int main(int argc, char* argv[])
{
bool verb; /* verbosity flag */
bool abc; /* absorbing boundary conditions flag */
bool free; /* free surface flag*/
bool snap; /* wavefield snapshots flag */
int jsnap;/* save wavefield every *jsnap* time steps */
/* cube axes */
sf_axis at,az,ax,as,ar,bt;
int it,iz,ix,is,ir, iop;
int nt,nz,nx,ns,nr,nz2,nx2;
float z0,x0,dt,dx,dz, idx,idz,dt2;
/* Laplacian */
int nop=2; /* Laplacian operator size */
float c0, c1, c2; /* Laplacian operator coefficients */
float co,c1x,c2x,c1z,c2z;
int nbz,nbx; /* boundary size */
float tz, tx; /* sponge boundary decay coefficients */
float dp;
float ws; /* injected data */
/* linear interpolation */
float *fzs,*fxs, *fzr,*fxr;
int *jzs,*jxs, *jzr,*jxr;
float *ws00,*ws01,*ws10,*ws11;
float *wr00,*wr01,*wr10,*wr11;
/* boundary */
float *bzl,*bzh,*bxl,*bxh;
/* I/O files */
sf_file Fw,Fs,Fr;
float *ww; /* wavelet */
pt2d *ss, *rr; /* source/receiver locations */
float **tt; /* taper */
/* background */
sf_file Bv,Bd,Bu; /* velocity, data, wavefield */
float **bvv,**bvo; /* velocity */
float *bdd; /* data */
float **bum,**buo,**bup,**bud; /* wavefields */
/* perturbation */
sf_file Pv,Pd,Pu;
float **pvv,**pvo;
float *pdd;
float **pum,**puo,**pup,**pud;
int ompchunk; /* OpenMP data chunk size */
/*------------------------------------------------------------*/
/* init RSF */
sf_init(argc,argv);
if(! sf_getint("ompchunk",&ompchunk)) ompchunk=1;
if(! sf_getbool("verb",&verb)) verb=false;
if(! sf_getbool( "abc",&abc )) abc=false;
if(! sf_getbool("snap",&snap)) snap=false;
if(! sf_getbool("free",&free)) free=false;
Fw = sf_input ("in" ); /* wavelet */
Fs = sf_input ("sou"); /* sources */
Fr = sf_input ("rec"); /* receivers */
Bv = sf_input ("vel"); /* velocity */
Bu = sf_output("wfl"); /* wavefield */
Bd = sf_output("out"); /* data */
Pv = sf_input ("ref"); /* velocity */
Pu = sf_output("liw"); /* linearized wavefield */
Pd = sf_output("lid"); /* linearized data */
/* read axes*/
at=sf_iaxa(Fw,1); sf_setlabel(at,"t");
nt=sf_n(at); dt=sf_d(at); if(verb) sf_raxa(at); /* time */
as=sf_iaxa(Fs,2); sf_setlabel(as,"s");
ns=sf_n(as); if(verb) sf_raxa(as); /* sources */
ar=sf_iaxa(Fr,2); sf_setlabel(ar,"r");
nr=sf_n(ar); if(verb) sf_raxa(ar); /* receivers */
az=sf_iaxa(Bv,1); sf_setlabel(az,"z");
nz=sf_n(az); dz=sf_d(az); if(verb) sf_raxa(az); /* depth */
ax=sf_iaxa(Bv,2); sf_setlabel(ax,"x");
nx=sf_n(ax); dx=sf_d(ax); if(verb) sf_raxa(ax); /* space */
/* configure wavefield snapshots */
if(snap) {
if(! sf_getint("jsnap",&jsnap)) jsnap=nt;
}
/*------------------------------------------------------------*/
/* expand domain for absorbing boundary conditions */
if(abc) {
if(! sf_getint("nbz",&nbz)) nbz=nop; if(nbz<nop) nbz=nop;
if(! sf_getint("nbx",&nbx)) nbx=nop; if(nbx<nop) nbx=nop;
if(! sf_getfloat("tz",&tz)) tz=0.025;
if(! sf_getfloat("tx",&tx)) tx=0.025;
} else {
nbz=nop;
nbx=nop;
}
/* expanded domain ( az+2 nz, ax+2 nx ) */
nz2=nz+2*nbz; z0=sf_o(az)-nbz*dz;
nx2=nx+2*nbx; x0=sf_o(ax)-nbx*dx;
sf_setn(az,nz2); sf_seto(az,z0); if(verb) sf_raxa(az);
sf_setn(ax,nx2); sf_seto(ax,x0); if(verb) sf_raxa(ax);
/*------------------------------------------------------------*/
/* setup output wavefield header */
if(snap) {
bt = sf_maxa(nt/jsnap,sf_o(at),dt*jsnap);
sf_setlabel(bt,"t");
sf_oaxa(Bu,az,1);
sf_oaxa(Bu,ax,2);
sf_oaxa(Bu,bt,3);
sf_oaxa(Pu,az,1);
sf_oaxa(Pu,ax,2);
sf_oaxa(Pu,bt,3);
}
/* setup output data header */
sf_oaxa(Bd,ar,1);
sf_oaxa(Bd,at,2);
sf_oaxa(Pd,ar,1);
sf_oaxa(Pd,at,2);
dt2 = dt*dt;
idz = 1/(dz*dz);
idx = 1/(dx*dx);
/* Laplacian coefficients */
c0=-30./12.;
c1=+16./12.;
c2=- 1./12.;
co = c0 * (idx+idz);
c1x= c1 * idx;
c2x= c2 * idx;
c1z= c1 * idz;
c2z= c2 * idz;
/*------------------------------------------------------------*/
/* allocate arrays */
ww=sf_floatalloc (nt); sf_floatread(ww ,nt ,Fw);
bvv=sf_floatalloc2(nz,nx); sf_floatread(bvv[0],nz*nx,Bv);
pvv=sf_floatalloc2(nz,nx); sf_floatread(pvv[0],nz*nx,Pv);
/* allocate source/receiver point arrays */
ss = (pt2d*) sf_alloc(ns,sizeof(*ss));
rr = (pt2d*) sf_alloc(nr,sizeof(*rr));
pt2dread1(Fs,ss,ns,3); /* read 3 elements (x,z,v) */
pt2dread1(Fr,rr,nr,2); /* read 2 elements (x,z) */
bdd=sf_floatalloc(nr);
pdd=sf_floatalloc(nr);
jzs=sf_intalloc(ns); fzs=sf_floatalloc(ns);
jzr=sf_intalloc(nr); fzr=sf_floatalloc(nr);
jxs=sf_intalloc(ns); fxs=sf_floatalloc(ns);
jxr=sf_intalloc(nr); fxr=sf_floatalloc(nr);
ws00 = sf_floatalloc(ns); wr00 = sf_floatalloc(nr);
ws01 = sf_floatalloc(ns); wr01 = sf_floatalloc(nr);
ws10 = sf_floatalloc(ns); wr10 = sf_floatalloc(nr);
ws11 = sf_floatalloc(ns); wr11 = sf_floatalloc(nr);
/*------------------------------------------------------------*/
for (is=0;is<ns;is++) {
if(ss[is].z >= z0 &&
ss[is].z < z0 + (nz2-1)*dz &&
ss[is].x >= x0 &&
ss[is].x < x0 + (nx2-1)*dx) {
jzs[is] = (int)( (ss[is].z-z0)/dz);
fzs[is] = (ss[is].z-z0)/dz - jzs[is];
jxs[is] = (int)( (ss[is].x-x0)/dx);
fxs[is] = (ss[is].x-x0)/dx - jxs[is];
} else {
jzs[is] = 0; jxs[is] = 0;
fzs[is] = 1; fxs[is] = 0;
ss[is].v= 0;
}
ws00[is] = (1-fzs[is])*(1-fxs[is]);
ws01[is] = ( fzs[is])*(1-fxs[is]);
ws10[is] = (1-fzs[is])*( fxs[is]);
ws11[is] = ( fzs[is])*( fxs[is]);
}
for (ir=0;ir<nr;ir++) {
if(rr[ir].z >= z0 &&
rr[ir].z < z0 + (nz2-1)*dz &&
rr[ir].x >= x0 &&
rr[ir].x < x0 + (nx2-1)*dx) {
jzr[ir] = (int)( (rr[ir].z-z0)/dz);
fzr[ir] = (rr[ir].z-z0)/dz - jzr[ir];
jxr[ir] = (int)( (rr[ir].x-x0)/dx);
fxr[ir] = (rr[ir].x-x0)/dx - jxr[ir];
rr[ir].v=1;
} else {
jzr[ir] = 0;
fzr[ir] = 1;
rr[ir].v= 0;
}
wr00[ir] = (1-fzr[ir])*(1-fxr[ir]);
wr01[ir] = ( fzr[ir])*(1-fxr[ir]);
wr10[ir] = (1-fzr[ir])*( fxr[ir]);
wr11[ir] = ( fzr[ir])*( fxr[ir]);
}
/*------------------------------------------------------------*/
/* allocate temporary arrays */
bum=sf_floatalloc2(nz2,nx2);
buo=sf_floatalloc2(nz2,nx2);
bup=sf_floatalloc2(nz2,nx2);
bud=sf_floatalloc2(nz2,nx2);
pum=sf_floatalloc2(nz2,nx2);
puo=sf_floatalloc2(nz2,nx2);
pup=sf_floatalloc2(nz2,nx2);
pud=sf_floatalloc2(nz2,nx2);
tt=sf_floatalloc2(nz2,nx2);
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(iz,ix) shared(nx2,nz2,bum,buo,bup,bud,pum,puo,pup,pud,tt)
#endif
for (iz=0; iz<nz2; iz++) {
for (ix=0; ix<nx2; ix++) {
bum[ix][iz]=pum[ix][iz]=0;
buo[ix][iz]=puo[ix][iz]=0;
bup[ix][iz]=pup[ix][iz]=0;
bud[ix][iz]=pud[ix][iz]=0;
tt[ix][iz]=1;
}
}
/*------------------------------------------------------------*/
/* velocity in the expanded domain (vo=vv^2)*/
bvo=sf_floatalloc2(nz2,nx2);
pvo=sf_floatalloc2(nz2,nx2);
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(iz,ix) shared(nz,nx,bvv,pvv,bvo,pvo)
#endif
for (iz=0; iz<nz; iz++) {
for (ix=0; ix<nx; ix++) {
bvo[nbx+ix][nbz+iz] = bvv[ix][iz] * bvv[ix][iz];
pvo[nbx+ix][nbz+iz] = pvv[ix][iz];
}
}
/* fill boundaries */
for (iz=0; iz<nbz; iz++) {
for (ix=0; ix<nx2; ix++) {
bvo[ix][ iz ] = bvo[ix][ nbz ];
bvo[ix][nz2-iz-1] = bvo[ix][nz2-nbz-1];
pvo[ix][ iz ] = pvo[ix][ nbz ];
pvo[ix][nz2-iz-1] = pvo[ix][nz2-nbz-1];
}
}
for (iz=0; iz<nz2; iz++) {
for (ix=0; ix<nbx; ix++) {
bvo[ ix ][iz] = bvo[ nbx ][iz];
bvo[nx2-ix-1][iz] = bvo[nx2-nbx-1][iz];
pvo[ ix ][iz] = pvo[ nbx ][iz];
pvo[nx2-ix-1][iz] = pvo[nx2-nbx-1][iz];
}
}
/*------------------------------------------------------------*/
/* free surface */
if(abc && free) {
for (iz=0; iz<nbz; iz++) {
for (ix=0; ix<nx2; ix++) {
bvo[ix][iz]=0;
pvo[ix][iz]=0;
}
}
}
/*------------------------------------------------------------*/
/* sponge ABC setup */
if(abc) {
for (iz=0; iz<nbz; iz++) {
for (ix=0; ix<nx2; ix++) {
tt[ix][ iz ] = exp( - (tz*(nbz-iz))*(tz*(nbz-iz)) );
tt[ix][nz2-iz-1] = tt[ix][iz];
}
}
for (iz=0; iz<nz2; iz++) {
for (ix=0; ix<nbx; ix++) {
tt[ ix ][iz] = exp( - (tx*(nbx-ix))*(tx*(nbx-ix)) );
tt[nx2-ix-1][iz] = tt[ix][iz];
}
}
}
/* one-way ABC setup */
bzl=sf_floatalloc(nx2);
bzh=sf_floatalloc(nx2);
bxl=sf_floatalloc(nz2);
bxh=sf_floatalloc(nz2);
for (ix=0;ix<nx2;ix++) {
dp = bvo[ix][ nop ] *dt/dz; bzl[ix] = (1-dp)/(1+dp);
dp = bvo[ix][nz2-nop-1] *dt/dz; bzh[ix] = (1-dp)/(1+dp);
}
for (iz=0;iz<nz2;iz++) {
dp = bvo[ nop ][iz] *dt/dx; bxl[iz] = (1-dp)/(1+dp);
dp = bvo[nx2-nop-1][iz] *dt/dx; bxh[iz] = (1-dp)/(1+dp);
}
/*------------------------------------------------------------*/
/*
* MAIN LOOP
*/
if(verb) fprintf(stderr,"\n");
for (it=0; it<nt; it++) {
if(verb) fprintf(stderr,"\b\b\b\b\b%d",it);
/* 4th order Laplacian operator */
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(iz,ix) shared(nop,nx2,nz2,bud,buo,pud,puo,co,c1x,c1z,c2x,c2z,idx,idz)
#endif
for (ix=nop; ix<nx2-nop; ix++) {
for (iz=nop; iz<nz2-nop; iz++) {
bud[ix][iz] =
co * buo[ix ][iz ] +
c1x*(buo[ix-1][iz ] + buo[ix+1][iz ]) +
c2x*(buo[ix-2][iz ] + buo[ix+2][iz ]) +
c1z*(buo[ix ][iz-1] + buo[ix ][iz+1]) +
c2z*(buo[ix ][iz-2] + buo[ix ][iz+2]);
pud[ix][iz] =
co * puo[ix ][iz ] +
c1x*(puo[ix-1][iz ] + puo[ix+1][iz ]) +
c2x*(puo[ix-2][iz ] + puo[ix+2][iz ]) +
c1z*(puo[ix ][iz-1] + puo[ix ][iz+1]) +
c2z*(puo[ix ][iz-2] + puo[ix ][iz+2]);
}
}
/* inject source */
for (is=0;is<ns;is++) {
ws = ww[it] * ss[is].v;
bud[ jxs[is] ][ jzs[is] ] -= ws * ws00[is];
bud[ jxs[is] ][ jzs[is]+1] -= ws * ws01[is];
bud[ jxs[is]+1][ jzs[is] ] -= ws * ws10[is];
bud[ jxs[is]+1][ jzs[is]+1] -= ws * ws11[is];
}
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(ix,iz) shared(nx2,nz2,pud,bud,pvo)
#endif
for (ix=0; ix<nx2; ix++) {
for (iz=0; iz<nz2; iz++) {
pud[ix][iz] -= bud[ix][iz] * 2*pvo[ix][iz];
}
}
/* velocity scale */
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(ix,iz) shared(nx2,nz2,bud,pud,bvo)
#endif
for (ix=0; ix<nx2; ix++) {
for (iz=0; iz<nz2; iz++) {
bud[ix][iz] *= bvo[ix][iz];
pud[ix][iz] *= bvo[ix][iz];
}
}
/* time step */
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(ix,iz) shared(nx2,nz2,bud,buo,bum,bup,pud,puo,pum,pup,dt2)
#endif
for (ix=0; ix<nx2; ix++) {
for (iz=0; iz<nz2; iz++) {
bup[ix][iz] = 2*buo[ix][iz] - bum[ix][iz] + bud[ix][iz] * dt2;
bum[ix][iz] = buo[ix][iz];
buo[ix][iz] = bup[ix][iz];
pup[ix][iz] = 2*puo[ix][iz] - pum[ix][iz] + pud[ix][iz] * dt2;
