int main()
{
int array[100];
/*
* Decreasing sequential numbers
*/
for (int i = 0; i < (int) LEN(array); i++) {
array[i] = 100 - i;
}
quicksort(array, 0, LEN(array) - 1);
PRINTARRAY(array);
printf("\n");
/*
* Random numbers from 0 - 100
*/
time_t t;
srand((unsigned) time(&t));
for (int i = 0; i < (int) LEN(array); i++) {
array[i] = rand() % 101;
}
quicksort(array, 0, LEN(array) - 1);
PRINTARRAY(array);
return 0;
}
int main()
{
int i;
int b[10] = {2,4,6,8,10,12,14,16,18,20};
printf("Array values are \n");
PRINTARRAY(b,10);
return 0;
}
void
oclHydroGodunov(long idimStart, real_t dt, const hydroparam_t H, hydrovar_t * Hv, hydrowork_t * Hw, hydrovarwork_t * Hvw)
{
cl_int status;
// Local variables
struct timespec start, end;
int i, j, idim, idimIndex;
int Hmin, Hmax, Hstep, Hnxystep;
int Hdimsize;
int Hndim_1;
int slices, iend;
real_t dtdx;
size_t lVarSz = H.arVarSz * H.nxystep * sizeof(real_t);
long Hnxyt = H.nxyt;
int clear = 0;
static FILE *fic = NULL;
if (fic == NULL && H.prt) {
char logname[256];
sprintf(logname, "TRACE.%04d_%04d.txt", H.nproc, H.mype);
fic = fopen(logname, "w");
}
WHERE("hydro_godunov");
if (H.prt) fprintf(fic, "loop dt=%lg\n", dt);
for (idimIndex = 0; idimIndex < 2; idimIndex++) {
idim = (idimStart - 1 + idimIndex) % 2 + 1;
// constant
// constant
dtdx = dt / H.dx;
// Update boundary conditions
if (H.prt) {
fprintf(fic, "godunov %d\n", idim);
PRINTUOLD(fic, H, Hv);
}
#define GETUOLD oclGetUoldFromDevice(H, Hv)
start = cclock();
oclMakeBoundary(idim, H, Hv, uoldDEV);
end = cclock();
functim[TIM_MAKBOU] += ccelaps(start, end);
if (H.prt) {fprintf(fic, "MakeBoundary\n");}
if (H.prt) {GETUOLD; PRINTUOLD(fic, H, Hv);}
if (idim == 1) {
Hmin = H.jmin + ExtraLayer;
Hmax = H.jmax - ExtraLayer;
Hdimsize = H.nxt;
Hndim_1 = H.nx + 1;
Hstep = H.nxystep;
} else {
Hmin = H.imin + ExtraLayer;
Hmax = H.imax - ExtraLayer;
Hdimsize = H.nyt;
Hndim_1 = H.ny + 1;
Hstep = H.nxystep;
}
Hnxystep = Hstep;
for (i = Hmin; i < Hmax; i += Hstep) {
long offsetIP = IHVWS(0, 0, IP);
long offsetID = IHVWS(0, 0, ID);
int Hnxyt = H.nxyt;
iend = i + Hstep;
if (iend >= Hmax) iend = Hmax;
slices = iend - i;
if (clear) oclMemset(uDEV, 0, lVarSz);
start = cclock();
oclGatherConservativeVars(idim, i, H.imin, H.imax, H.jmin, H.jmax, H.nvar, H.nxt, H.nyt, H.nxyt, slices, Hnxystep, uoldDEV, uDEV);
end = cclock();
functim[TIM_GATCON] += ccelaps(start, end);
if (H.prt) {fprintf(fic, "ConservativeVars %ld %ld %ld %ld %d %d\n", H.nvar, H.nxt, H.nyt, H.nxyt, slices, Hstep);}
if (H.prt) { GETARRV(uDEV, Hvw->u); }
PRINTARRAYV2(fic, Hvw->u, Hdimsize, "u", H);
// Convert to primitive variables
