0.5*(ubc[i+1]+ubc[i]);
double due = idx*(ubc[i+1] - ubc[i]);
double dbe = idx*(bedbc[i+1] - bedbc[i]);
uer = ue;
uel = ue;
duer = due;
duel = due;
dber = dbe;
dbel = dbe;
//duer = idx*(uip2l - uip1l);
dber = idx*(bip2l - bip1l);
//duel = idx*(uir - uim1r);
dbel = idx*(bir - bim1r);
sqrtghel = sqrt(g*hel);
sqrtgher = sqrt(g*her);
sl = fmin(0, fmin(uel - sqrtghel, uer - sqrtgher));
sr = fmax(0,fmax(uel + sqrtghel, uer + sqrtgher));
felh = uel*hel;
felG = Gel*uel + 0.5*g*hel*hel - 2*i3*hel*hel*hel*duel*duel + hel*hel*uel*duel*dbel;
ferh = uer*her;
ferG = Ger*uer + 0.5*g*her*her - 2*i3*her*her*her*duer*duer + her*her*uer*duer*dber;
//printf("%d || hihm - hir : %e || hihp - wip1l : %e\n", i, hihm - hir,hihp - wip1l);
isrmsl = 0.0;
if(sr != sl) isrmsl = 1.0 / (sr - sl);
foh = isrmsl*(sr*felh - sl*ferh + sl*sr*(her - hel ));
foG = isrmsl*(sr*felG - sl*ferG + sl*sr*(Ger - Gel ));
fih = foh;
fiG = foG;
himhp = hihp;
him1r = hir;
wim1r = wir;
bim1r = bir;
Gim1r = Gir;
uim1r = uir;
hil = hip1l;
wil = wip1l;
bil = bip1l;
Gil = Gip1l;
uil = uip1l;
hir = hip1r;
wir = wip1r;
bir = bip1r;
Gir = Gip1r;
uir = uip1r;
hip1l = hip2l;
wip1l = wip2l;
bip1l = bip2l;
Gip1l = Gip2l;
uip1l = uip2l;
dhip1 = dhip2;
dwip1 = dwip2;
duip1 = duip2;
dGip1 = dGip2;
for(i = nBCn; i < n + nBCn; i++)
{
//update both forces at same time
//define the stage
wi = hbc[i] + bedbc[i];
wip1 = hbc[i+1] + bedbc[i+1];
wip2 = hbc[i+2] + bedbc[i+2];
wip3 = hbc[i+3] + bedbc[i+3];
wim1 = hbc[i-1] + bedbc[i-1];
wim2 = hbc[i-2] + bedbc[i-2];
//reconstruct common values first
//only left of i+1 common but do both
//reconstruct right
hip1r = hbc[i+1] + 0.5*dhip1;
wip1r = wip1 + 0.5*dwip1;
Gip1r = Gbc[i+1] + 0.5*dGip1;
uip1r = ubc[i+1] + 0.5*duip1;
bip1r = wip1r - hip1r;
//reconstruct i+2 left
//gradients
dwip2b = (wip2 - wip1);
dwip2f = (wip3 - wip2);
dwip2m = 0.5*(wip3 - wip1);
dhip2b = (hbc[i+2] - hbc[i+1]);
dhip2f = (hbc[i+3] - hbc[i+2]);
dhip2m = 0.5*(hbc[i+3] - hbc[i+1]);
dGip2b = (Gbc[i+2] - Gbc[i+1]);
dGip2f = (Gbc[i+3] - Gbc[i+2]);
dGip2m = 0.5*(Gbc[i+3] - Gbc[i+1]);
duip2b = (ubc[i+2] - ubc[i+1]);
duip2f = (ubc[i+3] - ubc[i+2]);
duip2m = 0.5*(ubc[i+3] - ubc[i+1]);
//limiting
dwip2 = minmod(theta*dwip2b,theta*dwip2f,dwip2m);
dhip2 = minmod(theta*dhip2b,theta*dhip2f,dhip2m);
dGip2 = minmod(theta*dGip2b,theta*dGip2f,dGip2m);
duip2 = minmod(theta*duip2b,theta*duip2f,duip2m);
//reconstruct left
