Revision 365487241b962a4c2bba472af24d581b81d96c80 authored by Eh Tan on 27 March 2007, 04:40:15 UTC, committed by Eh Tan on 27 March 2007, 04:40:15 UTC
1 parent 78e36d1
Global_operations.c
/*
*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
*
*<LicenseText>
*
* CitcomS by Louis Moresi, Shijie Zhong, Lijie Han, Eh Tan,
* Clint Conrad, Michael Gurnis, and Eun-seo Choi.
* Copyright (C) 1994-2005, California Institute of Technology.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
*</LicenseText>
*
*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
*/
#include <mpi.h>
#include <math.h>
#include <sys/types.h>
#include "element_definitions.h"
#include "global_defs.h"
/* ===============================================
strips horizontal average from nodal field X.
Assumes orthogonal mesh, otherwise, horizontals
aren't & another method is required.
=============================================== */
void remove_horiz_ave(E,X,H,store_or_not)
struct All_variables *E;
double **X, *H;
int store_or_not;
{
int m,i,j,k,n,nox,noz,noy;
void return_horiz_ave();
const int dims = E->mesh.nsd;
noy = E->lmesh.noy;
noz = E->lmesh.noz;
nox = E->lmesh.nox;
return_horiz_ave(E,X,H);
for(m=1;m<=E->sphere.caps_per_proc;m++)
for(k=1;k<=noy;k++)
for(j=1;j<=nox;j++)
for(i=1;i<=noz;i++) {
n = i+(j-1)*noz+(k-1)*noz*nox;
X[m][n] -= H[i];
}
return;
}
void return_horiz_ave(E,X,H)
struct All_variables *E;
double **X, *H;
{
const int dims = E->mesh.nsd;
int m,i,j,k,d,nint,noz,nox,noy,el,elz,elx,ely,j1,j2,i1,i2,k1,k2,nproc;
int top,lnode[5], sizeofH, noz2,iroot;
double *Have,*temp,aa[5];
struct Shape_function1 M;
struct Shape_function1_dA dGamma;
void get_global_1d_shape_fn();
sizeofH = (2*E->lmesh.noz+2)*sizeof(double);
Have = (double *)malloc(sizeofH);
temp = (double *)malloc(sizeofH);
noz = E->lmesh.noz;
noy = E->lmesh.noy;
elz = E->lmesh.elz;
elx = E->lmesh.elx;
ely = E->lmesh.ely;
noz2 = 2*noz;
for (i=1;i<=elz;i++) {
temp[i] = temp[i+noz] = 0.0;
temp[i+1] = temp[i+1+noz] = 0.0;
top = 0;
if (i==elz) top = 1;
for (m=1;m<=E->sphere.caps_per_proc;m++)
for (k=1;k<=ely;k++)
for (j=1;j<=elx;j++) {
el = i + (j-1)*elz + (k-1)*elx*elz;
get_global_1d_shape_fn(E,el,&M,&dGamma,top,m);
lnode[1] = E->ien[m][el].node[1];
lnode[2] = E->ien[m][el].node[2];
lnode[3] = E->ien[m][el].node[3];
lnode[4] = E->ien[m][el].node[4];
for(nint=1;nint<=onedvpoints[E->mesh.nsd];nint++) {
for(d=1;d<=onedvpoints[E->mesh.nsd];d++)
temp[i] += X[m][lnode[d]] * E->M.vpt[GMVINDEX(d,nint)]
* dGamma.vpt[GMVGAMMA(0,nint)];
temp[i+noz] += dGamma.vpt[GMVGAMMA(0,nint)];
}
if (i==elz) {
lnode[1] = E->ien[m][el].node[5];
lnode[2] = E->ien[m][el].node[6];
lnode[3] = E->ien[m][el].node[7];
lnode[4] = E->ien[m][el].node[8];
for(nint=1;nint<=onedvpoints[E->mesh.nsd];nint++) {
for(d=1;d<=onedvpoints[E->mesh.nsd];d++)
temp[i+1] += X[m][lnode[d]] * E->M.vpt[GMVINDEX(d,nint)]
* dGamma.vpt[GMVGAMMA(1,nint)];
temp[i+1+noz] += dGamma.vpt[GMVGAMMA(1,nint)];
}
} /* end of if i==elz */
} /* end of j and k, and m */
} /* Done for i */
MPI_Allreduce(temp,Have,noz2+1,MPI_DOUBLE,MPI_SUM,E->parallel.horizontal_comm);
for (i=1;i<=noz;i++) {