start = cclock();
oclConstoprim(Hdimsize, H.nxyt, H.nvar, H.smallr, slices, Hnxystep, uDEV, qDEV, eDEV);
end = cclock();
functim[TIM_CONPRI] += ccelaps(start, end);
if (H.prt) { GETARR (eDEV, Hw->e); }
if (H.prt) { GETARRV(qDEV, Hvw->q); }
PRINTARRAY(fic, Hw->e, Hdimsize, "e", H);
PRINTARRAYV2(fic, Hvw->q, Hdimsize, "q", H);
start = cclock();
oclEquationOfState(offsetIP, offsetID, 0, Hdimsize, H.smallc, H.gamma, slices, H.nxyt, qDEV, eDEV, cDEV);
end = cclock();
functim[TIM_EOS] += ccelaps(start, end);
if (H.prt) { GETARR (cDEV, Hw->c); }
PRINTARRAY(fic, Hw->c, Hdimsize, "c", H);
if (H.prt) { GETARRV (qDEV, Hvw->q); }
PRINTARRAYV2(fic, Hvw->q, Hdimsize, "q", H);
if (clear) oclMemset(dqDEV, 0, H.arVarSz * H.nxystep);
// Characteristic tracing
if (H.iorder != 1) {
if (clear) oclMemset(dqDEV, 0, H.arVarSz);
start = cclock();
oclSlope(Hdimsize, H.nvar, H.nxyt, H.slope_type, slices, Hstep, qDEV, dqDEV);
end = cclock();
functim[TIM_SLOPE] += ccelaps(start, end);
if (H.prt) { GETARRV(dqDEV, Hvw->dq); }
PRINTARRAYV2(fic, Hvw->dq, Hdimsize, "dq", H);
}
start = cclock();
oclTrace(dtdx, Hdimsize, H.scheme, H.nvar, H.nxyt, slices, Hstep, qDEV, dqDEV, cDEV, qxmDEV, qxpDEV);
end = cclock();
functim[TIM_TRACE] += ccelaps(start, end);
if (H.prt) { GETARRV(qxmDEV, Hvw->qxm); }
if (H.prt) { GETARRV(qxpDEV, Hvw->qxp); }
PRINTARRAYV2(fic, Hvw->qxm, Hdimsize, "qxm", H);
PRINTARRAYV2(fic, Hvw->qxp, Hdimsize, "qxp", H);
start = cclock();
oclQleftright(idim, H.nx, H.ny, H.nxyt, H.nvar, slices, Hstep, qxmDEV, qxpDEV, qleftDEV, qrightDEV);
end = cclock();
functim[TIM_QLEFTR] += ccelaps(start, end);
if (H.prt) { GETARRV(qleftDEV, Hvw->qleft); }
if (H.prt) { GETARRV(qrightDEV, Hvw->qright); }
PRINTARRAYV2(fic, Hvw->qleft, Hdimsize, "qleft", H);
PRINTARRAYV2(fic, Hvw->qright, Hdimsize, "qright", H);
// Solve Riemann problem at interfaces
start = cclock();
oclRiemann(Hndim_1, H.smallr, H.smallc, H.gamma, H.niter_riemann, H.nvar, H.nxyt, slices, Hstep,
qleftDEV, qrightDEV, qgdnvDEV,sgnmDEV);
end = cclock();
functim[TIM_RIEMAN] += ccelaps(start, end);
if (H.prt) { GETARRV(qgdnvDEV, Hvw->qgdnv); }
PRINTARRAYV2(fic, Hvw->qgdnv, Hdimsize, "qgdnv", H);
// Compute fluxes
if (clear) oclMemset(fluxDEV, 0, H.arVarSz);
start = cclock();
oclCmpflx(Hdimsize, H.nxyt, H.nvar, H.gamma, slices, Hnxystep, qgdnvDEV, fluxDEV);
end = cclock();
functim[TIM_CMPFLX] += ccelaps(start, end);
if (H.prt) { GETARRV(fluxDEV, Hvw->flux); }
PRINTARRAYV2(fic, Hvw->flux, Hdimsize, "flux", H);
if (H.prt) { GETARRV(uDEV, Hvw->u); }
PRINTARRAYV2(fic, Hvw->u, Hdimsize, "u", H);
// if (H.prt) {
// GETUOLD; PRINTUOLD(fic, H, Hv);
// }