hip2l = hbc[i+2] - 0.5*dhip2;
wip2l = wip2 - 0.5*dwip2;
Gip2l = Gbc[i+2] - 0.5*dGip2;
uip2l = ubc[i+2] - 0.5*duip2;
bip2l = wip2l - hip2l;
//calculate forces
//right force i
nbi = fmax(bip1l,bir);
hihm = fmax(0.0,wir-nbi);
hihp = fmax(0.0,wip1l-nbi);
//printf("%d || hihm - hir : %e || hihp - wip1l : %e\n", i, hihm - hir,hihp - wip1l);
her = hihp;
Ger = Gip1l;
uer = uip1l;
hel = hihm;
Gel = Gir;
uel = uir;
duer = idx*(uip2l - uip1l);
dber = idx*(bip2l - bip1l);
duel = idx*(uir - uim1r);
dbel = idx*(bir - bim1r);
ue = 0.5*(ubc[i+1]+ubc[i]);
due = idx*(ubc[i+1] - ubc[i]);
dbe = idx*(bedbc[i+1] - bedbc[i]);
uer = ue;
uel = ue;
duer = due;
duel = due;
dber = dbe;
dbel = dbe;
sqrtghel = sqrt(g*hel);
sqrtgher = sqrt(g*her);
sl = fmin(0,fmin(uel - sqrtghel, uer - sqrtgher));
sr = fmax(0,fmax(uel + sqrtghel, uer + sqrtgher));
felh = uel*hel;
felG = Gel*uel + 0.5*g*hel*hel - 2*i3*hel*hel*hel*duel*duel + hel*hel*uel*duel*dbel;
ferh = uer*her;
ferG = Ger*uer + 0.5*g*her*her - 2*i3*her*her*her*duer*duer + her*her*uer*duer*dber;
isrmsl = 0.0;
if(sr != sl) isrmsl = 1.0 / (sr - sl);
foh = isrmsl*(sr*felh - sl*ferh + sl*sr*(her - hel ));
foG = isrmsl*(sr*felG - sl*ferG + sl*sr*(Ger - Gel ));
double th,tu,tux,tbx,tbxx, sourcer, sourcec, sourcel;
//calculate the source term
th = hbc[i];
tu = ubc[i];
tux = 0.5*idx*(ubc[i+1] - ubc[i-1]);
tbx = 0.5*idx*(bedbc[i+1] - bedbc[i-1]);
tbxx = idx*idx*(bedbc[i+1] - 2*bedbc[i] + bedbc[i-1]);
sourcer = g*0.5*(hihm*hihm - hir*hir);
sourcec = g*th*tbx + 0.5*th*th*tu*tux*tbxx - th*tu*tu*tbx*tbxx ;
sourcel = g*0.5*(hil*hil - himhp*himhp);
//printf("%d | Source: %e\n",i, dt*idx*(sourcer+sourcel + sourcec));
hblank[i-nBCn] = hbc[i] - dt*idx*(foh - fih);
Gblank[i-nBCn] = Gbc[i] - dt*idx*(foG - fiG) + dt*idx*(sourcer+sourcel + sourcec);
fih = foh;
fiG = foG;
himhp = hihp;
him1r = hir;
wim1r = wir;
bim1r = bir;
Gim1r = Gir;
uim1r = uir;
hil = hip1l;
wil = wip1l;
bil = bip1l;
Gil = Gip1l;
uil = uip1l;
hir = hip1r;
wir = wip1r;
bir = bip1r;
Gir = Gip1r;
uir = uip1r;
hip1l = hip2l;
wip1l = wip2l;
bip1l = bip2l;
Gip1l = Gip2l;
uip1l = uip2l;
dhip1 = dhip2;
dwip1 = dwip2;
duip1 = duip2;
dGip1 = dGip2;
}
free(u);
free(ubeg);
free(uend);
free(bbeg);
free(bend);
free(hbeg);
free(hend);
free(gbeg);
free(gend);
free(Gbc);
free(hbc);
free(ubc);
free(bedbc);
}
/**
* @brief
* Use minmod function to limit the gradient and get the linear
* limited result.