if(Have[i+noz] != 0.0)
H[i] = Have[i]/Have[i+noz];
}
/* if (E->parallel.me==0)
for(i=1;i<=noz;i++)
fprintf(stderr,"area %d %d %g\n",E->parallel.me,i,Have[i+noz]);
*/
free ((void *) Have);
free ((void *) temp);
return;
}
void return_horiz_ave_f(E,X,H)
struct All_variables *E;
float **X, *H;
{
const int dims = E->mesh.nsd;
int m,i,j,k,d,nint,noz,nox,noy,el,elz,elx,ely,j1,j2,i1,i2,k1,k2,nproc;
int top,lnode[5], sizeofH, noz2,iroot;
float *Have,*temp,aa[5];
struct Shape_function1 M;
struct Shape_function1_dA dGamma;
void get_global_1d_shape_fn();
sizeofH = (2*E->lmesh.noz+2)*sizeof(float);
Have = (float *)malloc(sizeofH);
temp = (float *)malloc(sizeofH);
noz = E->lmesh.noz;
noy = E->lmesh.noy;
elz = E->lmesh.elz;
elx = E->lmesh.elx;
ely = E->lmesh.ely;
noz2 = 2*noz;
for (i=1;i<=elz;i++) {
temp[i] = temp[i+noz] = 0.0;
temp[i+1] = temp[i+1+noz] = 0.0;
top = 0;
if (i==elz) top = 1;
for (m=1;m<=E->sphere.caps_per_proc;m++)
for (k=1;k<=ely;k++)
for (j=1;j<=elx;j++) {
el = i + (j-1)*elz + (k-1)*elx*elz;
get_global_1d_shape_fn(E,el,&M,&dGamma,top,m);
lnode[1] = E->ien[m][el].node[1];
lnode[2] = E->ien[m][el].node[2];
lnode[3] = E->ien[m][el].node[3];
lnode[4] = E->ien[m][el].node[4];
for(nint=1;nint<=onedvpoints[E->mesh.nsd];nint++) {
for(d=1;d<=onedvpoints[E->mesh.nsd];d++)
temp[i] += X[m][lnode[d]] * E->M.vpt[GMVINDEX(d,nint)]
* dGamma.vpt[GMVGAMMA(0,nint)];
temp[i+noz] += dGamma.vpt[GMVGAMMA(0,nint)];
}
if (i==elz) {
lnode[1] = E->ien[m][el].node[5];
lnode[2] = E->ien[m][el].node[6];
lnode[3] = E->ien[m][el].node[7];
lnode[4] = E->ien[m][el].node[8];
for(nint=1;nint<=onedvpoints[E->mesh.nsd];nint++) {
for(d=1;d<=onedvpoints[E->mesh.nsd];d++)
temp[i+1] += X[m][lnode[d]] * E->M.vpt[GMVINDEX(d,nint)]
* dGamma.vpt[GMVGAMMA(1,nint)];
temp[i+1+noz] += dGamma.vpt[GMVGAMMA(1,nint)];
}
} /* end of if i==elz */
} /* end of j and k, and m */
} /* Done for i */
MPI_Allreduce(temp,Have,noz2+1,MPI_FLOAT,MPI_SUM,E->parallel.horizontal_comm);
for (i=1;i<=noz;i++) {
if(Have[i+noz] != 0.0)
H[i] = Have[i]/Have[i+noz];
}
/* if (E->parallel.me==0)
for(i=1;i<=noz;i++)
fprintf(stderr,"area %d %d %g\n",E->parallel.me,i,Have[i+noz]);
*/
free ((void *) Have);
free ((void *) temp);
return;
}
/******* RETURN ELEMENTWISE HORIZ AVE ********************************/
/* */
/* This function is similar to return_horiz_ave in the citcom code */
/* however here, elemental horizontal averages are given rather than */
/* nodal averages. Also note, here is average per element */
void return_elementwise_horiz_ave(E,X,H)
struct All_variables *E;
double **X, *H;
{
int m,i,j,k,d,noz,noy,el,elz,elx,ely,nproc;
int sizeofH;
int elz2;
double *Have,*temp;
sizeofH = (2*E->lmesh.elz+2)*sizeof(double);
Have = (double *)malloc(sizeofH);
temp = (double *)malloc(sizeofH);
noz = E->lmesh.noz;
noy = E->lmesh.noy;
elz = E->lmesh.elz;
elx = E->lmesh.elx;
ely = E->lmesh.ely;
elz2 = 2*elz;
for (i=0;i<=(elz*2+1);i++)
{
temp[i]=0.0;
}
for (i=1;i<=elz;i++)
{
for (m=1;m<=E->sphere.caps_per_proc;m++)
{
for (k=1;k<=ely;k++)
{
for (j=1;j<=elx;j++)
{
el = i + (j-1)*elz + (k-1)*elx*elz;
temp[i] += X[m][el]*E->ECO[E->mesh.levmax][m][el].area;