if (H.prt) fprintf(fic, "dxdt=%lg\n", dtdx);
start = cclock();
oclUpdateConservativeVars(idim, i, dtdx, H.imin, H.imax, H.jmin, H.jmax, H.nvar, H.nxt, H.nyt, H.nxyt, slices, Hnxystep,
uoldDEV, uDEV, fluxDEV);
end = cclock();
functim[TIM_UPDCON] += ccelaps(start, end);
if (H.prt) {
GETUOLD; PRINTUOLD(fic, H, Hv);
}
} // for j
if (H.prt) {
// printf("After pass %d\n", idim);
PRINTUOLD(fic, H, Hv);
}
}
} // hydro_godunov
void
hydro_godunov(long idim, double dt, const hydroparam_t H, hydrovar_t * Hv, hydrowork_t * Hw, hydrovarwork_t * Hvw)
{
// Local variables
long i, j;
double dtdx;
double *dq;
double *e;
double *flux;
double *q;
double *qleft, *qright;
double *qxm, *qxp;
double *u;
double *c;
double *uold;
double *rl, *ul, *pl, *cl, *wl, *rr, *ur, *pr, *cr, *wr, *ro, *uo, *po, *co, *wo,
*rstar, *ustar, *pstar, *cstar, *spin, *spout, *ushock, *frac, *scr, *delp, *pold;
long *sgnm, *ind, *ind2;
double *qgdnv;
WHERE("hydro_godunov");
// constant
dtdx = dt / H.dx;
// Update boundary conditions
if (H.prt) {
fprintf(stdout, "godunov %ld\n", idim);
PRINTUOLD(H, Hv);
}
make_boundary(idim, H, Hv);
PRINTUOLD(H, Hv);
// Allocate work space for 1D sweeps
allocate_work_space(H, Hw, Hvw);
uold = Hv->uold;
qgdnv = Hvw->qgdnv;
flux = Hvw->flux;
c = Hw->c;
q = Hvw->q;
e = Hw->e;
u = Hvw->u;
qxm = Hvw->qxm;
qxp = Hvw->qxp;
qleft = Hvw->qleft;
qright = Hvw->qright;
dq = Hvw->dq;
rl = Hw->rl;
ul = Hw->ul;
pl = Hw->pl;
cl = Hw->cl;
wl = Hw->wl;
rr = Hw->rr;
ur = Hw->ur;
pr = Hw->pr;
cr = Hw->cr;
wr = Hw->wr;
ro = Hw->ro;
uo = Hw->uo;
po = Hw->po;
co = Hw->co;
wo = Hw->wo;
rstar = Hw->rstar;
ustar = Hw->ustar;
pstar = Hw->pstar;
cstar = Hw->cstar;
sgnm = Hw->sgnm;
spin = Hw->spin;
spout = Hw->spout;
ushock = Hw->ushock;
frac = Hw->frac;
scr = Hw->scr;
delp = Hw->delp;
pold = Hw->pold;
ind = Hw->ind;
ind2 = Hw->ind2;
if (idim == 1) {
for (j = H.jmin + ExtraLayer; j < H.jmax - ExtraLayer; j++) {
double *qID = &q[IHvw(0, ID)];
double *qIP = &q[IHvw(0, IP)];
gatherConservativeVars(idim, j, uold, u, H.imin, H.imax, H.jmin, H.jmax, H.nvar, H.nxt, H.nyt, H.nxyt);
// Convert to primitive variables
constoprim(u, q, e, H.nxt, H.nxyt, H.nvar, H.smallr);
equation_of_state(qID, e, qIP, c, 0, H.nxt, H.smallc, H.gamma);
Dmemset(dq, 0, (H.nxyt + 2) * H.nvar);
// Characteristic tracing
if (H.iorder != 1) {
slope(q, dq, H.nxt, H.nvar, H.nxyt, H.slope_type);
}
trace(q, dq, c, qxm, qxp, dtdx, H.nxt, H.scheme, H.nvar, H.nxyt);
qleftright(idim, H.nx, H.ny, H.nxyt, H.nvar, qxm, qxp, qleft, qright);
// Solve Riemann problem at interfaces
riemann(qleft, qright, qgdnv,
rl, ul, pl, cl, wl, rr, ur, pr, cr, wr, ro, uo, po, co, wo,