pum[ix][iz] = puo[ix][iz];
puo[ix][iz] = pup[ix][iz];
}
}
/* one-way ABC apply */
if(abc) {
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(ix,iz,iop) shared(nx2,nz2,nop,buo,bum,puo,pum,bzl,bzh)
#endif
for(ix=0;ix<nx2;ix++) {
for(iop=0;iop<nop;iop++) {
iz = nop-iop;
buo [ix][iz ]
= bum[ix][iz+1]
+(bum[ix][iz ]
- buo[ix][iz+1]) * bzl[ix];
puo [ix][iz ]
= pum[ix][iz+1]
+(pum[ix][iz ]
- puo[ix][iz+1]) * bzl[ix];
iz = nz2-nop+iop-1;
buo [ix][iz ]
= bum[ix][iz-1]
+(bum[ix][iz ]
- buo[ix][iz-1]) * bzh[ix];
puo [ix][iz ]
= pum[ix][iz-1]
+(pum[ix][iz ]
- puo[ix][iz-1]) * bzh[ix];
}
}
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,1) private(ix,iz,iop) shared(nx2,nz2,nop,buo,bum,puo,pum,bzl,bzh)
#endif
for(iop=0;iop<nop;iop++) {
for(iz=0;iz<nz2;iz++) {
ix = nop-iop;
buo [ix ][iz]
= bum[ix+1][iz]
+(bum[ix ][iz]
- buo[ix+1][iz]) * bxl[iz];
puo [ix ][iz]
= pum[ix+1][iz]
+(pum[ix ][iz]
- puo[ix+1][iz]) * bxl[iz];
ix = nx2-nop+iop-1;
buo [ix ][iz]
= bum[ix-1][iz]
+(bum[ix ][iz]
- buo[ix-1][iz]) * bxh[iz];
puo [ix ][iz]
= pum[ix-1][iz]
+(pum[ix ][iz]
- puo[ix-1][iz]) * bxh[iz];
}
}
}
/* sponge ABC apply */
if(abc) {
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(ix,iz) shared(nx2,nz2,buo,bum,bud,puo,pum,pud,tt)
#endif
for (ix=0; ix<nx2; ix++) {
for (iz=0; iz<nz2; iz++) {
buo[ix][iz] *= tt[ix][iz];
bum[ix][iz] *= tt[ix][iz];
bud[ix][iz] *= tt[ix][iz];
puo[ix][iz] *= tt[ix][iz];
pum[ix][iz] *= tt[ix][iz];
bud[ix][iz] *= tt[ix][iz];
}
}
}
/* write wavefield */
if(snap && it%jsnap==0) {
sf_floatwrite(buo[0],nz2*nx2,Bu);
sf_floatwrite(puo[0],nz2*nx2,Pu);
}
/* write data */
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,1) private(ir) shared(bdd,pdd,rr,buo,puo,jzr,wr00,wr01,wr10,wr11)
#endif
for (ir=0;ir<nr;ir++) {
bdd[ir] =
buo[ jxr[ir] ][ jzr[ir] ] * wr00[ir] +
buo[ jxr[ir] ][ jzr[ir]+1] * wr01[ir] +
buo[ jxr[ir]+1][ jzr[ir] ] * wr10[ir] +
buo[ jxr[ir]+1][ jzr[ir]+1] * wr11[ir];
bdd[ir] *= rr[ir].v;
pdd[ir] =
puo[ jxr[ir] ][ jzr[ir] ] * wr00[ir] +
puo[ jxr[ir] ][ jzr[ir]+1] * wr01[ir] +
puo[ jxr[ir]+1][ jzr[ir] ] * wr10[ir] +
puo[ jxr[ir]+1][ jzr[ir]+1] * wr11[ir];
pdd[ir] *= rr[ir].v;
}
/* write data */
sf_floatwrite(bdd,nr,Bd);
sf_floatwrite(pdd,nr,Pd);
}
if(verb) fprintf(stderr,"\n");
exit (0);
}
int
main(int argc, char** argv)
{
bool verb, fsrf, snap, expl, dabc, cden, adj;
bool optfd, hybrid, sinc;
int jsnap, jdata;
/* I/O files */
sf_file file_wav=NULL; /* wavelet */
sf_file file_vel=NULL; /* velocity */
sf_file file_den=NULL; /* density */
sf_file file_wfl=NULL; /* wavefield */
sf_file file_dat=NULL; /* data */
sf_file file_src=NULL; /* sources */
sf_file file_rec=NULL; /* receivers */
/* cube axes */
sf_axis at = NULL, az = NULL, ax = NULL, ay = NULL;
sf_axis as = NULL, ar = NULL;
int nbd; /* ABC boundary size */
int fdorder; /* finite difference spatial accuracy order */
int nzpad,nxpad,nypad; /* boundary padded model size */
int ix,iy,it,is,nx,ny,nz,nt,ns,nr;
float dx,dy,dz,dt,dt2;
float* damp=NULL; /* damping profile for hybrid bc */
float* ws; /* wavelet */
float*** vel=NULL; /* velocity */
float*** rho=NULL; /* density */
float*** u0=NULL; /* wavefield array u@t-1 (u@t+1) */
float*** u1=NULL; /* wavefield array u@t */
float* u_dat=NULL; /* output data */
float*** ptr_tmp=NULL;
pt3d* src3d=NULL; /* source position */
pt3d* rec3d=NULL; /*receiver position*/
scoef3d cssinc = NULL, crsinc = NULL;
lint3d cslint = NULL, crlint = NULL;
/* FDM structure */
fdm3d fdm = NULL;
abcone3d abc = NULL;
sponge spo = NULL;
int nbell;
float* fdcoef_d2;
float* fdcoef_d1;
sf_axis acz = NULL, acx = NULL, acy = NULL;
int nqz, nqx, nqy;
float oqz, oqx, oqy, dqz, dqx, dqy;
float** oslice = NULL; /* output 3D wavefield slice-by-slice */
float*** tmp_array;
double wall_clock_time_s, wall_clock_time_e;
const int SECOND_DERIV = 2;
const int FIRST_DERIV = 1;
int nop;
#if defined _OPENMP && _DEBUG
double tic;
double toc;
#endif
/* init RSF */
sf_init(argc,argv);
#ifdef _OPENMP
omp_init();
wall_clock_time_s = omp_get_wtime();
#else
wall_clock_time_s = (double) clock() / CLOCKS_PER_SEC;
#endif
if (!sf_getbool("verb",&verb)) verb=false; /* Verbosity flag */
if (!sf_getbool("snap",&snap)) snap=false; /* Wavefield snapshots flag */
if (!sf_getbool("expl",&expl)) expl=false; /* Multiple sources, one wvlt*/
if (!sf_getbool("dabc",&dabc)) dabc=false; /* Absorbing BC */
if (!sf_getbool("cden",&cden)) cden=false; /* Constant density */
if (!sf_getbool("adj",&adj)) adj=false; /* adjoint flag */
if (!sf_getbool("free",&fsrf) && !sf_getbool("fsrf",&fsrf)) fsrf=false; /* Free surface flag */
if (!sf_getint("nbell",&nbell)) nbell=5; /* gaussian for source injection */
if (!sf_getbool("optfd",&optfd)) optfd=false; /* optimized FD coefficients flag */
if (!sf_getint("fdorder",&fdorder)) fdorder=4; /* spatial FD order */
if (!sf_getbool("hybridbc",&hybrid)) hybrid=false; /* hybrid Absorbing BC */
if (!sf_getbool("sinc",&sinc)) sinc=false; /* sinc source injection */
/* Initialize variables */
file_wav = sf_input("in"); /* wavelet */
file_vel = sf_input("vel"); /* velocity */
file_src = sf_input("sou"); /* sources */
file_rec = sf_input("rec"); /* receivers */
file_dat = sf_output("out"); /* data */
if (snap) file_wfl = sf_output("wfl"); /* wavefield */
if (!cden) {
if (sf_getstring("cden")) {
file_den = sf_input ("den"); /* density */
} else {
cden = true;
if (verb) sf_warning("No density file provided, running with constant density");
}
}
at = sf_iaxa(file_wav,2); sf_setlabel(at,"t"); if(verb) sf_raxa(at); /* time */
az = sf_iaxa(file_vel,1); sf_setlabel(az,"z"); if(verb) sf_raxa(az); /* depth */
ax = sf_iaxa(file_vel,2); sf_setlabel(ax,"x"); if(verb) sf_raxa(ax); /* space */
ay = sf_iaxa(file_vel,3); sf_setlabel(ay,"y"); if(verb) sf_raxa(ay); /* space */
as = sf_iaxa(file_src,2); sf_setlabel(as,"s"); if(verb) sf_raxa(as); /* sources */
ar = sf_iaxa(file_rec,2); sf_setlabel(ar,"r"); if(verb) sf_raxa(ar); /* receivers */
nt = sf_n(at); dt = sf_d(at);
nz = sf_n(az); dz = sf_d(az);
nx = sf_n(ax); dx = sf_d(ax);
ny = sf_n(ay); dy = sf_d(ay);
ns = sf_n(as);
nr = sf_n(ar);
/* other execution parameters */
if (snap) {
if (!sf_getint("jsnap",&jsnap)) jsnap=nt;
/* # of t steps at which to save wavefield */
}
if (!sf_getint("jdata",&jdata)) jdata=1;
/* # of t steps at which to save receiver data */
/* setup output data header */
sf_oaxa(file_dat,ar,1);
sf_setn(at,(nt-1)/jdata+1);
sf_setd(at,dt*jdata);
sf_oaxa(file_dat,at,2);
/* wavefield cut params */
/* setup output wavefield header */
if (snap) {
if (!sf_getint ("nqz",&nqz)) nqz=sf_n(az); /* Saved wfld window nz */
if (!sf_getint ("nqx",&nqx)) nqx=sf_n(ax); /* Saved wfld window nx */
if (!sf_getint ("nqy",&nqy)) nqy=sf_n(ay); /* Saved wfld window ny */
if (!sf_getfloat("oqz",&oqz)) oqz=sf_o(az); /* Saved wfld window oz */
if (!sf_getfloat("oqx",&oqx)) oqx=sf_o(ax); /* Saved wfld window ox */
if (!sf_getfloat("oqy",&oqy)) oqy=sf_o(ay); /* Saved wfld window oy */
if (!sf_getfloat("dqz",&dqz)) dqz=sf_d(az); /* Saved wfld window dz */
if (!sf_getfloat("dqx",&dqx)) dqx=sf_d(ax); /* Saved wfld window dx */
if (!sf_getfloat("dqy",&dqy)) dqy=sf_d(ay); /* Saved wfld window dy */
acz = sf_maxa(nqz,oqz,dqz); if (verb) sf_raxa(acz);
acx = sf_maxa(nqx,oqx,dqx); if (verb) sf_raxa(acx);
acy = sf_maxa(nqy,oqy,dqy); if (verb) sf_raxa(acy);
/* check if the imaging window fits in the wavefield domain */
sf_setn(at,(nt-1)/jsnap+1);
sf_setd(at,dt*jsnap);
if (verb) sf_raxa(at);
sf_oaxa(file_wfl,acz,1);
sf_oaxa(file_wfl,acx,2);
sf_oaxa(file_wfl,acy,3);
sf_oaxa(file_wfl,at,4);
}
/* 2-2N finite difference coefficient */
nop = fdorder/2; /* fd half-length stencil */
if (!sf_getint("nb",&nbd) || nbd<nop) nbd=nop;
if (dabc && hybrid && nbd<=nop) nbd = 2*nop;
/* expand domain for FD operators and ABC */
fdm = fdutil3d_init(verb,fsrf,az,ax,ay,nbd,1);
sf_setn(az,fdm->nzpad); sf_seto(az,fdm->ozpad); if (verb) sf_raxa(az);
sf_setn(ax,fdm->nxpad); sf_seto(ax,fdm->oxpad); if (verb) sf_raxa(ax);
sf_setn(ay,fdm->nypad); sf_seto(ay,fdm->oypad); if (verb) sf_raxa(ay);
/* Precompute coefficients */
dt2 = dt*dt;
nzpad = nz+2*nbd; nxpad = nx+2*nbd; nypad = ny+2*nbd;
fdcoef_d2 = compute_fdcoef(nop,dz,dx,dy,optfd,SECOND_DERIV);
fdcoef_d1 = compute_fdcoef(nop,dz,dx,dy,optfd,FIRST_DERIV);
/* Allocate memories */
if (expl) ws = sf_floatalloc(1);
else ws = sf_floatalloc(ns);
vel = sf_floatalloc3(nzpad,nxpad,nypad);
if (!cden) rho = sf_floatalloc3(nzpad,nxpad,nypad);
u_dat = sf_floatalloc(nr);
src3d = pt3dalloc1(ns);
rec3d = pt3dalloc1(nr);
if (snap) oslice = sf_floatalloc2(sf_n(acz),sf_n(acx));
/* source and receiver position */
pt3dread1(file_src,src3d,ns,3); /* read format: (x,y,z) */
if (sinc) cssinc = sinc3d_make(ns,src3d,fdm);
else cslint = lint3d_make(ns,src3d,fdm);
pt3dread1(file_rec,rec3d,nr,3); /* read format: (x,y,z) */
if (sinc) crsinc = sinc3d_make(nr,rec3d,fdm);
else crlint = lint3d_make(nr,rec3d,fdm);
if (!sinc) fdbell3d_init(nbell);
/* temperary array */
tmp_array = sf_floatalloc3(nz,nx,ny);
/* read velocity and pad */
sf_floatread(tmp_array[0][0],nz*nx*ny,file_vel);
expand3d(tmp_array,vel,fdm);
/* read density and pad */
if (!cden) {
sf_floatread(tmp_array[0][0],nz*nx*ny,file_den);
expand3d(tmp_array,rho,fdm);
}
free(**tmp_array); free(*tmp_array); free(tmp_array);
/* A1 one-way ABC implicit scheme coefficients */
if (dabc) {
abc = abcone3d_make(nbd,dt,vel,fsrf,fdm);
if (hybrid)
damp = damp_make(nbd-nop); /* compute damping profiles for hybrid bc */
else
spo = sponge_make(fdm->nb);
}
/* allocate memory for wavefield variables */
u0 = sf_floatalloc3(nzpad,nxpad,nypad);
u1 = sf_floatalloc3(nzpad,nxpad,nypad);
/* initialize variables */
memset(u0[0][0],0,sizeof(float)*nzpad*nxpad*nypad);
memset(u1[0][0],0,sizeof(float)*nzpad*nxpad*nypad);
memset(u_dat,0,sizeof(float)*nr);
/* v = (v*dt)^2 */
for (ix=0;ix<nzpad*nxpad*nypad;ix++)
*(vel[0][0]+ix) *= *(vel[0][0]+ix)*dt2;
if (fsrf && !hybrid) {
for (iy=0; iy<nypad; iy++)