temp[i+elz] += E->ECO[E->mesh.levmax][m][el].area;
}
}
}
}
/* determine which processors should get the message from me for
computing the layer averages */
MPI_Allreduce(temp,Have,elz2+1,MPI_DOUBLE,MPI_SUM,E->parallel.horizontal_comm);
for (i=1;i<=elz;i++) {
if(Have[i+elz] != 0.0)
H[i] = Have[i]/Have[i+elz];
}
free ((void *) Have);
free ((void *) temp);
return;
}
float return_bulk_value(E,Z,average)
struct All_variables *E;
float **Z;
int average;
{
void get_global_shape_fn();
void float_global_operation();
double rtf[4][9];
int n,i,j,k,el,m;
float volume,integral,volume1,integral1;
struct Shape_function GN;
struct Shape_function_dx GNx;
struct Shape_function_dA dOmega;
const int vpts = vpoints[E->mesh.nsd];
const int ends = enodes[E->mesh.nsd];
const int sphere_key=1;
volume1=0.0;
integral1=0.0;
for (m=1;m<=E->sphere.caps_per_proc;m++)
for (el=1;el<=E->lmesh.nel;el++) {
get_global_shape_fn(E,el,&GN,&GNx,&dOmega,0,sphere_key,rtf,E->mesh.levmax,m);
for(j=1;j<=vpts;j++)
for(i=1;i<=ends;i++) {
n = E->ien[m][el].node[i];
volume1 += E->N.vpt[GNVINDEX(i,j)] * dOmega.vpt[j];
integral1 += Z[m][n] * E->N.vpt[GNVINDEX(i,j)] * dOmega.vpt[j];
}
}
MPI_Allreduce(&volume1 ,&volume ,1,MPI_FLOAT,MPI_SUM,E->parallel.world);
MPI_Allreduce(&integral1,&integral,1,MPI_FLOAT,MPI_SUM,E->parallel.world);
if(average && volume != 0.0)
integral /= volume;
return((float)integral);
}
/************ RETURN BULK VALUE_D *****************************************/
/* */
/* Same as return_bulk_value but allowing double instead of float. */
/* I think when integer average =1, volume average is returned. */
/* when integer average =0, integral is returned. */
double return_bulk_value_d(E,Z,average)
struct All_variables *E;
double **Z;
int average;
{
void get_global_shape_fn();
double rtf[4][9];
int n,i,j,el,m;
double volume,integral,volume1,integral1;
struct Shape_function GN;
struct Shape_function_dx GNx;
struct Shape_function_dA dOmega;
const int vpts = vpoints[E->mesh.nsd];
const int ends = enodes[E->mesh.nsd];
const int sphere_key=1;
volume1=0.0;
integral1=0.0;
for (m=1;m<=E->sphere.caps_per_proc;m++)
for (el=1;el<=E->lmesh.nel;el++) {
get_global_shape_fn(E,el,&GN,&GNx,&dOmega,0,sphere_key,rtf,E->mesh.levmax,m);
for(j=1;j<=vpts;j++)
for(i=1;i<=ends;i++) {
n = E->ien[m][el].node[i];
volume1 += E->N.vpt[GNVINDEX(i,j)] * dOmega.vpt[j];
integral1 += Z[m][n] * E->N.vpt[GNVINDEX(i,j)] * dOmega.vpt[j];
}
}
MPI_Allreduce(&volume1 ,&volume ,1,MPI_DOUBLE,MPI_SUM,E->parallel.world);
MPI_Allreduce(&integral1,&integral,1,MPI_DOUBLE,MPI_SUM,E->parallel.world);
if(average && volume != 0.0)
integral /= volume;
return((double)integral);
}
/* ================================================== */
float find_max_horizontal(E,Tmax)
struct All_variables *E;
float Tmax;
{
float ttmax;
MPI_Allreduce(&Tmax,&ttmax,1,MPI_FLOAT,MPI_MAX,E->parallel.horizontal_comm);
return(ttmax);
}
/* ================================================== */
void sum_across_surface(E,data,total)
struct All_variables *E;
float *data;
int total;
{
int j,d;
float *temp;
temp = (float *)malloc((total+1)*sizeof(float));