*
* Usages:
* shape = mesh.Shape;
* hlim = Minmod1d_Mex...
* (h, mesh.J, shape.M, shape.Fmask, mesh.EToE, mesh.x)
*/
void mexFunction (int nlhs, mxArray *plhs[],
int nrhs, const mxArray *prhs[]){
/* check input & output */
if (nrhs != 6)
mexErrMsgTxt("Wrong number of input arguments.");
if (nlhs != 1)
mexErrMsgTxt("Wrong number of output arguments");
/* get inputs */
real *h = mxGetPr(prhs[0]);
real *J = mxGetPr(prhs[1]);
real *M = mxGetPr(prhs[2]);
real *Fmask= mxGetPr(prhs[3]);
real *EToE = mxGetPr(prhs[4]);
real *x = mxGetPr(prhs[5]);
/* get dimensions */
size_t Np, K;
Np = mxGetM(prhs[0]);
K = mxGetN(prhs[0]);
size_t Nfaces,Nfp;
Nfaces = mxGetM(prhs[3]);
Nfp = mxGetN(prhs[3]);
/* allocation of output */
plhs[0] = mxCreateDoubleMatrix((mwSize)Np, (mwSize)K, mxREAL);
real *hlim = mxGetPr(plhs[0]);
/* cell averages */
real *hmean = (real*) malloc(sizeof(real)*K );
real *area = (real*) malloc(sizeof(real)*K );
//elemental integral coefficient
real *w = (real*) malloc(sizeof(real)*Np);
int i,j,k,sk;
for(i=0;i<Np;i++){
w[i] = 0.0;
for(j=0;j<Np;j++){
w[i] += M[i*Np + j];
}
}
for (k=0;k<K;k++){
area [k] = 0.0;
hmean[k] = 0.0;
for(i=0;i<Np;i++){
sk = (k*Np + i);
hmean[k] += w[i]*J[sk]*h[sk];
area [k] += w[i]*J[sk];
}
hmean[k] /= area[k];
}
// reconstruct over each cell
int v1 = (int) Fmask[0] - 1;
int v2 = (int) Fmask[1] - 1;
for(k=0;k<K;k++){
/* cell gradient */
int sk1 = k*Np + v1;
int sk2 = k*Np + v2;
real xc =(x[sk2] + x[sk1])/2.0;
real dh = h[sk2] - h[sk1];
/* limited gradient */
int e1 = (int)EToE[k ] - 1;
int e2 = (int)EToE[k+K] - 1;
real a[3];
a[0] = dh/area[k];
a[1] = (hmean[e2] - hmean[k] )/area[k];
a[2] = (hmean[k] - hmean[e1])/area[k];
real dhlim;
minmod(3, a, &dhlim);
/* reconstruction */
for(i=0;i<Np;i++){
sk = (k*Np + i);
real delta = (real)(x[sk] - xc);
hlim[sk] = hmean[k] + delta*dhlim;
}
}
free(w); free(hmean); free(area);
return;
}
void test_minmod()
{
double a[3];
double b[3];
a[2] = 2.;
a[1] = 1.;
a[0] = 0.;
b[2] = -2.;
b[1] = -1.;
b[0] = -0.;
for(int i=0;i<3;i++)
{
printf(
" signbit(a[%d]) = %d"
" signbit(b[%d]) = %d\n",
i, signbit(a[i]),
i, signbit(b[i])
);
for(int j=0;j<3;j++)
{
printf(
" minmod(a[%d],b[%d]) = %f "
" minmod(a[%d],a[%d]) = %f "
" minmod(b[%d],b[%d]) = %f "
" minmod(b[%d],a[%d]) = %f \n",
i,j, minmod(a[i],b[j]),
i,j, minmod(a[i],a[j]),