for (ix=0; ix<nxpad; ix++)
memset(vel[iy][ix],0,sizeof(float)*fdm->nb);
}
for (it=0; it<nt; it++) {
if (verb) sf_warning("it=%d;",it+1);
#if defined _OPENMP && _DEBUG
tic=omp_get_wtime();
#endif
step_forward(u0,u1,vel,rho,fdcoef_d2,fdcoef_d1,nop,nzpad,nxpad,nypad);
if (adj) { /* backward inject source wavelet */
if (expl) {
sf_seek(file_wav,(off_t)(nt-it-1)*sizeof(float),SEEK_SET);
sf_floatread(ws,1,file_wav);
ws[0] *= dt2;
if (sinc) sinc3d_inject1(u0,ws[0],cssinc);
else lint3d_inject1(u0,ws[0],cslint);
} else {
sf_seek(file_wav,(off_t)(nt-it-1)*ns*sizeof(float),SEEK_SET);
sf_floatread(ws,ns,file_wav);
for (is=0; is<ns; is++) ws[is] *= dt2;
if (sinc) sinc3d_inject(u0,ws,cssinc);
else lint3d_inject(u0,ws,cslint);
}
} else { /* forward inject source wavelet */
if (expl) {
sf_floatread(ws,1,file_wav);
ws[0] *= dt2;
if (sinc) sinc3d_inject1(u0,ws[0],cssinc);
else lint3d_inject1(u0,ws[0],cslint);
} else {
sf_floatread(ws,ns,file_wav);
for (is=0; is<ns; is++) ws[is] *= dt2;
if (sinc) sinc3d_inject(u0,ws,cssinc);
else lint3d_inject(u0,ws,cslint);
}
}
/* apply abc */
if (dabc) {
if (hybrid) apply_abc(u0,u1,nz,nx,ny,nbd,abc,nop,damp);
else {
abcone3d_apply(u0,u1,nop,abc,fdm);
sponge3d_apply(u0,spo,fdm);
sponge3d_apply(u1,spo,fdm);
}
}
/* loop over pointers */
ptr_tmp = u0; u0 = u1; u1 = ptr_tmp;
/* extract snapshot */
if (snap && it%jsnap==0) {
int fy = (floor)((sf_o(acy)-fdm->oypad)/fdm->dy);
int jy = floor(sf_d(acy)/fdm->dy);
float **ptr_slice;
for (iy=0; iy<sf_n(acy); iy++) {
ptr_slice = u0[fy+iy*jy];
cut3d_slice(ptr_slice,oslice,fdm,acz,acx);
sf_floatwrite(oslice[0],sf_n(acz)*sf_n(acx),file_wfl);
}
}
/* extract receiver data */
if (sinc) sinc3d_extract(u0,u_dat,crsinc);
else lint3d_extract(u0,u_dat,crlint);
sf_floatwrite(u_dat,nr,file_dat);
#if defined _OPENMP && _DEBUG
toc=omp_get_wtime();
fprintf(stderr,"%5.2gs",(float)(toc-tic));
#endif
}
#ifdef _OPENMP
wall_clock_time_e = omp_get_wtime();
#else
wall_clock_time_e = (double) clock() / CLOCKS_PER_SEC;
#endif
if (verb)
fprintf(stderr,"\nElapsed time: %lf s\n",wall_clock_time_e-wall_clock_time_s);
free(**u0); free(*u0); free(u0);
free(**u1); free(*u1); free(u1);
free(**vel); free(*vel); free(vel);
free(u_dat);
free(ws);
free(fdcoef_d2); free(fdcoef_d1);
if (snap) { free(*oslice); free(oslice); }
if(!cden) { free(**rho); free(*rho); free(rho); }
if (hybrid) free(damp);
free(src3d); free(rec3d);
return 0;
}
int main(int argc, char* argv[])
{
bool verb,fsrf,snap,expl;
int jsnap,ntsnap;
int jdata;
/* I/O files */
sf_file Fwav=NULL; /* wavelet */
sf_file Fsou=NULL; /* sources */
sf_file Frec=NULL; /* receivers */
sf_file Fvel=NULL; /* velocity */
sf_file Fref=NULL; /* reflectivity */
sf_file Fden=NULL; /* density */
sf_file Fdat=NULL; /* data (background) */
sf_file Fwfl=NULL; /* wavefield (background) */
sf_file Flid=NULL; /* data (scattered) */
sf_file Fliw=NULL; /* wavefield (scattered) */
/* I/O arrays */
float *ww=NULL; /* wavelet */
pt2d *ss=NULL; /* sources */
pt2d *rr=NULL; /* receivers */
float **vpin=NULL; /* velocity */
float **roin=NULL; /* density */
float **rfin=NULL; /* reflectivity */
float **vp=NULL; /* velocity in expanded domain */
float **ro=NULL; /* density in expanded domain */
float **ro1=NULL; /* normalized 1st derivative of density on axis 1 */
float **ro2=NULL; /* normalized 1st derivative of density on axis 2 */
float **rf=NULL; /* reflectivity in expanded domain */
float *bdd=NULL; /* data (background) */
float *sdd=NULL; /* data (scattered) */
float **vt=NULL; /* temporary vp*vp * dt*dt */
float **bum,**buo,**bup,**bua,**but; /* wavefield: um = U @ t-1; uo = U @ t; up = U @ t+1 */
float **sum,**suo,**sup,**sua,**sut; /* wavefield: um = U @ t-1; uo = U @ t; up = U @ t+1 */
/* cube axes */
sf_axis at,a1,a2,as,ar;
int nt,n1,n2,ns,nr,nb;
int it,i1,i2;
float dt,d1,d2,id1,id2,dt2;
/* linear interpolation weights/indices */
lint2d cs,cr;
fdm2d fdm;
abcone2d abc; /* abc */
sponge spo;
/* FD operator size */
float co,ca2,cb2,ca1,cb1;
int ompchunk;
#ifdef _OPENMP
int ompnth,ompath;
#endif
sf_axis ac1=NULL,ac2=NULL;
int nqz,nqx;
float oqz,oqx;
float dqz,dqx;
float **uc=NULL;
/*------------------------------------------------------------*/
/* init RSF */
sf_init(argc,argv);
if(! sf_getint("ompchunk",&ompchunk)) ompchunk=1;
/* OpenMP data chunk size */
#ifdef _OPENMP
if(! sf_getint("ompnth", &ompnth)) ompnth=0;
/* OpenMP available threads */
#pragma omp parallel
ompath=omp_get_num_threads();
if(ompnth<1) ompnth=ompath;
omp_set_num_threads(ompnth);
sf_warning("using %d threads of a total of %d",ompnth,ompath);
#endif
if(! sf_getbool("verb",&verb)) verb=false; /* verbosity flag */
if(! sf_getbool("snap",&snap)) snap=false; /* wavefield snapshots flag */
if(! sf_getbool("free",&fsrf)) fsrf=false; /* free surface flag */
if(! sf_getbool("expl",&expl)) expl=false; /* "exploding reflector" */
Fwav = sf_input ("in" ); /* wavelet */
Fsou = sf_input ("sou"); /* sources */
Frec = sf_input ("rec"); /* receivers */
Fvel = sf_input ("vel"); /* velocity */
Fden = sf_input ("den"); /* density */
Fref = sf_input ("ref"); /* reflectivity */
Fwfl = sf_output("wfl"); /* wavefield */
Fdat = sf_output("out"); /* data */
Fliw = sf_output("liw"); /* wavefield (scattered) */
Flid = sf_output("lid"); /* data (scattered) */
/* axes */
at = sf_iaxa(Fwav,2); sf_setlabel(at,"t"); if(verb) sf_raxa(at); /* time */
as = sf_iaxa(Fsou,2); sf_setlabel(as,"s"); if(verb) sf_raxa(as); /* sources */
ar = sf_iaxa(Frec,2); sf_setlabel(ar,"r"); if(verb) sf_raxa(ar); /* receivers */
a1 = sf_iaxa(Fvel,1); sf_setlabel(a1,"z"); if(verb) sf_raxa(a1); /* depth */
a2 = sf_iaxa(Fvel,2); sf_setlabel(a2,"x"); if(verb) sf_raxa(a2); /* space */
nt = sf_n(at); dt = sf_d(at);
ns = sf_n(as);
nr = sf_n(ar);
n1 = sf_n(a1); d1 = sf_d(a1);
n2 = sf_n(a2); d2 = sf_d(a2);
if(! sf_getint("jdata",&jdata)) jdata=1;
if(snap) { /* save wavefield every *jsnap* time steps */
if(! sf_getint("jsnap",&jsnap)) jsnap=nt;
}
/*------------------------------------------------------------*/
/* setup output data header */
sf_oaxa(Fdat,ar,1);
sf_oaxa(Flid,ar,1);
sf_setn(at,nt/jdata);
sf_setd(at,dt*jdata);
sf_oaxa(Fdat,at,2);
sf_oaxa(Flid,at,2);
/* setup output wavefield header */
if(snap) {
if(!sf_getint ("nqz",&nqz)) nqz=sf_n(a1);
if(!sf_getint ("nqx",&nqx)) nqx=sf_n(a2);
if(!sf_getfloat("oqz",&oqz)) oqz=sf_o(a1);
if(!sf_getfloat("oqx",&oqx)) oqx=sf_o(a2);
dqz=sf_d(a1);
dqx=sf_d(a2);
ac1 = sf_maxa(nqz,oqz,dqz);
ac2 = sf_maxa(nqx,oqx,dqx);
/* check if the imaging window fits in the wavefield domain */
uc=sf_floatalloc2(sf_n(ac1),sf_n(ac2));
ntsnap=0;
for(it=0; it<nt; it++) {
if(it%jsnap==0) ntsnap++;
}
sf_setn(at, ntsnap);
sf_setd(at,dt*jsnap);
if(verb) sf_raxa(at);
/* sf_setn(at,nt/jsnap);
sf_setd(at,dt*jsnap); */
sf_oaxa(Fwfl,ac1,1);
sf_oaxa(Fwfl,ac2,2);
sf_oaxa(Fwfl,at, 3);
sf_oaxa(Fliw,ac1,1);
sf_oaxa(Fliw,ac2,2);
sf_oaxa(Fliw,at, 3);
}
/*------------------------------------------------------------*/
/* expand domain for FD operators and ABC */
if( !sf_getint("nb",&nb) || nb<NOP) nb=NOP;
fdm=fdutil_init(verb,fsrf,a1,a2,nb,ompchunk);
sf_setn(a1,fdm->nzpad); sf_seto(a1,fdm->ozpad); if(verb) sf_raxa(a1);
sf_setn(a2,fdm->nxpad); sf_seto(a2,fdm->oxpad); if(verb) sf_raxa(a2);
/*------------------------------------------------------------*/
if(expl) ww = sf_floatalloc( 1);
else ww = sf_floatalloc(ns);
bdd =sf_floatalloc(nr);
sdd =sf_floatalloc(nr);
/*------------------------------------------------------------*/
/* setup source/receiver coordinates */
ss = (pt2d*) sf_alloc(ns,sizeof(*ss));
rr = (pt2d*) sf_alloc(nr,sizeof(*rr));
pt2dread1(Fsou,ss,ns,2); /* read (x,z) coordinates */
pt2dread1(Frec,rr,nr,2); /* read (x,z) coordinates */
cs = lint2d_make(ns,ss,fdm);
cr = lint2d_make(nr,rr,fdm);
/*------------------------------------------------------------*/
/* setup FD coefficients */
dt2 = dt*dt;
id1 = 1/d1;
id2 = 1/d2;
co = C0 * (id2*id2+id1*id1);
ca2= CA * id2*id2;
cb2= CB * id2*id2;
ca1= CA * id1*id1;
cb1= CB * id1*id1;
/*------------------------------------------------------------*/
/* input density */
roin=sf_floatalloc2(n1, n2 );
ro =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
ro1 =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
ro2 =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sf_floatread(roin[0],n1*n2,Fden);
expand(roin,ro,fdm);
/* normalized density derivatives */
for (i2=NOP; i2<fdm->nxpad-NOP; i2++) {
for(i1=NOP; i1<fdm->nzpad-NOP; i1++) {
ro1[i2][i1] = D1(ro,i2,i1,id1) / ro[i2][i1];
ro2[i2][i1] = D2(ro,i2,i1,id2) / ro[i2][i1];
}
}
free(*roin); free(roin);
/*------------------------------------------------------------*/
/* input velocity */
vpin=sf_floatalloc2(n1, n2 );
vp =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
vt =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sf_floatread(vpin[0],n1*n2,Fvel);
expand(vpin,vp,fdm);
free(*vpin); free(vpin);
/*------------------------------------------------------------*/
/* input reflectivity */
rfin=sf_floatalloc2(n1, n2 );
rf =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sf_floatread(rfin[0],n1*n2,Fref);
expand(rfin,rf,fdm);
free(*rfin); free(rfin);
for (i2=0; i2<fdm->nxpad; i2++) {
for(i1=0; i1<fdm->nzpad; i1++) {
vt[i2][i1] = vp[i2][i1] * vp[i2][i1] * dt2;
}
}
/* free surface */
if(fsrf) {
for (i2=0; i2<fdm->nxpad; i2++) {
for(i1=0; i1<fdm->nb; i1++) {
vt[i2][i1]=0;
}
}
}
/*------------------------------------------------------------*/
/* allocate wavefield arrays */
bum=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
buo=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
bup=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
bua=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sum=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
suo=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sup=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sua=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
for (i2=0; i2<fdm->nxpad; i2++) {