MPI_Allreduce(data,temp,total,MPI_FLOAT,MPI_SUM,E->parallel.horizontal_comm);
for (j=0;j<total;j++) {
data[j] = temp[j];
}
free((void *)temp);
return;
}
/* ================================================== */
/* ================================================== */
/* ================================================== */
void sum_across_surf_sph1(E,sphc,sphs)
struct All_variables *E;
float *sphc,*sphs;
{
int jumpp,total,j,d;
float *sphcs,*temp;
temp = (float *) malloc((E->sphere.hindice*2+3)*sizeof(float));
sphcs = (float *) malloc((E->sphere.hindice*2+3)*sizeof(float));
/* pack */
jumpp = E->sphere.hindice;
total = E->sphere.hindice*2+3;
for (j=0;j<E->sphere.hindice;j++) {
sphcs[j] = sphc[j];
sphcs[j+jumpp] = sphs[j];
}
/* sum across processors in horizontal direction */
MPI_Allreduce(sphcs,temp,total,MPI_FLOAT,MPI_SUM,E->parallel.horizontal_comm);
/* unpack */
for (j=0;j<E->sphere.hindice;j++) {
sphc[j] = temp[j];
sphs[j] = temp[j+jumpp];
}
free((void *)temp);
free((void *)sphcs);
return;
}
/* ================================================== */
float global_fvdot(E,A,B,lev)
struct All_variables *E;
float **A,**B;
int lev;
{
int m,i,neq;
float prod, temp,temp1;
neq=E->lmesh.NEQ[lev];
temp = 0.0;
temp1 = 0.0;
prod = 0.0;
for (m=1;m<=E->sphere.caps_per_proc;m++) {
neq=E->lmesh.NEQ[lev];
temp1 = 0.0;
for (i=0;i<neq;i++)
temp += A[m][i]*B[m][i];
for (i=1;i<=E->parallel.Skip_neq[lev][m];i++)
temp1 += A[m][E->parallel.Skip_id[lev][m][i]]*B[m][E->parallel.Skip_id[lev][m][i]];
temp -= temp1;
}
MPI_Allreduce(&temp, &prod,1,MPI_FLOAT,MPI_SUM,E->parallel.world);
return (prod);
}
double kineticE_radial(E,A,lev)
struct All_variables *E;
double **A;
int lev;
{
int m,i,neq;
double prod, temp,temp1;
temp = 0.0;
prod = 0.0;
for (m=1;m<=E->sphere.caps_per_proc;m++) {
neq=E->lmesh.NEQ[lev];
temp1 = 0.0;
for (i=0;i<neq;i++)
if ((i+1)%3==0)
temp += A[m][i]*A[m][i];
for (i=1;i<=E->parallel.Skip_neq[lev][m];i++)
if ((E->parallel.Skip_id[lev][m][i]+1)%3==0)
temp1 += A[m][E->parallel.Skip_id[lev][m][i]]*A[m][E->parallel.Skip_id[lev][m][i]];
temp -= temp1;
}
MPI_Allreduce(&temp, &prod,1,MPI_DOUBLE,MPI_SUM,E->parallel.world);
return (prod);
}
double global_vdot(E,A,B,lev)
struct All_variables *E;
double **A,**B;
int lev;
{
int m,i,neq;
double prod, temp,temp1;
temp = 0.0;
prod = 0.0;
for (m=1;m<=E->sphere.caps_per_proc;m++) {
neq=E->lmesh.NEQ[lev];
temp1 = 0.0;
for (i=0;i<neq;i++)
temp += A[m][i]*B[m][i];
for (i=1;i<=E->parallel.Skip_neq[lev][m];i++)
temp1 += A[m][E->parallel.Skip_id[lev][m][i]]*B[m][E->parallel.Skip_id[lev][m][i]];
temp -= temp1;
}
MPI_Allreduce(&temp, &prod,1,MPI_DOUBLE,MPI_SUM,E->parallel.world);
return (prod);
}
double global_pdot(E,A,B,lev)
struct All_variables *E;
double **A,**B;
int lev;
{
int i,m,npno;
double prod, temp;
npno=E->lmesh.NPNO[lev];
temp = 0.0;
prod = 0.0;
for (m=1;m<=E->sphere.caps_per_proc;m++) {
npno=E->lmesh.NPNO[lev];
for (i=1;i<=npno;i++)
temp += A[m][i]*B[m][i];
}
MPI_Allreduce(&temp, &prod,1,MPI_DOUBLE,MPI_SUM,E->parallel.world);
return (prod);
}
double global_tdot_d(E,A,B,lev)
struct All_variables *E;
double **A,**B;
int lev;
{
int i,nno,m;
double prod, temp;