rstar, ustar, pstar, cstar,
sgnm, spin, spout, ushock, frac,
scr, delp, pold, ind, ind2, H.nx + 1, H.smallr, H.smallc, H.gamma, H.niter_riemann, H.nvar, H.nxyt);
// Compute fluxes
cmpflx(qgdnv, flux, H.nxt, H.nxyt, H.nvar, H.gamma);
updateConservativeVars(idim, j, dtdx, uold, u, flux, H.imin, H.imax, H.jmin, H.jmax, H.nvar, H.nxt, H.nyt, H.nxyt);
} // for j
if (H.prt) {
printf("After pass %ld\n", idim);
PRINTUOLD(H, Hv);
}
} else {
for (i = H.imin + ExtraLayer; i < H.imax - ExtraLayer; i++) {
double *qID = &Hvw->q[IHvw(0, ID)];
double *qIP = &Hvw->q[IHvw(0, IP)];
gatherConservativeVars(idim, i, uold, u, H.imin, H.imax, H.jmin, H.jmax, H.nvar, H.nxt, H.nyt, H.nxyt);
PRINTARRAYV(Hvw->u, H.nyt, "uY", H);
// Convert to primitive variables
constoprim(u, q, e, H.nyt, H.nxyt, H.nvar, H.smallr);
equation_of_state(qID, e, qIP, c, 0, H.nyt, H.smallc, H.gamma);
PRINTARRAY(Hw->c, H.nyt, "cY", H);
// Characteristic tracing
// compute slopes
Dmemset(dq, 0, H.nyt * H.nvar);
if (H.iorder != 1) {
slope(q, dq, H.nyt, H.nvar, H.nxyt, H.slope_type);
}
PRINTARRAYV(Hvw->dq, H.nyt, "dqY", H);
trace(q, dq, c, qxm, qxp, dtdx, H.nyt, H.scheme, H.nvar, H.nxyt);
qleftright(idim, H.nx, H.ny, H.nxyt, H.nvar, qxm, qxp, qleft, qright);
PRINTARRAYV(Hvw->qleft, H.ny + 1, "qleftY", H);
PRINTARRAYV(Hvw->qright, H.ny + 1, "qrightY", H);
// Solve Riemann problem at interfaces
riemann(qleft, qright, qgdnv, rl, ul,
pl, cl, wl, rr, ur, pr,
cr, wr, ro, uo, po, co,
wo, rstar, ustar, pstar, cstar,
sgnm, spin, spout, ushock, frac,
scr, delp, pold, ind, ind2, H.ny + 1, H.smallr, H.smallc, H.gamma, H.niter_riemann, H.nvar, H.nxyt);
// Compute fluxes
cmpflx(qgdnv, flux, H.nyt, H.nxyt, H.nvar, H.gamma);
PRINTARRAYV(Hvw->flux, H.ny + 1, "fluxY", H);
// updateConservativeVars(idim, i, dtdx, H, Hv, Hvw);
updateConservativeVars(idim, i, dtdx, uold, u, flux, H.imin, H.imax, H.jmin, H.jmax, H.nvar, H.nxt, H.nyt, H.nxyt);
} // else
if (H.prt) {
printf("After pass %ld\n", idim);
PRINTUOLD(H, Hv);
}
}
// Deallocate work space
deallocate_work_space(H, Hw, Hvw);
} // hydro_godunov
// variables auxiliaires pour mettre en place le mode resident de HMPP
void
hydro_godunov(int idimStart, real_t dt, const hydroparam_t H, hydrovar_t * Hv, hydrowork_t * Hw, hydrovarwork_t * Hvw) {
// Local variables
struct timespec start, end;
int j;
real_t dtdx;
int clear=0;
real_t (*e)[H.nxyt];
real_t (*flux)[H.nxystep][H.nxyt];
real_t (*qleft)[H.nxystep][H.nxyt];
real_t (*qright)[H.nxystep][H.nxyt];
real_t (*c)[H.nxyt];
real_t *uold;
int (*sgnm)[H.nxyt];
real_t (*qgdnv)[H.nxystep][H.nxyt];
real_t (*u)[H.nxystep][H.nxyt];
real_t (*qxm)[H.nxystep][H.nxyt];