for(i1=0; i1<fdm->nzpad; i1++) {
bum[i2][i1]=0;
buo[i2][i1]=0;
bup[i2][i1]=0;
bua[i2][i1]=0;
sum[i2][i1]=0;
suo[i2][i1]=0;
sup[i2][i1]=0;
sua[i2][i1]=0;
}
}
/*------------------------------------------------------------*/
/* one-way abc setup */
abc = abcone2d_make(NOP,dt,vp,fsrf,fdm);
/* sponge abc setup */
spo = sponge_make(fdm->nb);
/*------------------------------------------------------------*/
/*
* MAIN LOOP
*/
/*------------------------------------------------------------*/
if(verb) fprintf(stderr,"\n");
for (it=0; it<nt; it++) {
if(verb) fprintf(stderr,"\b\b\b\b\b%d",it);
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,fdm->ompchunk) private(i2,i1) shared(fdm,bua,buo,sua,suo,co,ca2,ca1,cb2,cb1,id2,id1)
#endif
for (i2=NOP; i2<fdm->nxpad-NOP; i2++) {
for(i1=NOP; i1<fdm->nzpad-NOP; i1++) {
/* 4th order Laplacian operator */
bua[i2][i1] =
co * buo[i2 ][i1 ] +
ca2*(buo[i2-1][i1 ] + buo[i2+1][i1 ]) +
cb2*(buo[i2-2][i1 ] + buo[i2+2][i1 ]) +
ca1*(buo[i2 ][i1-1] + buo[i2 ][i1+1]) +
cb1*(buo[i2 ][i1-2] + buo[i2 ][i1+2]);
sua[i2][i1] =
co * suo[i2 ][i1 ] +
ca2*(suo[i2-1][i1 ] + suo[i2+1][i1 ]) +
cb2*(suo[i2-2][i1 ] + suo[i2+2][i1 ]) +
ca1*(suo[i2 ][i1-1] + suo[i2 ][i1+1]) +
cb1*(suo[i2 ][i1-2] + suo[i2 ][i1+2]);
/* density term */
bua[i2][i1] -= (
D1(buo,i2,i1,id1) * ro1[i2][i1] +
D2(buo,i2,i1,id2) * ro2[i2][i1] );
sua[i2][i1] -= (
D1(suo,i2,i1,id1) * ro1[i2][i1] +
D2(suo,i2,i1,id2) * ro2[i2][i1] );
}
}
/* inject acceleration source */
if(expl) {
sf_floatread(ww, 1,Fwav);
lint2d_inject1(bua,ww[0],cs);
} else {
sf_floatread(ww,ns,Fwav);
lint2d_inject(bua,ww,cs);
}
/* single scattering */
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(i2,i1) shared(fdm,buo,sua,rf)
#endif
for (i2=0; i2<fdm->nxpad; i2++) {
for (i1=0; i1<fdm->nzpad; i1++) {
sua[i2][i1] -= bua[i2][i1] * 2*rf[i2][i1];
}
}
/* step forward in time */
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,fdm->ompchunk) private(i2,i1) shared(fdm,bua,buo,bum,bup,sua,suo,sum,sup,vt,dt2)
#endif
for (i2=0; i2<fdm->nxpad; i2++) {
for(i1=0; i1<fdm->nzpad; i1++) {
bup[i2][i1] = 2*buo[i2][i1]
- bum[i2][i1]
+ bua[i2][i1] * vt[i2][i1];
sup[i2][i1] = 2*suo[i2][i1]
- sum[i2][i1]
+ sua[i2][i1] * vt[i2][i1];
}
}
/* circulate wavefield arrays */
but=bum;
bum=buo;
buo=bup;
bup=but;
sut=sum;
sum=suo;
suo=sup;
sup=sut;
/* one-way abc apply*/
abcone2d_apply(buo,bum,NOP,abc,fdm);
sponge2d_apply(bum, spo,fdm);
sponge2d_apply(buo, spo,fdm);
abcone2d_apply(suo,sum,NOP,abc,fdm);
sponge2d_apply(sum, spo,fdm);
sponge2d_apply(suo, spo,fdm);
/* extract data at receivers */
lint2d_extract(buo,bdd,cr);
lint2d_extract(suo,sdd,cr);
if( it%jdata==0) {
sf_floatwrite(bdd,nr,Fdat);
sf_floatwrite(sdd,nr,Flid);
}
/* extract wavefield in the "box" */
if(snap && it%jsnap==0) {
cut2d(buo,uc,fdm,ac1,ac2);
sf_floatwrite(uc[0],sf_n(ac1)*sf_n(ac2),Fwfl);
cut2d(suo,uc,fdm,ac1,ac2);
sf_floatwrite(uc[0],sf_n(ac1)*sf_n(ac2),Fliw);
}
}
if(verb) fprintf(stderr,"\n");
exit (0);
}
int main(int argc, char *argv[])
{
clock_t tstart, tend;
double duration;
int numprocs, rank;
float *sendbuf, *recvbuf;
MPI_Comm Comm=MPI_COMM_WORLD;
bool verb, wantrecord, wantwf, onlyrecord;
sf_file Ffvel, Ffden, Fbvel, Fbden;
sf_file Fsrc, Frcd, Fimg1, Fimg2;
sf_file FGx, FGz, Fsxx, Fsxz, Fszx, Fszz;
sf_file Ftmpfwf, Ftmpbwf;
sf_axis at, ax, az, atau;
int shtbgn, shtinv, shtnmb, shtpad, shtnmb0;
int snapturn, tmpint;
float **fvel, **bvel;
float ***fwf, ***record, **localrec;
float ***img1, **img2, ***mig1, **mig2;
float *tmpsxx, *tmpsxz, *tmpszx, *tmpszz;
sf_init(argc, argv);
MPI_Init(&argc, &argv);
MPI_Comm_size(Comm, &numprocs);
MPI_Comm_rank(Comm, &rank);
tstart=clock();
if(rank==0) sf_warning("numprocs=%d", numprocs);
if(!sf_getbool("verb", &verb)) verb=true;
if(!sf_getbool("wantrecord", &wantrecord)) wantrecord=false;
if(!sf_getbool("wantwf", &wantwf)) wantwf=false;
if(!sf_getbool("onlyrecord", &onlyrecord)) onlyrecord=false;
Fsrc=sf_input("-input");
Fimg1=sf_output("-output");
Fimg2=sf_output("img2");
Ffvel=sf_input("fvel");
Ffden=sf_input("fden");
Fbvel=sf_input("bvel");
Fbden=sf_input("bden");
if(wantrecord)
Frcd=sf_input("record");
else
Frcd=sf_output("record");
if(wantwf){
Ftmpfwf=sf_output("tmpfwf");
Ftmpbwf=sf_output("tmpbwf");
}
FGx=sf_input("Gx");
FGz=sf_input("Gz");
Fsxx=sf_input("sxx");
Fsxz=sf_input("sxz");
Fszx=sf_input("szx");
Fszz=sf_input("szz");
at=sf_iaxa(Fsrc, 1); nt=sf_n(at); dt=sf_d(at);
if(!sf_getbool("srcdecay", &srcdecay)) srcdecay=true;
if(!sf_getint("srcrange", &srcrange)) srcrange=3;
if(!sf_getfloat("srctrunc", &srctrunc)) srctrunc=0.2;
if(!sf_getfloat("srcalpha", &srcalpha)) srcalpha=0.5;
wavelet=sf_floatalloc(nt);
sf_floatread(wavelet, nt, Fsrc);
if(!sf_getint("pmlsize", &pmlsize)) pmlsize=30;
if(!sf_getint("nfd", &nfd)) sf_error("Need half of the FD order!");
if(!sf_getfloat("pmld0", &pmld0)) pmld0=200;
if(!sf_getint("shtnmb", &shtnmb)) sf_error("Need shot number!");
if(!sf_getint("shtinv", &shtinv)) sf_error("Need shot interval!");
if(!sf_getint("shtbgn", &shtbgn)) shtbgn=0;
shtpad=numprocs-shtnmb%numprocs;
shtnmb0=shtnmb+shtpad;
az=sf_iaxa(Ffvel, 1); nzb=sf_n(az);
ax=sf_iaxa(Ffvel, 2); nxb=sf_n(ax);
nxzb=nxb*nzb;
nz=nzb-2*nfd-2*pmlsize;
nx=nxb-2*nfd-2*pmlsize;
if(!sf_getint("snapturn", &snapturn)) snapturn=1;
if(!sf_getint("ginv", &ginv)) ginv=1;
if(!sf_getint("wfinv", &wfinv)) wfinv=1;
if(!sf_getint("spz", &spz)) spz=6;
if(!sf_getint("gp", &gp)) gp=0;
ng=(nx-1)/ginv+1;
wfnt=(nt-1)/wfinv+1;
wfdt=dt*wfinv;
if(!sf_getint("ntau", &ntau)) ntau=1;
if(!sf_getfloat("dtau", &dtau)) dtau=wfdt;
if(!sf_getfloat("tau0", &tau0)) tau0=0;
atau=sf_iaxa(Fsrc, 1);
sf_setn(atau, ntau);
sf_setd(atau, dtau);
sf_seto(atau, tau0);
if(!sf_histint(FGx, "n1", &nxz)) sf_error("No n1= in FGx!");
if(nxz != nxzb) sf_error("Dimension error!");
if(!sf_histint(FGx, "n2", &lenx)) sf_error("No n2= in FGx!");
if(!sf_histint(FGz, "n2", &lenz)) sf_error("No n2= in FGz!");
Gx=sf_floatalloc3(nzb, nxb, lenx);
Gz=sf_floatalloc3(nzb, nxb, lenz);
sxx=sf_intalloc(lenx);
sxz=sf_intalloc(lenx);
szx=sf_intalloc(lenz);
szz=sf_intalloc(lenz);
tmpsxx=sf_floatalloc(lenx);
tmpsxz=sf_floatalloc(lenx);
tmpszx=sf_floatalloc(lenz);
tmpszz=sf_floatalloc(lenz);
sf_floatread(Gx[0][0], nxzb*lenx, FGx);
sf_floatread(Gz[0][0], nxzb*lenz, FGz);
sf_floatread(tmpsxx, lenx, Fsxx);
sf_floatread(tmpsxz, lenx, Fsxz);
sf_floatread(tmpszx, lenz, Fszx);
sf_floatread(tmpszz, lenz, Fszz);
for (ix=0; ix<lenx; ix++){
sxx[ix]=(int)tmpsxx[ix];
sxz[ix]=(int)tmpsxz[ix];
}
for (iz=0; iz<lenz; iz++){
szx[iz]=(int)tmpszx[iz];
szz[iz]=(int)tmpszz[iz];
}
fvel=sf_floatalloc2(nzb, nxb);
fden=sf_floatalloc2(nzb, nxb);
fc11=sf_floatalloc2(nzb, nxb);
bvel=sf_floatalloc2(nzb, nxb);
bden=sf_floatalloc2(nzb, nxb);
bc11=sf_floatalloc2(nzb, nxb);
sf_floatread(fvel[0], nxzb, Ffvel);
sf_floatread(fden[0], nxzb, Ffden);
sf_floatread(bvel[0], nxzb, Fbvel);
sf_floatread(bden[0], nxzb, Fbden);
for (ix=0; ix<nxb; ix++){
for (iz=0; iz<nzb; iz++){
fc11[ix][iz]=fden[ix][iz]*fvel[ix][iz]*fvel[ix][iz];
bc11[ix][iz]=bden[ix][iz]*bvel[ix][iz]*bvel[ix][iz];
}
}
if(wantrecord){
/* check record data */
sf_histint(Frcd, "n1", &tmpint);
if(tmpint != nt) sf_error("Not matched dimensions!");
sf_histint(Frcd, "n2", &tmpint);
if(tmpint != ng) sf_error("Not matched dimensions!");
sf_histint(Frcd, "n3", &tmpint);
if(tmpint != shtnmb) sf_error("Not matched dimensions!");
}
if(rank==0){
record=sf_floatalloc3(nt, ng, shtnmb0);
if(wantrecord){
sf_floatread(record[0][0], nt*ng*shtnmb, Frcd);
for(is=shtnmb; is<shtnmb0; is++)
for(ix=0; ix<ng; ix++)
for(it=0; it<nt; it++)
record[is][ix][it]=0.0;
}
}
img1=sf_floatalloc3(nz, nx, ntau);
mig1=sf_floatalloc3(nz, nx, ntau);
img2=sf_floatalloc2(nz, nx);
mig2=sf_floatalloc2(nz, nx);
zero3(img1, nz, nx, ntau);
zero2(img2, nz, nx);
sf_setn(az, nz);
sf_setn(ax, ng);
if(!wantrecord){
sf_oaxa(Frcd, at, 1);
sf_oaxa(Frcd, ax, 2);
sf_putint(Frcd, "n3", shtnmb);
sf_putint(Frcd, "d3", shtinv);
sf_putint(Frcd, "o3", shtbgn);
}
sf_setn(ax, nx);
if(wantwf){
sf_setn(at, wfnt);
sf_setd(at, wfdt);
sf_oaxa(Ftmpfwf, az, 1);
sf_oaxa(Ftmpfwf, ax, 2);
sf_oaxa(Ftmpfwf, at, 3);
sf_oaxa(Ftmpbwf, az, 1);
sf_oaxa(Ftmpbwf, ax, 2);
sf_oaxa(Ftmpbwf, at, 3);
}
sf_oaxa(Fimg1, az, 1);
sf_oaxa(Fimg1, ax, 2);
sf_oaxa(Fimg1, atau, 3);
sf_oaxa(Fimg2, az, 1);
sf_oaxa(Fimg2, ax, 2);
fwf=sf_floatalloc3(nz, nx, wfnt);
localrec=sf_floatalloc2(nt, ng);
if(verb){
sf_warning("==================================");
sf_warning("nx=%d nz=%d nt=%d", nx, nz, nt);
sf_warning("wfnt=%d wfdt=%f wfinv=%d dt=%f", wfnt, wfdt, wfinv, dt);
sf_warning("nxb=%d nzb=%d pmlsize=%d nfd=%d", nxb, nzb, pmlsize, nfd);
sf_warning("ntau=%d dtau=%f tau0=%f", ntau, dtau, tau0);
sf_warning("shtnmb=%d shtbgn=%d shtinv=%d", shtnmb, shtbgn, shtinv);
sf_warning("lenx=%d lenz=%d spz=%d gp=%d", lenx, lenz, spz, gp);
sf_warning("==================================");
}
init();
for(iturn=0; iturn*numprocs<shtnmb; iturn++){
is=iturn*numprocs+rank;
if(is<shtnmb){
sf_warning("ishot/nshot: %d/%d", is+1, shtnmb);
spx=is*shtinv+shtbgn;
sglfdfor2(fwf, localrec, verb);
}
if(wantrecord){
recvbuf=localrec[0];
if(rank==0) sendbuf=record[iturn*numprocs][0];
else sendbuf=NULL;
MPI_Scatter(sendbuf, ng*nt, MPI_FLOAT, recvbuf, ng*nt, MPI_FLOAT, 0, Comm);
}else{
sendbuf=localrec[0];
if(rank==0) recvbuf=record[iturn*numprocs][0];
else recvbuf=NULL;
MPI_Gather(sendbuf, ng*nt, MPI_FLOAT, recvbuf, ng*nt, MPI_FLOAT, 0, Comm);
}
if(wantwf && rank==0 && iturn==snapturn-1) wantwf=true;
else wantwf=false;
if(wantwf) sf_floatwrite(fwf[0][0], wfnt*nx*nz, Ftmpfwf);
if(!onlyrecord && is<shtnmb){
sglfdback2(mig1, mig2, fwf, localrec, verb, wantwf, Ftmpbwf);
for(itau=0; itau<ntau; itau++){
for(ix=0; ix<nx; ix++){
for(iz=0; iz<nz; iz++){
img1[itau][ix][iz]+=mig1[itau][ix][iz];
}
}
}
for(ix=0; ix<nx; ix++){
for(iz=0; iz<nz; iz++){
img2[ix][iz]+=mig2[ix][iz];
}
}
}
MPI_Barrier(Comm);
} //end of iturn
if(!onlyrecord){
if(rank==0){
sendbuf=(float *)MPI_IN_PLACE;
recvbuf=img1[0][0];
}else{
sendbuf=img1[0][0];