nno=E->lmesh.NNO[lev];
temp = 0.0;
prod = 0.0;
for (m=1;m<=E->sphere.caps_per_proc;m++) {
nno=E->lmesh.NNO[lev];
for (i=1;i<=nno;i++)
if (!(E->NODE[lev][m][i] & SKIP))
temp += A[m][i];
}
MPI_Allreduce(&temp, &prod,1,MPI_DOUBLE,MPI_SUM,E->parallel.world);
return (prod);
}
float global_tdot(E,A,B,lev)
struct All_variables *E;
float **A,**B;
int lev;
{
int i,nno,m;
float prod, temp;
temp = 0.0;
prod = 0.0;
for (m=1;m<=E->sphere.caps_per_proc;m++) {
nno=E->lmesh.NNO[lev];
for (i=1;i<=nno;i++)
if (!(E->NODE[lev][m][i] & SKIP))
temp += A[m][i]*B[m][i];
}
MPI_Allreduce(&temp, &prod,1,MPI_FLOAT,MPI_SUM,E->parallel.world);
return (prod);
}
float global_fmin(E,a)
struct All_variables *E;
float a;
{
float temp;
MPI_Allreduce(&a, &temp,1,MPI_FLOAT,MPI_MIN,E->parallel.world);
return (temp);
}
double global_dmax(E,a)
struct All_variables *E;
double a;
{
double temp;
MPI_Allreduce(&a, &temp,1,MPI_DOUBLE,MPI_MAX,E->parallel.world);
return (temp);
}
float global_fmax(E,a)
struct All_variables *E;
float a;
{
float temp;
MPI_Allreduce(&a, &temp,1,MPI_FLOAT,MPI_MAX,E->parallel.world);
return (temp);
}
double Tmaxd(E,T)
struct All_variables *E;
double **T;
{
double global_dmax(),temp,temp1;
int i,m;
temp = -10.0;
for (m=1;m<=E->sphere.caps_per_proc;m++)
for(i=1;i<=E->lmesh.nno;i++)
temp = max(T[m][i],temp);
temp1 = global_dmax(E,temp);
return (temp1);
}
float Tmax(E,T)
struct All_variables *E;
float **T;
{
float global_fmax(),temp,temp1;
int i,m;
temp = -10.0;
for (m=1;m<=E->sphere.caps_per_proc;m++)
for(i=1;i<=E->lmesh.nno;i++)
temp = max(T[m][i],temp);
temp1 = global_fmax(E,temp);
return (temp1);
}
float vnorm_nonnewt(E,dU,U,lev)
struct All_variables *E;
float **dU,**U;
int lev;
{
float temp1,temp2,dtemp,temp;
int a,e,i,m,node;
const int dims = E->mesh.nsd;
const int ends = enodes[dims];
const int nel=E->lmesh.nel;
dtemp=0.0;
temp=0.0;
for (m=1;m<=E->sphere.caps_per_proc;m++)
for (e=1;e<=nel;e++)
/*if (E->mat[m][e]==1)*/
for (i=1;i<=dims;i++)
for (a=1;a<=ends;a++) {
node = E->IEN[lev][m][e].node[a];
dtemp += dU[m][ E->ID[lev][m][node].doff[i] ]*
dU[m][ E->ID[lev][m][node].doff[i] ];
temp += U[m][ E->ID[lev][m][node].doff[i] ]*
U[m][ E->ID[lev][m][node].doff[i] ];
}
MPI_Allreduce(&dtemp, &temp2,1,MPI_FLOAT,MPI_SUM,E->parallel.world);
MPI_Allreduce(&temp, &temp1,1,MPI_FLOAT,MPI_SUM,E->parallel.world);
temp1 = sqrt(temp2/temp1);
return (temp1);
}
void sum_across_depth_sph1(E,sphc,sphs)
struct All_variables *E;
float *sphc,*sphs;
{
int jumpp,total,j;
float *sphcs,*temp;
if (E->parallel.nprocz > 1) {
total = E->sphere.hindice*2+3;
temp = (float *) malloc(total*sizeof(float));
sphcs = (float *) malloc(total*sizeof(float));
/* pack sphc[] and sphs[] into sphcs[] */
jumpp = E->sphere.hindice;
for (j=0;j<E->sphere.hindice;j++) {
sphcs[j] = sphc[j];
sphcs[j+jumpp] = sphs[j];
}
/* sum across processors in z direction */
MPI_Allreduce(sphcs, temp, total, MPI_FLOAT, MPI_SUM,
E->parallel.vertical_comm);
/* unpack */
for (j=0;j<E->sphere.hindice;j++) {
sphc[j] = temp[j];
sphs[j] = temp[j+jumpp];
}
free(temp);
free(sphcs);
}
return;
}
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...