real_t (*qxp)[H.nxystep][H.nxyt];
real_t (*q)[H.nxystep][H.nxyt];
real_t (*dq)[H.nxystep][H.nxyt];
static FILE *fic = NULL;
if (fic == NULL && H.prt == 1) {
char logname[256];
sprintf(logname, "TRACE.%04d_%04d.txt", H.nproc, H.mype);
fic = fopen(logname, "w");
}
WHERE("hydro_godunov");
// int hmppGuard = 1;
int idimIndex = 0;
for (idimIndex = 0; idimIndex < 2; idimIndex++) {
int idim = (idimStart - 1 + idimIndex) % 2 + 1;
// constant
dtdx = dt / H.dx;
// Update boundary conditions
if (H.prt) {
fprintf(fic, "godunov %d\n", idim);
PRINTUOLD(fic, H, Hv);
}
// if (H.mype == 1) fprintf(fic, "Hydro makes boundary.\n");
start = cclock();
make_boundary(idim, H, Hv);
end = cclock();
functim[TIM_MAKBOU] += ccelaps(start, end);
if (H.prt) {fprintf(fic, "MakeBoundary\n");}
PRINTUOLD(fic, H, Hv);
uold = Hv->uold;
qgdnv = (real_t (*)[H.nxystep][H.nxyt]) Hvw->qgdnv;
flux = (real_t (*)[H.nxystep][H.nxyt]) Hvw->flux;
c = (real_t (*)[H.nxyt]) Hw->c;
e = (real_t (*)[H.nxyt]) Hw->e;
qleft = (real_t (*)[H.nxystep][H.nxyt]) Hvw->qleft;
qright = (real_t (*)[H.nxystep][H.nxyt]) Hvw->qright;
sgnm = (int (*)[H.nxyt]) Hw->sgnm;
q = (real_t (*)[H.nxystep][H.nxyt]) Hvw->q;
dq = (real_t (*)[H.nxystep][H.nxyt]) Hvw->dq;
u = (real_t (*)[H.nxystep][H.nxyt]) Hvw->u;
qxm = (real_t (*)[H.nxystep][H.nxyt]) Hvw->qxm;
qxp = (real_t (*)[H.nxystep][H.nxyt]) Hvw->qxp;
int Hmin, Hmax, Hstep;
int Hdimsize;
int Hndim_1;
if (idim == 1) {
Hmin = H.jmin + ExtraLayer;
Hmax = H.jmax - ExtraLayer;
Hdimsize = H.nxt;
Hndim_1 = H.nx + 1;
Hstep = H.nxystep;
} else {
Hmin = H.imin + ExtraLayer;
Hmax = H.imax - ExtraLayer;
Hdimsize = H.nyt;
Hndim_1 = H.ny + 1;
Hstep = H.nxystep;
}
if (!H.nstep && idim == 1) {
/* LM -- HERE a more secure implementation should be used: a new parameter ? */
}
// if (H.mype == 1) fprintf(fic, "Hydro computes slices.\n");
for (j = Hmin; j < Hmax; j += Hstep) {
// we try to compute many slices each pass
int jend = j + Hstep;
if (jend >= Hmax)
jend = Hmax;
int slices = jend - j; // numbre of slices to compute
// fprintf(stderr, "Godunov idim=%d, j=%d %d \n", idim, j, slices);
if (clear) Dmemset((H.nxyt) * H.nxystep * H.nvar, (real_t *) dq, 0);
start = cclock();
gatherConservativeVars(idim, j, H.imin, H.imax, H.jmin, H.jmax, H.nvar, H.nxt, H.nyt, H.nxyt, slices, Hstep, uold,
u);
end = cclock();
functim[TIM_GATCON] += ccelaps(start, end);
if (H.prt) {fprintf(fic, "ConservativeVars %d %d %d %d %d %d\n", H.nvar, H.nxt, H.nyt, H.nxyt, slices, Hstep);}
PRINTARRAYV2(fic, u, Hdimsize, "u", H);
if (clear) Dmemset((H.nxyt) * H.nxystep * H.nvar, (real_t *) dq, 0);
// Convert to primitive variables