i,j, minmod(b[i],b[j]),
i,j, minmod(b[i],a[j])
);
}
}
for(int i=0;i<3;i++)
for(int j=0;j<3;j++)
for(int k=0;k<3;k++)
{
printf(
" minmod(a[%d],a[%d],a[%d]) = %f "
" minmod(b[%d],b[%d],b[%d]) = %f \n",
i,j,k, minmod(a[i],a[j],a[k]),
i,j,k, minmod(b[i],b[j],b[k]));
}
for(int i=0;i<3;i++)
for(int j=0;j<3;j++)
{
printf(
" minmod(a[ 0],a[%d],a[%d]) = %f "
" minmod(a[%d],a[ 0],a[%d]) = %f "
" minmod(a[%d],a[%d],a[ 0]) = %f "
" minmod(b[ 0],b[%d],b[%d]) = %f "
" minmod(b[%d],b[ 0],b[%d]) = %f "
" minmod(b[%d],b[%d],b[ 0]) = %f \n",
i,j, minmod(a[ 0],a[ i],a[ j]),
i,j, minmod(a[ i],a[ 0],a[ j]),
i,j, minmod(a[ i],a[ j],a[ 0]),
i,j, minmod(b[ 0],b[ i],b[ j]),
i,j, minmod(b[ i],b[ 0],b[ j]),
i,j, minmod(b[ i],b[ j],b[ 0]));
}
}
// -------------------------------------------------------------------------- //
//
// Limiter based on a modification of the following paper:
// L. Krivodonova. "Limiters for high-order discontinuous
// Galerkin methods." J. Comp. Phys., Vol. 226, pg 879-896.
//
// For systems, we have orders 1--3 implemented. For scalar problems, we have
// orders 1--5 implemented. This routine calls ApplyScalarLimiter in the scalar
// case.
//
// -------------------------------------------------------------------------- //
void ApplyLimiterKrivodonova(dTensorBC4& aux, dTensorBC4& q,
void (*ProjectRightEig)(int ixy, // Component of flux function (1 or 2)
const dTensor1& Aux_ave, const dTensor1& Q_ave, // Riemann states
const dTensor2& Wvals, // Characteristic variables
dTensor2& Qvals // Conserved variables
),
void (*ProjectLeftEig)(int ixy, // Component of flux function (1 or 2)
const dTensor1& Aux_ave, const dTensor1& Q_ave, // Riemann states
const dTensor2& Qvals, // Conserved variables
dTensor2& Wvals) // Characteristic variables
)
{
// Problem dimensions
const int mx = q.getsize(1);
const int my = q.getsize(2);
const int meqn = q.getsize(3);
const int kmax = q.getsize(4);
const int mbc = q.getmbc();
const int maux = aux.getsize(3);
// Number of basis functions
const int space_order = dogParams.get_space_order();
assert_eq(int((sqrt(1+8*kmax)-1)/2),space_order);
const int ksize = (space_order==2)?1:3;
// Nothing to limit in the first-order case
if(space_order==1)
{ return; }
// Scalar limiter is implemented for all orders up to 5. Call this routine
// instead if this is a scalar problem.