recvbuf=NULL;
}
MPI_Reduce(sendbuf, recvbuf, ntau*nx*nz, MPI_FLOAT, MPI_SUM, 0, Comm);
if(rank==0){
sendbuf=MPI_IN_PLACE;
recvbuf=img2[0];
}else{
sendbuf=img2[0];
recvbuf=NULL;
}
MPI_Reduce(sendbuf, recvbuf, nx*nz, MPI_FLOAT, MPI_SUM, 0, Comm);
}
if(rank==0){
if(!wantrecord){
sf_floatwrite(record[0][0], shtnmb*ng*nt, Frcd);
}
sf_floatwrite(img1[0][0], ntau*nx*nz, Fimg1);
sf_floatwrite(img2[0], nx*nz, Fimg2);
}
tend=clock();
duration=(double)(tend-tstart)/CLOCKS_PER_SEC;
sf_warning(">>The CPU time of sfmpilfdrtm2 is: %f seconds<<", duration);
MPI_Finalize();
exit(0);
}
int main (int argc, char* argv[])
{
bool verb, fsrf, snap, expl, dabc;
int jsnap, ntsnap, jdata;
/* I/O files */
sf_file Fwav = NULL; /* wavelet */
sf_file Fsou = NULL; /* sources */
sf_file Frec = NULL; /* receivers */
sf_file Fvel = NULL; /* velocity */
sf_file Fden = NULL; /* density */
sf_file Fdat = NULL; /* data */
sf_file Fwfl = NULL; /* wavefield */
/* cube axes */
sf_axis at, az, ax;
sf_axis as, ar;
int cpuid;
int nt, nz, nx, ns, nr;
int it, ia;
float dt, dz, dx;
/* FDM structure */
/* Wee keep these structures for compatibility,
as soon as PETSC version is finished,
this will be removed */
fdm2d fdm = NULL;
/* I/O arrays */
float *ww = NULL; /* wavelet */
pt2d *ss = NULL; /* sources */
pt2d *rr = NULL; /* receivers */
float *dd = NULL; /* data */
float **u,**v;
/* linear interpolation weights/indices */
lint2d cs,cr;
/* wavefield cut params */
sf_axis acz = NULL, acx = NULL;
int nqz, nqx;
float oqz, oqx;
float dqz, dqx;
float **uc = NULL;
sf_petsc_aimplfd2 aimplfd;
PetscErrorCode ierr;
/* PETSc Initialization */
ierr = PetscInitialize (&argc, &argv, 0, 0); CHKERRQ(ierr);
MPI_Comm_rank (MPI_COMM_WORLD, &cpuid);
/*------------------------------------------------------------*/
/* init RSF */
sf_init (argc, argv);
/*------------------------------------------------------------*/
if (!sf_getbool ("verb", &verb)) verb = false; /* verbosity flag */
if (!sf_getbool ("snap", &snap)) snap = false; /* wavefield snapshots flag */
if (!sf_getbool ("free", &fsrf)) fsrf = false; /* free surface flag */
if (!sf_getbool ("expl", &expl)) expl = false; /* "exploding reflector" */
if (!sf_getbool ("dabc", &dabc)) dabc = false; /* absorbing BC */
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* I/O files */
Fwav = sf_input ("in" ); /* wavelet */
Fvel = sf_input ("vel"); /* velocity */
Fsou = sf_input ("sou"); /* sources */
Frec = sf_input ("rec"); /* receivers */
Fwfl = sf_output("wfl"); /* wavefield */
Fdat = sf_output("out"); /* data */
Fden = sf_input ("den"); /* density */
/*------------------------------------------------------------*/
/* axes */
at = sf_iaxa (Fwav,2); sf_setlabel (at,"t"); if (verb && 0 == cpuid) sf_raxa (at); /* time */
az = sf_iaxa (Fvel,1); sf_setlabel (az,"z"); if (verb && 0 == cpuid) sf_raxa (az); /* depth */
ax = sf_iaxa (Fvel,2); sf_setlabel (ax,"x"); if (verb && 0 == cpuid) sf_raxa (ax); /* space */
as = sf_iaxa (Fsou, 2); sf_setlabel (as, "s"); if (verb && 0 == cpuid) sf_raxa (as); /* sources */
ar = sf_iaxa (Frec, 2); sf_setlabel (ar, "r"); if (verb && 0 == cpuid) sf_raxa (ar); /* receivers */
nt = sf_n (at); dt = sf_d (at);
nz = sf_n (az); dz = sf_d (az);
nx = sf_n (ax); dx = sf_d (ax);
ns = sf_n (as);
nr = sf_n (ar);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* other execution parameters */
if (!sf_getint ("jdata", &jdata)) jdata = 1;
if (snap) { /* save wavefield every *jsnap* time steps */
if (!sf_getint ("jsnap", &jsnap)) jsnap = nt;
}
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* expand domain for FD operators and ABC */
/*if (!sf_getint("nb",&nb) || nb<NOP) nb=NOP;*/
fdm = fdutil_init (verb, true, az, ax, 0, 1);
/*------------------------------------------------------------*/
if (0 == cpuid) {
sf_setn (az, fdm->nzpad); sf_seto (az, fdm->ozpad); if(verb) sf_raxa (az);
sf_setn (ax, fdm->nxpad); sf_seto (ax, fdm->oxpad); if(verb) sf_raxa (ax);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* setup output data header */
sf_oaxa (Fdat, ar, 1);
sf_setn (at, nt/jdata);
sf_setd (at, dt*jdata);
sf_oaxa (Fdat, at, 2);
}
/* setup output wavefield header */
if (snap && 0 == cpuid) {
if (!sf_getint ("nqz", &nqz)) nqz = sf_n (az);
if (!sf_getint ("nqx", &nqx)) nqx = sf_n (ax);
if (!sf_getfloat ("oqz", &oqz)) oqz = sf_o (az);
if (!sf_getfloat ("oqx", &oqx)) oqx = sf_o (ax);
dqz = sf_d (az);
dqx = sf_d (ax);
acz = sf_maxa (nqz, oqz, dqz); sf_raxa (acz);
acx = sf_maxa (nqx, oqx, dqx); sf_raxa (acx);
/* check if the imaging window fits in the wavefield domain */
uc = sf_floatalloc2 (sf_n (acz), sf_n (acx));
ntsnap = 0;
for (it = 0; it < nt; it++) {
if (it % jsnap == 0)
ntsnap++;
}
sf_setn (at, ntsnap);
sf_setd (at, dt*jsnap);
if (verb)
sf_raxa(at);
sf_oaxa (Fwfl, acz, 1);
sf_oaxa (Fwfl, acx, 2);
sf_oaxa (Fwfl, at, 3);
}
if (expl) {
ww = sf_floatalloc (1);
} else {
ww = sf_floatalloc (ns);
}
dd = sf_floatalloc (nr);
/*------------------------------------------------------------*/
/* setup source/receiver coordinates */
ss = (pt2d*) sf_alloc (ns, sizeof (*ss));
rr = (pt2d*) sf_alloc (nr, sizeof (*rr));
pt2dread1 (Fsou, ss, ns, 2); /* read (x,z) coordinates */
pt2dread1 (Frec, rr, nr, 2); /* read (x,z) coordinates */
cs = lint2d_make (ns, ss, fdm);
cr = lint2d_make (nr, rr, fdm);
v = sf_floatalloc2 (nz, nx);
u = sf_floatalloc2 (nz, nx);
/* input velocity */
sf_floatread (v[0], nz*nx, Fvel);
/*------------------------------------------------------------*/
PetscFPrintf (MPI_COMM_WORLD, stderr, "Initializing GMRES solver\n");
aimplfd = sf_petsc_aimplfd2_init (nz, nx, dz, dx, dt, &v[0][0], 100, true);
/*------------------------------------------------------------*/
/*
* MAIN LOOP
*/
/*------------------------------------------------------------*/
for (it = 0; it < nt; it++) {
PetscFPrintf (MPI_COMM_WORLD, stderr, "Timestep #%d, t=%f\n", it, it*dt);
sf_petsc_aimplfd2_next_step (aimplfd);
/* inject acceleration source */
if (expl) {
sf_floatread (ww, 1, Fwav);
for (ia = 0; ia < cs->n; ia++) {
sf_petsc_aimplfd2_add_source_ut1 (aimplfd, ww[0], cs->jz[ia], cs->jx[ia]);
}
} else {
sf_floatread (ww, ns, Fwav);
for (ia = 0; ia < cs->n; ia++) {
/*
PetscFPrintf (MPI_COMM_WORLD, stderr, "Source #%d [%d, %d], f=%f\n", ia, cs->jz[ia], cs->jx[ia], ww[0]);
*/
sf_petsc_aimplfd2_add_source_ut1 (aimplfd, ww[ia], cs->jz[ia], cs->jx[ia]);
}
}
sf_petsc_aimplfd2_get_wavefield_ut2 (aimplfd, &u[0][0]);
/* extract data */
/*
lint2d_extract (u, dd, cr);
*/
for (ia = 0; ia < cr->n; ia++) {
dd[ia] = u[cr->jx[ia]][cr->jz[ia]];
}
if (snap && it % jsnap == 0 && 0 == cpuid) {
cut2d (u, uc, fdm, acz, acx);
sf_floatwrite (uc[0], sf_n(acz)*sf_n(acx), Fwfl);
}
if (it % jdata == 0 && 0 == cpuid)
sf_floatwrite (dd, nr, Fdat);
}
exit (0);
}
int main(int argc, char* argv[])
{
/*------------------------------------------------------------*/
/* Execution control, I/O files and geometry */
/*------------------------------------------------------------*/
bool verb,fsrf,snap,back,esou; /* execution flags */
int jsnap,ntsnap,jdata; /* jump along axes */
int shft; /* time shift for wavefield matching in RTM */
/* I/O files */
sf_file Fwav=NULL; /* wavelet */
sf_file Fdat=NULL; /* data */
sf_file Fsou=NULL; /* sources */
sf_file Frec=NULL; /* receivers */
sf_file Fccc=NULL; /* velocity */
sf_file Frkp=NULL; /* app. rank */
sf_file Fltp=NULL; /* left mat */
sf_file Frtp=NULL; /* right mat */
sf_file Fwfp=NULL; /* wavefield */
sf_file Frks=NULL; /* app. rank */
sf_file Flts=NULL; /* left mat */
sf_file Frts=NULL; /* right mat */
sf_file Fwfs=NULL; /* wavefield */
/* cube axes */
sf_axis at,ax,ay,az; /* time, x, y, z */
sf_axis asx,asy,arx,ary,ac; /* sou, rec-x, rec-y, component */
/* dimension, index and interval */
int nt,nz,nx,ny,ns,nr,nc,nb;
int it,iz,ix,iy;
float dt,dz,dx,dy;
int nxyz, nk;
/* FDM and KSP structure */ //!!!JS
fdm3d fdm=NULL;
dft3d dft=NULL;
clr3d clr_p=NULL, clr_s=NULL;
/* I/O arrays for sou & rec */
sf_complex***ww=NULL; /* wavelet */
pt3d *ss=NULL; /* sources */
pt3d *rr=NULL; /* receivers */
sf_complex **dd=NULL; /* data */
/*------------------------------------------------------------*/
/* displacement: uo = U @ t; up = U @ t+1 */
/*------------------------------------------------------------*/
sf_complex ***uox, ***uoy, ***uoz, **uo;
sf_complex ***uox_p, ***uoy_p, ***uoz_p, **uo_p;
sf_complex ***uox_s, ***uoy_s, ***uoz_s, **uo_s;
/*sf_complex ***upx, ***upy, ***upz, **up;*/
/*------------------------------------------------------------*/
/* lowrank decomposition arrays */
/*------------------------------------------------------------*/
int ntmp, *n2s_p, *n2s_s;
sf_complex **lt_p, **rt_p, **lt_s, **rt_s;
/*------------------------------------------------------------*/
/* linear interpolation weights/indices */
/*------------------------------------------------------------*/
lint3d cs,cr; /* for injecting source and extracting data */
/* Gaussian bell */
int nbell;
/*------------------------------------------------------------*/
/* wavefield cut params */
/*------------------------------------------------------------*/
sf_axis acz=NULL,acx=NULL,acy=NULL;
int nqz,nqx,nqy;
float oqz,oqx,oqy;
float dqz,dqx,dqy;
sf_complex***uc=NULL; /* tmp array for output wavefield snaps */
/*------------------------------------------------------------*/
/* init RSF */
/*------------------------------------------------------------*/
sf_init(argc,argv);
/*------------------------------------------------------------*/
/* OMP parameters */
/*------------------------------------------------------------*/
#ifdef _OPENMP
omp_init();
#endif
/*------------------------------------------------------------*/
/* read execution flags */
/*------------------------------------------------------------*/
if(! sf_getbool("verb",&verb)) verb=false; /* verbosity flag */
if(! sf_getbool("snap",&snap)) snap=false; /* wavefield snapshots flag */
if(! sf_getbool("free",&fsrf)) fsrf=false; /* free surface flag */