start = cclock();
constoprim(Hdimsize, H.nxyt, H.nvar, H.smallr, slices, Hstep, u, q, e);
end = cclock();
functim[TIM_CONPRI] += ccelaps(start, end);
PRINTARRAY(fic, e, Hdimsize, "e", H);
PRINTARRAYV2(fic, q, Hdimsize, "q", H);
start = cclock();
equation_of_state(0, Hdimsize, H.nxyt, H.nvar, H.smallc, H.gamma, slices, Hstep, e, q, c);
end = cclock();
functim[TIM_EOS] += ccelaps(start, end);
PRINTARRAY(fic, c, Hdimsize, "c", H);
PRINTARRAYV2(fic, q, Hdimsize, "q", H);
// Characteristic tracing
if (H.iorder != 1) {
start = cclock();
slope(Hdimsize, H.nvar, H.nxyt, H.slope_type, slices, Hstep, q, dq);
end = cclock();
functim[TIM_SLOPE] += ccelaps(start, end);
PRINTARRAYV2(fic, dq, Hdimsize, "dq", H);
}
if (clear) Dmemset(H.nxyt * H.nxystep * H.nvar, (real_t *) qxm, 0);
if (clear) Dmemset(H.nxyt * H.nxystep * H.nvar, (real_t *) qxp, 0);
if (clear) Dmemset(H.nxyt * H.nxystep * H.nvar, (real_t *) qleft, 0);
if (clear) Dmemset(H.nxyt * H.nxystep * H.nvar, (real_t *) qright, 0);
if (clear) Dmemset(H.nxyt * H.nxystep * H.nvar, (real_t *) flux, 0);
if (clear) Dmemset(H.nxyt * H.nxystep * H.nvar, (real_t *) qgdnv, 0);
start = cclock();
trace(dtdx, Hdimsize, H.scheme, H.nvar, H.nxyt, slices, Hstep, q, dq, c, qxm, qxp);
end = cclock();
functim[TIM_TRACE] += ccelaps(start, end);
PRINTARRAYV2(fic, qxm, Hdimsize, "qxm", H);
PRINTARRAYV2(fic, qxp, Hdimsize, "qxp", H);
start = cclock();
qleftright(idim, H.nx, H.ny, H.nxyt, H.nvar, slices, Hstep, qxm, qxp, qleft, qright);
end = cclock();
functim[TIM_QLEFTR] += ccelaps(start, end);
PRINTARRAYV2(fic, qleft, Hdimsize, "qleft", H);
PRINTARRAYV2(fic, qright, Hdimsize, "qright", H);
start = cclock();
riemann(Hndim_1, H.smallr, H.smallc, H.gamma, H.niter_riemann, H.nvar, H.nxyt, slices, Hstep, qleft, qright, qgdnv, sgnm, Hw);
end = cclock();
functim[TIM_RIEMAN] += ccelaps(start, end);
PRINTARRAYV2(fic, qgdnv, Hdimsize, "qgdnv", H);
start = cclock();
cmpflx(Hdimsize, H.nxyt, H.nvar, H.gamma, slices, Hstep, qgdnv, flux);
end = cclock();
functim[TIM_CMPFLX] += ccelaps(start, end);
PRINTARRAYV2(fic, flux, Hdimsize, "flux", H);
PRINTARRAYV2(fic, u, Hdimsize, "u", H);
start = cclock();
updateConservativeVars(idim, j, dtdx, H.imin, H.imax, H.jmin, H.jmax, H.nvar, H.nxt, H.nyt, H.nxyt, slices, Hstep,
uold, u, flux);
end = cclock();
functim[TIM_UPDCON] += ccelaps(start, end);
PRINTUOLD(fic, H, Hv);
} // for j
if (H.prt) {
// printf("[%d] After pass %d\n", H.mype, idim);
PRINTUOLD(fic, H, Hv);
}
}
if ((H.t + dt >= H.tend) || (H.nstep + 1 >= H.nstepmax)) {
/* LM -- HERE a more secure implementation should be used: a new parameter ? */
}
} // hydro_godunov