if( meqn == 1 )
{
void ApplyScalarLimiter(const dTensorBC4& aux, dTensorBC4& q);
ApplyScalarLimiter(aux, q);
return;
}
// Storage
dTensorBC4 wx_cent(mx,my,meqn,ksize,mbc);
dTensorBC4 wy_cent(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_right(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_left(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_up(mx,my,meqn,ksize,mbc);
dTensorBC4 dw_down(mx,my,meqn,ksize,mbc);
void ConvertQtoW(int ixy,
const dTensorBC4& aux,
const dTensorBC4& qold,
dTensorBC4& dwp,
dTensorBC4& dwm,
dTensorBC4& w_cent,
void (*ProjectLeftEig)(int,
const dTensor1&,
const dTensor1&,
const dTensor2&,
dTensor2&));
void ConvertWtoQ(int ixy,
const dTensorBC4& aux,
const dTensorBC4& qold,
const dTensorBC4& win,
dTensorBC4& qout,
void (*ProjectRightEig)(int,
const dTensor1&,
const dTensor1&,
const dTensor2&,
dTensor2&));
// ----------------------------------------------------------
// Key: storage of characteristic variables
//
// wx_cent(i,j,me,1) = Lx * q(i,j,me,2)
// wx_cent(i,j,me,2) = Lx * q(i,j,me,4)
// wx_cent(i,j,me,3) = Lx * q(i,j,me,5)
//
// dw_right(i,j,me,1) = Lx * (q(i+1,j,me,1)-q(i,j,me,1))
// dw_right(i,j,me,2) = Lx * (q(i+1,j,me,2)-q(i,j,me,2))
// dw_right(i,j,me,3) = Lx * (q(i,j+1,me,2)-q(i,j,me,2))
//
// dw_left(i,j,me,1) = Lx * (q(i,j,me,1)-q(i-1,j,me,1))
// dw_left(i,j,me,2) = Lx * (q(i,j,me,2)-q(i-1,j,me,2))
// dw_left(i,j,me,3) = Lx * (q(i,j,me,2)-q(i,j-1,me,2))
//
// wy_cent(i,j,me,1) = Ly * q(i,j,me,3)
// wy_cent(i,j,me,2) = Ly * q(i,j,me,4)
// wy_cent(i,j,me,3) = Ly * q(i,j,me,6)
//
// dw_up(i,j,me,1) = Ly * (q(i,j+1,me,1)-q(i,j,me,1))
// dw_up(i,j,me,2) = Ly * (q(i+1,j,me,3)-q(i,j,me,3))
// dw_up(i,j,me,3) = Ly * (q(i,j+1,me,3)-q(i,j,me,3))
//
// dw_down(i,j,me,1) = Ly * (q(i,j,me,1)-q(i,j-1,me,1))
// dw_down(i,j,me,2) = Ly * (q(i,j,me,3)-q(i-1,j,me,3))
// dw_down(i,j,me,3) = Ly * (q(i,j,me,3)-q(i,j-1,me,3))
//
// ----------------------------------------------------------
// Moment limiter (highest to lowest)
// Reminder: this is for 3rd and 2nd-order case only.
switch ( space_order )
{
default:
unsupported_value_error(space_order);
break;
case 2: // 2nd order in space
// Convert to characteristic variables in x-direction
ConvertQtoW(1,aux,q,dw_right,dw_left,wx_cent,ProjectLeftEig);
// Convert to characteristic variables in y-direction
ConvertQtoW(2,aux,q,dw_up,dw_down,wy_cent,ProjectLeftEig);
// Limit in both the x-direction and the y-direction
#pragma omp parallel for
for (int i=(3-mbc); i<=(mx+mbc-2); i++)
for (int j=(3-mbc); j<=(my+mbc-2); j++)
for (int m=1; m<=meqn; m++)
{
// x-direction: limit linear term
{
const double dwp = dw_right.get(i,j,m,1);
const double dwm = dw_left.get(i,j,m,1);
const double wx_now = wx_cent.get(i,j,m,1);
const double wx_limited = minmod(wx_now, dwp/sq3, dwm/sq3);
wx_cent.set(i,j,m,1, wx_limited );
}
// y-direction: limit linear term
{
const double dwp = dw_up.get(i,j,m,1);
const double dwm = dw_down.get(i,j,m,1);
const double wy_now = wy_cent.get(i,j,m,1);
const double wy_limited = minmod(wy_now, dwp/sq3, dwm/sq3);
wy_cent.set(i,j,m,1, wy_limited );
}
}