if(! sf_getbool("back",&back)) back=false; /* backward extrapolation flag (for rtm) */
if(! sf_getbool("esou",&esou)) esou=false; /* explosive force source */
/*------------------------------------------------------------*/
/* I/O files */
/*------------------------------------------------------------*/
Fwav = sf_input ("in" ); /* wavelet */
Fdat = sf_output("out"); /* data */
Fsou = sf_input ("sou"); /* sources */
Frec = sf_input ("rec"); /* receivers */
Fccc = sf_input ("ccc"); /* stiffness */
Frkp = sf_input ("rkp"); /* app. rank */
Fltp = sf_input ("ltp"); /* left mat */
Frtp = sf_input ("rtp"); /* right mat */
Fwfp = sf_output("wfp"); /* wavefield */
Frks = sf_input ("rks"); /* app. rank */
Flts = sf_input ("lts"); /* left mat */
Frts = sf_input ("rts"); /* right mat */
Fwfs = sf_output("wfs"); /* wavefield */
/*------------------------------------------------------------*/
/* axes */
/*------------------------------------------------------------*/
at = sf_iaxa(Fwav,4); sf_setlabel(at,"t"); if(verb) sf_raxa(at); /* time */
az = sf_iaxa(Fccc,1); sf_setlabel(az,"z"); if(verb) sf_raxa(az); /* depth */
ax = sf_iaxa(Fccc,2); sf_setlabel(ax,"x"); if(verb) sf_raxa(ax); /* space x */
ay = sf_iaxa(Fccc,3); sf_setlabel(ay,"y"); if(verb) sf_raxa(ay); /* space y */
asx = sf_iaxa(Fsou,2); sf_setlabel(asx,"sx"); if(verb) sf_raxa(asx); /* sources x */
asy = sf_iaxa(Fsou,3); sf_setlabel(asy,"sy"); if(verb) sf_raxa(asy); /* sources y */
arx = sf_iaxa(Frec,2); sf_setlabel(arx,"rx"); if(verb) sf_raxa(arx); /* receivers x */
ary = sf_iaxa(Frec,3); sf_setlabel(ary,"ry"); if(verb) sf_raxa(ary); /* receivers y */
nt = sf_n(at); dt = sf_d(at);
nz = sf_n(az); dz = sf_d(az);
nx = sf_n(ax); dx = sf_d(ax);
ny = sf_n(ay); dy = sf_d(ay);
ns = sf_n(asx)*sf_n(asy);
nr = sf_n(arx)*sf_n(ary);
/*------------------------------------------------------------*/
/* other execution parameters */
/*------------------------------------------------------------*/
if(! sf_getint("nbell",&nbell)) nbell=NOP; /* bell size */
if(verb) sf_warning("nbell=%d",nbell);
if(! sf_getint("jdata",&jdata)) jdata=1;
if(snap) { /* save wavefield every *jsnap* time steps */
if(! sf_getint("jsnap",&jsnap)) jsnap=nt;
}
if(back) {
shft = (nt-1)%jsnap;
sf_warning("For backward extrapolation, make sure nbell(%d)=0",nbell);
} else shft = 0;
/*------------------------------------------------------------*/
/* expand domain for FD operators and ABC */
/*------------------------------------------------------------*/
if( !sf_getint("nb",&nb)) nb=NOP;
fdm=fdutil3d_init(verb,fsrf,az,ax,ay,nb,1);
if(nbell) fdbell3d_init(nbell);
sf_setn(az,fdm->nzpad); sf_seto(az,fdm->ozpad); if(verb) sf_raxa(az);
sf_setn(ax,fdm->nxpad); sf_seto(ax,fdm->oxpad); if(verb) sf_raxa(ax);
sf_setn(ay,fdm->nypad); sf_seto(ay,fdm->oypad); if(verb) sf_raxa(ay);
/*------------------------------------------------------------*/
/* 3D vector components */
/*------------------------------------------------------------*/
nc=3;
ac=sf_maxa(nc,0,1); /* output 3 cartesian components */
/*------------------------------------------------------------*/
/* setup output data header */
/*------------------------------------------------------------*/
sf_settype(Fdat,SF_COMPLEX);
sf_oaxa(Fdat,arx,1);
sf_oaxa(Fdat,ary,2);
sf_oaxa(Fdat,ac,3);
sf_setn(at,nt/jdata);
sf_setd(at,dt*jdata);
sf_oaxa(Fdat,at,4);
/* setup output wavefield header */
if(snap) {
if(!sf_getint ("nqz",&nqz)) nqz=sf_n(az);
if(!sf_getint ("nqx",&nqx)) nqx=sf_n(ax);
if(!sf_getint ("nqy",&nqy)) nqy=sf_n(ay);
if(!sf_getfloat("oqz",&oqz)) oqz=sf_o(az);
if(!sf_getfloat("oqx",&oqx)) oqx=sf_o(ax);
if(!sf_getfloat("oqy",&oqy)) oqy=sf_o(ay);
dqz=sf_d(az);
dqx=sf_d(ax);
dqy=sf_d(ay);
acz = sf_maxa(nqz,oqz,dqz); sf_raxa(acz);
acx = sf_maxa(nqx,oqx,dqx); sf_raxa(acx);
acy = sf_maxa(nqy,oqy,dqy); sf_raxa(acy);
uc=sf_complexalloc3(sf_n(acz),sf_n(acx),sf_n(acy));
ntsnap=0; /* ntsnap = it/jsnap+1; */
for(it=0; it<nt; it++) {
if(it%jsnap==0) ntsnap++;
}
sf_setn(at, ntsnap);
sf_setd(at,dt*jsnap);
if(verb) sf_raxa(at);
sf_settype(Fwfp,SF_COMPLEX);
sf_oaxa(Fwfp,acz,1);
sf_oaxa(Fwfp,acx,2);
sf_oaxa(Fwfp,acy,3);
sf_oaxa(Fwfp,ac, 4);
sf_oaxa(Fwfp,at, 5);
sf_settype(Fwfs,SF_COMPLEX);
sf_oaxa(Fwfs,acz,1);
sf_oaxa(Fwfs,acx,2);
sf_oaxa(Fwfs,acy,3);
sf_oaxa(Fwfs,ac, 4);
sf_oaxa(Fwfs,at, 5);
}
/*------------------------------------------------------------*/
/* source and data array */
/*------------------------------------------------------------*/
ww=sf_complexalloc3(ns,nc,nt); /* Fast axis: n_sou > n_comp > n_time */
sf_complexread(ww[0][0],nt*nc*ns,Fwav);
dd=sf_complexalloc2(nr,nc);
/*------------------------------------------------------------*/
/* setup source/receiver coordinates */
/*------------------------------------------------------------*/
ss = (pt3d*) sf_alloc(ns,sizeof(*ss));
rr = (pt3d*) sf_alloc(nr,sizeof(*rr));
pt3dread1(Fsou,ss,ns,3); /* read (x,y,z) coordinates */
pt3dread1(Frec,rr,nr,3); /* read (x,y,z) coordinates */
/* calculate 3d linear interpolation coef for sou & rec */
cs = lint3d_make(ns,ss,fdm);
cr = lint3d_make(nr,rr,fdm);
/*------------------------------------------------------------*/
/* allocate and initialize wavefield arrays */
/*------------------------------------------------------------*/
/* z-component */
uoz=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uoz_p=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uoz_s=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
/*upz=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);*/
/* x-component */
uox=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uox_p=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uox_s=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
/*upx=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);*/
/* y-component */
uoy=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uoy_p=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uoy_s=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
/*upy=sf_complexalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);*/
/* wavefield vector */
uo = (sf_complex**) sf_alloc(3,sizeof(sf_complex*));
uo[0] = uox[0][0]; uo[1] = uoy[0][0]; uo[2] = uoz[0][0];
uo_p = (sf_complex**) sf_alloc(3,sizeof(sf_complex*));
uo_p[0] = uox_p[0][0]; uo_p[1] = uoy_p[0][0]; uo_p[2] = uoz_p[0][0];
uo_s = (sf_complex**) sf_alloc(3,sizeof(sf_complex*));
uo_s[0] = uox_s[0][0]; uo_s[1] = uoy_s[0][0]; uo_s[2] = uoz_s[0][0];
/*up = (sf_complex**) sf_alloc(3,sizeof(sf_complex*));*/
/*up[0] = upx[0][0]; up[1] = upy[0][0]; up[2] = upz[0][0];*/
/* initialize fft and lrk */
dft = dft3d_init(1,false,false,fdm);
nxyz= fdm->nypad*fdm->nxpad*fdm->nzpad;
nk = dft->nky *dft->nkx *dft->nkz;
/*------------------------------------------------------------*/
/* allocation I/O arrays */
/*------------------------------------------------------------*/
n2s_p = sf_intalloc(9);
sf_intread(n2s_p,9,Frkp);
clr_p = clr3d_make2(n2s_p,fdm);
n2s_s = sf_intalloc(9);
sf_intread(n2s_s,9,Frks);
clr_s = clr3d_make2(n2s_s,fdm);
if (clr_p->n2_max > clr_s->n2_max) clr3d_init(fdm,dft,clr_p);
else clr3d_init(fdm,dft,clr_s);
/* check the dimension */
if (!sf_histint(Fltp,"n1",&ntmp) || ntmp != nxyz) sf_error("Need n1=%d in left",nxyz);
if (!sf_histint(Fltp,"n2",&ntmp) || ntmp != clr_p->n2_sum) sf_error("Need n2=%d in left",clr_p->n2_sum);
if (!sf_histint(Frtp,"n1",&ntmp) || ntmp != nk) sf_error("Need n1=%d in right",nk);
if (!sf_histint(Frtp,"n2",&ntmp) || ntmp != clr_p->n2_sum) sf_error("Need n2=%d in right",clr_p->n2_sum);
lt_p = sf_complexalloc2(nxyz,clr_p->n2_sum);
rt_p = sf_complexalloc2(nk ,clr_p->n2_sum);
sf_complexread(lt_p[0],nxyz*clr_p->n2_sum,Fltp);
sf_complexread(rt_p[0],nk *clr_p->n2_sum,Frtp);
if (!sf_histint(Flts,"n1",&ntmp) || ntmp != nxyz) sf_error("Need n1=%d in left",nxyz);
if (!sf_histint(Flts,"n2",&ntmp) || ntmp != clr_s->n2_sum) sf_error("Need n2=%d in left",clr_s->n2_sum);
if (!sf_histint(Frts,"n1",&ntmp) || ntmp != nk) sf_error("Need n1=%d in right",nk);
if (!sf_histint(Frts,"n2",&ntmp) || ntmp != clr_s->n2_sum) sf_error("Need n2=%d in right",clr_s->n2_sum);
lt_s = sf_complexalloc2(nxyz,clr_s->n2_sum);
rt_s = sf_complexalloc2(nk ,clr_s->n2_sum);
sf_complexread(lt_s[0],nxyz*clr_s->n2_sum,Flts);
sf_complexread(rt_s[0],nk *clr_s->n2_sum,Frts);
/* initialize to zero */
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,1) \
private(iy,ix,iz) \
shared(fdm,uoz,uox,uoy,uoz_p,uox_p,uoy_p,uoz_s,uox_s,uoy_s)
#endif
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
uoz [iy][ix][iz]=sf_cmplx(0.,0.); uox [iy][ix][iz]=sf_cmplx(0.,0.); uoy [iy][ix][iz]=sf_cmplx(0.,0.);
uoz_p[iy][ix][iz]=sf_cmplx(0.,0.); uox_p[iy][ix][iz]=sf_cmplx(0.,0.); uoy_p[iy][ix][iz]=sf_cmplx(0.,0.);
uoz_s[iy][ix][iz]=sf_cmplx(0.,0.); uox_s[iy][ix][iz]=sf_cmplx(0.,0.); uoy_s[iy][ix][iz]=sf_cmplx(0.,0.);
/*upz[iy][ix][iz]=sf_cmplx(0.,0.); upx[iy][ix][iz]=sf_cmplx(0.,0.); upy[iy][ix][iz]=sf_cmplx(0.,0.);*/
}
}
}
/*------------------------------------------------------------*/
/*------------------------ MAIN LOOP -------------------------*/
/*------------------------------------------------------------*/
if(verb) fprintf(stderr,"\n");
for (it=0; it<nt; it++) {
if(verb) sf_warning("it=%d/%d;",it,nt); /*fprintf(stderr,"\b\b\b\b\b%d",it);*/
/*------------------------------------------------------------*/
/* apply lowrank matrix to wavefield vector */
/*------------------------------------------------------------*/
clr3d_apply(uo_p, uo, lt_p, rt_p, fdm, dft, clr_p);
clr3d_apply(uo_s, uo, lt_s, rt_s, fdm, dft, clr_s);
/*------------------------------------------------------------*/
/* combine P and S wave modes */
/*------------------------------------------------------------*/
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,1) \
private(iy,ix,iz) \
shared(fdm,uoz,uox,uoy,uoz_p,uox_p,uoy_p,uoz_s,uox_s,uoy_s)
#endif
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