// Convert back to conserved variables in x-direction
ConvertWtoQ(1,aux,q,wx_cent,q,ProjectRightEig);
// Convert back to conserved variables in y-direction
ConvertWtoQ(2,aux,q,wy_cent,q,ProjectRightEig);
break;
case 3: // 3rd order in space
// Convert to characteristic variables in x-direction
ConvertQtoW(1, aux, q, dw_right, dw_left, wx_cent, ProjectLeftEig);
// Convert to characteristic variables in y-direction
ConvertQtoW(2, aux, q, dw_up, dw_down, wy_cent, ProjectLeftEig);
// Limit
const double molifier = 0.6;
const double fac1 = molifier*sq3*osq5; // osq3*osq5
const double fac2 = molifier*osq3; // osq3
#pragma omp parallel for
for (int i=(3-mbc); i<=(mx+mbc-2); i++)
for (int j=(3-mbc); j<=(my+mbc-2); j++)
for (int m=1; m<=meqn; m++)
{
int mflag = 0;
double dwp,dwm,wx_limited,wx_now,wy_limited,wy_now;
// ---------------------------------------------------
// x-direction: limit quadratic term
dwp = dw_right.get(i,j,m,2);
dwm = dw_left.get(i,j,m,2);
wx_now = wx_cent.get(i,j,m,3);
wx_limited = minmod(wx_now, fac1*dwp, fac1*dwm);
wx_cent.set(i,j,m,3, wx_limited );
if (fabs(wx_now - wx_limited)>=1.0e-14 ||
fabs(wx_limited)<1.0e-14)
{ mflag = 1; }
// ---------------------------------------------------
// ---------------------------------------------------
// y-direction: limit quadratic term
dwp = dw_up.get(i,j,m,3);
dwm = dw_down.get(i,j,m,3);
wy_now = wy_cent.get(i,j,m,3);
wy_limited = minmod(wy_now, fac1*dwp, fac1*dwm);
wy_cent.set(i,j,m,3, wy_limited );
if (fabs(wy_now - wy_limited)>=1.0e-14 ||
fabs(wy_limited)<1.0e-14)
{ mflag = 1; }
// ---------------------------------------------------
if (mflag==1)
{
// ---------------------------------------------------
// x-direction: limit mixed term
dwp = dw_right.get(i,j,m,3);
dwm = dw_left.get(i,j,m,3);
wx_now = wx_cent.get(i,j,m,2);
wx_limited = minmod(wx_now, fac1*dwp, fac1*dwm);
wx_cent.set(i,j,m,2, wx_limited );
// ---------------------------------------------------
// ---------------------------------------------------
// x-direction: limit linear term
dwp = dw_right.get(i,j,m,1);
dwm = dw_left.get(i,j,m,1);
wx_now = wx_cent.get(i,j,m,1);
wx_limited = minmod(wx_now, fac2*dwp, fac2*dwm);
wx_cent.set(i,j,m,1, wx_limited );
// ---------------------------------------------------
// ---------------------------------------------------
// y-direction: limit mixed term
dwp = dw_up.get(i,j,m,2);
dwm = dw_down.get(i,j,m,2);
wy_now = wy_cent.get(i,j,m,2);
wy_limited = minmod(wy_now, fac1*dwp, fac1*dwm);
wy_cent.set(i,j,m,2, wy_limited );
// ---------------------------------------------------
// ---------------------------------------------------
// y-direction: limit linear term
dwp = dw_up.get(i,j,m,1);
dwm = dw_down.get(i,j,m,1);
wy_now = wy_cent.get(i,j,m,1);
wy_limited = minmod(wy_now, fac2*dwp, fac2*dwm);
wy_cent.set(i,j,m,1, wy_limited );
// ---------------------------------------------------
}
}
// Convert back to conserved variables in x-direction
ConvertWtoQ(1, aux, q, wx_cent, q, ProjectRightEig);
// Convert back to conserved variables in y-direction
ConvertWtoQ(2, aux, q, wy_cent, q, ProjectRightEig);
break;
}
}