uoz[iy][ix][iz] = uoz_p[iy][ix][iz] + uoz_s[iy][ix][iz];
uox[iy][ix][iz] = uox_p[iy][ix][iz] + uox_s[iy][ix][iz];
uoy[iy][ix][iz] = uoy_p[iy][ix][iz] + uoy_s[iy][ix][iz];
/*upz[iy][ix][iz]=sf_cmplx(0.,0.); upx[iy][ix][iz]=sf_cmplx(0.,0.); upy[iy][ix][iz]=sf_cmplx(0.,0.);*/
}
}
}
/*------------------------------------------------------------*/
/* free surface */
/*------------------------------------------------------------*/
if(fsrf) { /* need to do something here */ }
/*------------------------------------------------------------*/
/* inject displacement source */
/*------------------------------------------------------------*/
if(esou) {
/* exploding force source */
lint3d_expl_complex(uoz,uox,uoy,ww[it],cs);
} else {
if(nbell) {
lint3d_bell_complex(uoz,ww[it][0],cs);
lint3d_bell_complex(uox,ww[it][1],cs);
lint3d_bell_complex(uoy,ww[it][2],cs);
} else {
lint3d_inject_complex(uoz,ww[it][0],cs);
lint3d_inject_complex(uox,ww[it][1],cs);
lint3d_inject_complex(uoy,ww[it][2],cs);
}
}
/*------------------------------------------------------------*/
/* cut wavefield and save */
/*------------------------------------------------------------*/
if(snap && (it-shft)%jsnap==0) {
/* P wave */
cut3d_complex(uoz_p,uc,fdm,acz,acx,acy);
sf_complexwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfp);
cut3d_complex(uox_p,uc,fdm,acz,acx,acy);
sf_complexwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfp);
cut3d_complex(uoy_p,uc,fdm,acz,acx,acy);
sf_complexwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfp);
/* S wave */
cut3d_complex(uoz_s,uc,fdm,acz,acx,acy);
sf_complexwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfs);
cut3d_complex(uox_s,uc,fdm,acz,acx,acy);
sf_complexwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfs);
cut3d_complex(uoy_s,uc,fdm,acz,acx,acy);
sf_complexwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfs);
}
lint3d_extract_complex(uoz,dd[0],cr);
lint3d_extract_complex(uox,dd[1],cr);
lint3d_extract_complex(uoy,dd[2],cr);
if(it%jdata==0) sf_complexwrite(dd[0],nr*nc,Fdat);
}
if(verb) sf_warning(".");
if(verb) fprintf(stderr,"\n");
/*------------------------------------------------------------*/
/* deallocate arrays */
free(dft);
dft3d_finalize();
free(clr_p); free(clr_s);
clr3d_finalize();
free(**ww); free(*ww); free(ww);
free(ss);
free(rr);
free(*dd); free(dd);
free(n2s_p); free(n2s_s);
free(*lt_p); free(lt_p); free(*rt_p); free(rt_p);
free(*lt_s); free(lt_s); free(*rt_s); free(rt_s);
free(**uoz ); free(*uoz ); free(uoz) ;
free(**uoz_p); free(*uoz_p); free(uoz_p);
free(**uoz_s); free(*uoz_s); free(uoz_s);
/*free(**upz); free(*upz); free(upz);*/
free(uo); free(uo_p); free(uo_s);
/*free(up);*/
if (snap) {
free(**uc); free(*uc); free(uc);
}
/*------------------------------------------------------------*/
exit (0);
}
int main(int argc, char* argv[])
{
bool verb,fsrf,snap,ssou,dabc,opot;
int jsnap,ntsnap,jdata;
/* I/O files */
sf_file Fwav=NULL; /* wavelet */
sf_file Fsou=NULL; /* sources */
sf_file Frec=NULL; /* receivers */
sf_file Fccc=NULL; /* velocity */
sf_file Fden=NULL; /* density */
sf_file Fdat=NULL; /* data */
sf_file Fwfl=NULL; /* wavefield */
/* cube axes */
sf_axis at,ax,ay,az;
sf_axis as,ar,ac;
int nt,nz,nx,ny,ns,nr,nc,nb;
int it,iz,ix,iy;
float dt,dz,dx,dy,idz,idx,idy;
/* FDM structure */
fdm3d fdm=NULL;
abcone3d /* abcp=NULL, */ abcs=NULL;
sponge spo=NULL;
/* I/O arrays */
float***ww=NULL; /* wavelet */
pt3d *ss=NULL; /* sources */
pt3d *rr=NULL; /* receivers */
float **dd=NULL; /* data */
/*------------------------------------------------------------*/
float ***tt=NULL;
float ***ro=NULL; /* density */
/* orthorombic footprint - 9 coefficients */
/* c11 c12 c13
. c22 c23
. . c33
c44
c55
c66 */
float ***c11=NULL;
float ***c22=NULL;
float ***c33=NULL;
float ***c44=NULL;
float ***c55=NULL;
float ***c66=NULL;
float ***c12=NULL;
float ***c13=NULL;
float ***c23=NULL;
float ***vp,***vs;
float ***qp=NULL,***qsx=NULL,***qsy=NULL,***qsz=NULL;
/*------------------------------------------------------------*/
/* displacement: um = U @ t-1; uo = U @ t; up = U @ t+1 */
float ***umz,***uoz,***upz,***uaz,***utz;
float ***umx,***uox,***upx,***uax,***utx;
float ***umy,***uoy,***upy,***uay,***uty;
/* stress/strain tensor */
float ***tzz,***txx,***tyy,***txy,***tyz,***tzx;
float szz, sxx, syy, sxy, syz, szx;
/*------------------------------------------------------------*/
/* linear interpolation weights/indices */
lint3d cs,cr;
/* Gaussian bell */
int nbell;
/* wavefield cut params */
sf_axis acz=NULL,acx=NULL,acy=NULL;
int nqz,nqx,nqy;
float oqz,oqx,oqy;
float dqz,dqx,dqy;
float ***uc=NULL;
/*------------------------------------------------------------*/
/* init RSF */
sf_init(argc,argv);
/*------------------------------------------------------------*/
/* OMP parameters */
#ifdef _OPENMP
omp_init();
#endif
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* execution flags */
if(! sf_getbool("verb",&verb)) verb=false; /* verbosity flag */
if(! sf_getbool("snap",&snap)) snap=false; /* wavefield snapshots flag */
if(! sf_getbool("free",&fsrf)) fsrf=false; /* free surface flag */
if(! sf_getbool("ssou",&ssou)) ssou=false; /* stress source */
if(! sf_getbool("dabc",&dabc)) dabc=false; /* absorbing BC */
if(! sf_getbool("opot",&opot)) opot=false; /* output potentials */
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* I/O files */
Fwav = sf_input ("in" ); /* wavelet */
Fccc = sf_input ("ccc"); /* stiffness */
Fden = sf_input ("den"); /* density */
Fsou = sf_input ("sou"); /* sources */
Frec = sf_input ("rec"); /* receivers */
Fwfl = sf_output("wfl"); /* wavefield */
Fdat = sf_output("out"); /* data */
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* axes */
at = sf_iaxa(Fwav,3); sf_setlabel(at,"t"); if(verb) sf_raxa(at); /* time */
az = sf_iaxa(Fccc,1); sf_setlabel(az,"z"); if(verb) sf_raxa(az); /* depth */
ax = sf_iaxa(Fccc,2); sf_setlabel(ax,"x"); if(verb) sf_raxa(ax); /* space x */
ay = sf_iaxa(Fccc,3); sf_setlabel(ay,"y"); if(verb) sf_raxa(ay); /* space y */
as = sf_iaxa(Fsou,2); sf_setlabel(as,"s"); if(verb) sf_raxa(as); /* sources */
ar = sf_iaxa(Frec,2); sf_setlabel(ar,"r"); if(verb) sf_raxa(ar); /* receivers */
nt = sf_n(at); dt = sf_d(at);
nz = sf_n(az); dz = sf_d(az);
nx = sf_n(ax); dx = sf_d(ax);
ny = sf_n(ay); dy = sf_d(ay);
ns = sf_n(as);
nr = sf_n(ar);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* other execution parameters */
if(! sf_getint("nbell",&nbell)) nbell=5; /* bell size */
if(verb) sf_warning("nbell=%d",nbell);
if(! sf_getint("jdata",&jdata)) jdata=1;
if(snap) { /* save wavefield every *jsnap* time steps */
if(! sf_getint("jsnap",&jsnap)) jsnap=nt;
}
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* expand domain for FD operators and ABC */
if( !sf_getint("nb",&nb) || nb<NOP) nb=NOP;
fdm=fdutil3d_init(verb,fsrf,az,ax,ay,nb,1);
fdbell3d_init(nbell);
sf_setn(az,fdm->nzpad); sf_seto(az,fdm->ozpad); if(verb) sf_raxa(az);
sf_setn(ax,fdm->nxpad); sf_seto(ax,fdm->oxpad); if(verb) sf_raxa(ax);
sf_setn(ay,fdm->nypad); sf_seto(ay,fdm->oypad); if(verb) sf_raxa(ay);
/*------------------------------------------------------------*/
/* 3D vector components */
nc=3;
if(opot) {
ac=sf_maxa(nc+1,0,1);
} else {
ac=sf_maxa(nc ,0,1);
}
/*------------------------------------------------------------*/
/* setup output data header */
sf_oaxa(Fdat,ar,1);
sf_oaxa(Fdat,ac,2);
sf_setn(at,nt/jdata);
sf_setd(at,dt*jdata);
sf_oaxa(Fdat,at,3);
/* setup output wavefield header */
if(snap) {
if(!sf_getint ("nqz",&nqz)) nqz=sf_n(az);
if(!sf_getint ("nqx",&nqx)) nqx=sf_n(ax);
if(!sf_getint ("nqy",&nqy)) nqy=sf_n(ay);
if(!sf_getfloat("oqz",&oqz)) oqz=sf_o(az);
if(!sf_getfloat("oqx",&oqx)) oqx=sf_o(ax);
if(!sf_getfloat("oqy",&oqy)) oqy=sf_o(ay);
dqz=sf_d(az);
dqx=sf_d(ax);
dqy=sf_d(ay);
acz = sf_maxa(nqz,oqz,dqz); sf_raxa(acz);
acx = sf_maxa(nqx,oqx,dqx); sf_raxa(acx);
acy = sf_maxa(nqy,oqy,dqy); sf_raxa(acy);
/* TODO: check if the imaging window fits in the wavefield domain */
uc=sf_floatalloc3(sf_n(acz),sf_n(acx),sf_n(acy));
ntsnap=0;
for(it=0; it<nt; it++) {
if(it%jsnap==0) ntsnap++;
}
sf_setn(at, ntsnap);
sf_setd(at,dt*jsnap);
if(verb) sf_raxa(at);
sf_oaxa(Fwfl,acz,1);
sf_oaxa(Fwfl,acx,2);
sf_oaxa(Fwfl,acy,3);
sf_oaxa(Fwfl,ac, 4);
sf_oaxa(Fwfl,at, 5);
}
/*------------------------------------------------------------*/
/* source array */
ww=sf_floatalloc3(ns,nc,nt);
sf_floatread(ww[0][0],nt*nc*ns,Fwav);
/* data array */
if(opot) {
dd=sf_floatalloc2(nr,nc+1);
} else {
dd=sf_floatalloc2(nr,nc );
}
/*------------------------------------------------------------*/
/* setup source/receiver coordinates */
ss = (pt3d*) sf_alloc(ns,sizeof(*ss));
rr = (pt3d*) sf_alloc(nr,sizeof(*rr));
pt3dread1(Fsou,ss,ns,3); /* read (x,y,z) coordinates */
pt3dread1(Frec,rr,nr,3); /* read (x,y,z) coordinates */
cs = lint3d_make(ns,ss,fdm);
cr = lint3d_make(nr,rr,fdm);
/*------------------------------------------------------------*/
/* setup FD coefficients */
/* idz = 2/dz;*/
/* idx = 2/dx;*/
/* idy = 2/dy;*/
idz = 1/dz;
idx = 1/dx;
idy = 1/dy;
/*------------------------------------------------------------*/
tt = sf_floatalloc3(nz,nx,ny);
ro =sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
c11=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
c22=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
c33=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
c44=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
c55=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
c66=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
c12=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
c13=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
c23=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
/* input density */
sf_floatread(tt[0][0],nz*nx*ny,Fden); expand3d(tt,ro ,fdm);
/* input stiffness */
sf_floatread(tt[0][0],nz*nx*ny,Fccc ); expand3d(tt,c11,fdm);
sf_floatread(tt[0][0],nz*nx*ny,Fccc ); expand3d(tt,c22,fdm);
sf_floatread(tt[0][0],nz*nx*ny,Fccc ); expand3d(tt,c33,fdm);
sf_floatread(tt[0][0],nz*nx*ny,Fccc ); expand3d(tt,c44,fdm);
sf_floatread(tt[0][0],nz*nx*ny,Fccc ); expand3d(tt,c55,fdm);
sf_floatread(tt[0][0],nz*nx*ny,Fccc ); expand3d(tt,c66,fdm);
sf_floatread(tt[0][0],nz*nx*ny,Fccc ); expand3d(tt,c12,fdm);
sf_floatread(tt[0][0],nz*nx*ny,Fccc ); expand3d(tt,c13,fdm);
sf_floatread(tt[0][0],nz*nx*ny,Fccc ); expand3d(tt,c23,fdm);
free(**tt); free(*tt); free(tt);
/*------------------------------------------------------------*/
if(dabc) {
/* one-way abc setup */
vp = sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
vs = sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
vp[iy][ix][iz] = sqrt( c11[iy][ix][iz]/ro[iy][ix][iz] );
vs[iy][ix][iz] = sqrt( c55[iy][ix][iz]/ro[iy][ix][iz] );
}
}
}
/* abcp = abcone3d_make(NOP,dt,vp,fsrf,fdm); */
abcs = abcone3d_make(NOP,dt,vs,fsrf,fdm);
free(**vp); free(*vp); free(vp);
free(**vs); free(*vs); free(vs);
/* sponge abc setup */
spo = sponge_make(fdm->nb);
}
/*------------------------------------------------------------*/
/* precompute 1/ro * dt^2 */
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
ro[iy][ix][iz] = dt*dt/ro[iy][ix][iz];
}
}
}
/*------------------------------------------------------------*/
/* allocate wavefield arrays */
umz=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uoz=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
upz=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uaz=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
umx=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uox=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
upx=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uax=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
umy=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uoy=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
upy=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uay=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
tzz=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
tyy=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
txx=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
txy=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
tyz=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
tzx=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
umz[iy][ix][iz]=0; umx[iy][ix][iz]=0; umy[iy][ix][iz]=0;
uoz[iy][ix][iz]=0; uox[iy][ix][iz]=0; uoy[iy][ix][iz]=0;
upz[iy][ix][iz]=0; upx[iy][ix][iz]=0; upy[iy][ix][iz]=0;
uaz[iy][ix][iz]=0; uax[iy][ix][iz]=0; uay[iy][ix][iz]=0;
}
}
}
if(opot) {
qp =sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
qsx=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
qsy=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
qsz=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
}
/*------------------------------------------------------------*/
/*
* MAIN LOOP
*/
/*------------------------------------------------------------*/
if(verb) fprintf(stderr,"\n");
for (it=0; it<nt; it++) {
if(verb) fprintf(stderr,"\b\b\b\b\b%d",it);
/*------------------------------------------------------------*/
/* from displacement to strain */
/*------------------------------------------------------------*/
/*
* exx = Fx(ux)
* eyy = Fy(uy)
* ezz = Fz(uz)
* exy = By(ux) + Bx(uy)
* eyz = Bz(uy) + By(uz)
* ezx = Bx(uz) + Bz(ux)
*/
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iy,iz) \
shared(fdm,txx,tyy,tzz,txy,tyz,tzx,uox,uoy,uoz,idx,idy,idz)
#endif
for (iy=NOP; iy<fdm->nypad-NOP; iy++) {
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
txx[iy][ix][iz] = Dx(uox,ix,iy,iz,idx);
tyy[iy][ix][iz] = Dy(uoy,ix,iy,iz,idy);
tzz[iy][ix][iz] = Dz(uoz,ix,iy,iz,idz);
txy[iy][ix][iz] = Dy(uox,ix,iy,iz,idy) + Dx(uoy,ix,iy,iz,idx);
tyz[iy][ix][iz] = Dz(uoy,ix,iy,iz,idz) + Dy(uoz,ix,iy,iz,idy);
tzx[iy][ix][iz] = Dx(uoz,ix,iy,iz,idx) + Dz(uox,ix,iy,iz,idz);
}
}
}
/*------------------------------------------------------------*/
/* from strain to stress */
/*------------------------------------------------------------*/
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iy,iz,sxx,syy,szz,sxy,syz,szx) \
shared(fdm,txx,tyy,tzz,txy,tyz,tzx,c11,c22,c33,c44,c55,c66,c12,c13,c23)
#endif
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
sxx = c11[iy][ix][iz] * txx[iy][ix][iz]
+ c12[iy][ix][iz] * tyy[iy][ix][iz]
+ c13[iy][ix][iz] * tzz[iy][ix][iz];
syy = c12[iy][ix][iz] * txx[iy][ix][iz]
+ c22[iy][ix][iz] * tyy[iy][ix][iz]
+ c23[iy][ix][iz] * tzz[iy][ix][iz];
szz = c13[iy][ix][iz] * txx[iy][ix][iz]
+ c23[iy][ix][iz] * tyy[iy][ix][iz]
+ c33[iy][ix][iz] * tzz[iy][ix][iz];
sxy = c66[iy][ix][iz] * txy[iy][ix][iz];
syz = c44[iy][ix][iz] * tyz[iy][ix][iz];
szx = c55[iy][ix][iz] * tzx[iy][ix][iz];
txx[iy][ix][iz] = sxx;
tyy[iy][ix][iz] = syy;
tzz[iy][ix][iz] = szz;
txy[iy][ix][iz] = sxy;
tyz[iy][ix][iz] = syz;
tzx[iy][ix][iz] = szx;
}
}
}
/*------------------------------------------------------------*/
/* free surface */
/*------------------------------------------------------------*/
if(fsrf) {
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iy,iz) \
shared(fdm,txx,tyy,tzz,txy,tyz,tzx)
#endif
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nb; iz++) {
txx[iy][ix][iz]=0;
tyy[iy][ix][iz]=0;
tzz[iy][ix][iz]=0;
txy[iy][ix][iz]=0;
tyz[iy][ix][iz]=0;
tzx[iy][ix][iz]=0;
}
}
}
}
/*------------------------------------------------------------*/
/* inject stress source */
/*------------------------------------------------------------*/
if(ssou) {
lint3d_bell(tzz,ww[it][0],cs);
lint3d_bell(txx,ww[it][1],cs);
lint3d_bell(tyy,ww[it][2],cs);
}
/*------------------------------------------------------------*/
/* from stress to acceleration */
/*------------------------------------------------------------*/
/*
* ax = Bx(txx) + Fy(txy) + Fz(txz)
* ay = Fx(txy) + By(tyy) + Fz(tyz)
* az = Fx(txz) + Fy(tyz) + Bz(tzz)
*/
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iy,iz) \
shared(fdm,txx,tyy,tzz,txy,tyz,tzx,uax,uay,uaz,idx,idy,idz)
#endif
for (iy=NOP; iy<fdm->nypad-NOP; iy++) {
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
uax[iy][ix][iz] = Dx( txx,ix,iy,iz,idx ) + Dy( txy,ix,iy,iz,idy ) + Dz( tzx,ix,iy,iz,idz ) ;
uay[iy][ix][iz] = Dx( txy,ix,iy,iz,idx ) + Dy( tyy,ix,iy,iz,idy ) + Dz( tyz,ix,iy,iz,idz ) ;
uaz[iy][ix][iz] = Dx( tzx,ix,iy,iz,idx ) + Dy( tyz,ix,iy,iz,idy ) + Dz( tzz,ix,iy,iz,idz ) ;
}
}
}
/*------------------------------------------------------------*/
/* inject acceleration source */
/*------------------------------------------------------------*/
if(!ssou) {
lint3d_bell(uaz,ww[it][0],cs);
lint3d_bell(uax,ww[it][1],cs);
lint3d_bell(uay,ww[it][2],cs);
}
/*------------------------------------------------------------*/
/* step forward in time */
/*------------------------------------------------------------*/
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iy,iz) \
shared(fdm,uox,uoy,uoz,umx,umy,umz,upx,upy,upz,uax,uay,uaz,ro)
#endif
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
upx[iy][ix][iz] = 2*uox[iy][ix][iz]
- umx[iy][ix][iz]
+ uax[iy][ix][iz] * ro[iy][ix][iz];
upy[iy][ix][iz] = 2*uoy[iy][ix][iz]
- umy[iy][ix][iz]
+ uay[iy][ix][iz] * ro[iy][ix][iz];
upz[iy][ix][iz] = 2*uoz[iy][ix][iz]
- umz[iy][ix][iz]
+ uaz[iy][ix][iz] * ro[iy][ix][iz];
}
}
}
/* circulate wavefield arrays */
utz=umz; uty=umy; utx=umx;
umz=uoz; umy=uoy; umx=uox;
uoz=upz; uoy=upy; uox=upx;
upz=utz; upy=uty; upx=utx;
if(dabc) {
/* one-way ABC */
/* abcone3d_apply(uoz,umz,NOP,abcp,fdm); */
/* abcone3d_apply(uox,umx,NOP,abcp,fdm); */
/* abcone3d_apply(uoy,umy,NOP,abcp,fdm); */
abcone3d_apply(uoz,umz,NOP,abcs,fdm);
abcone3d_apply(uox,umx,NOP,abcs,fdm);
abcone3d_apply(uoy,umy,NOP,abcs,fdm);
/* sponge ABC */
sponge3d_apply(umz,spo,fdm);
sponge3d_apply(uoz,spo,fdm);
sponge3d_apply(umx,spo,fdm);
sponge3d_apply(uox,spo,fdm);
sponge3d_apply(umy,spo,fdm);
sponge3d_apply(uoy,spo,fdm);
}
/*------------------------------------------------------------*/
/* cut wavefield and save */
/*------------------------------------------------------------*/
if(opot) {
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iy,iz) \
shared(fdm,uox,uoy,uoz,idx,idy,idz)
#endif
for (iy=NOP; iy<fdm->nypad-NOP; iy++) {
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
qp [iy][ix][iz] = Dx( uox,ix,iy,iz,idx )
+ Dy( uoy,ix,iy,iz,idy )
+ Dz( uoz,ix,iy,iz,idz );
qsx[iy][ix][iz] = Dy( uoz,ix,iy,iz,idy ) - Dz( uoy,ix,iy,iz,idz );
qsy[iy][ix][iz] = Dz( uox,ix,iy,iz,idz ) - Dx( uoz,ix,iy,iz,idx );
qsz[iy][ix][iz] = Dx( uoy,ix,iy,iz,idx ) - Dy( uox,ix,iy,iz,idy );
}
}
}
if(snap && it%jsnap==0) {
cut3d(qp ,uc,fdm,acz,acx,acy);
sf_floatwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfl);
cut3d(qsz,uc,fdm,acz,acx,acy);
sf_floatwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfl);
cut3d(qsx,uc,fdm,acz,acx,acy);
sf_floatwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfl);
cut3d(qsy,uc,fdm,acz,acx,acy);
sf_floatwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfl);
}
lint3d_extract(qp , dd[0],cr);
lint3d_extract(qsx, dd[1],cr);
lint3d_extract(qsy, dd[2],cr);
lint3d_extract(qsz, dd[3],cr);
if(it%jdata==0) sf_floatwrite(dd[0],nr*(nc+1),Fdat);
} else {
if(snap && it%jsnap==0) {
cut3d(uoz,uc,fdm,acz,acx,acy);
sf_floatwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfl);
cut3d(uox,uc,fdm,acz,acx,acy);
sf_floatwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfl);
cut3d(uoy,uc,fdm,acz,acx,acy);
sf_floatwrite(uc[0][0],sf_n(acx)*sf_n(acy)*sf_n(acz),Fwfl);
}
lint3d_extract(uoz,dd[0],cr);
lint3d_extract(uox,dd[1],cr);
lint3d_extract(uoy,dd[2],cr);
if(it%jdata==0) sf_floatwrite(dd[0],nr*nc,Fdat);
}
}
if(verb) fprintf(stderr,"\n");
/*------------------------------------------------------------*/
/* deallocate arrays */
free(**ww); free(*ww); free(ww);
free(ss);
free(rr);
free(*dd); free(dd);
free(**ro); free(*ro); free(ro);
free(**c11); free(*c11); free(c11);
free(**c22); free(*c22); free(c22);
free(**c33); free(*c33); free(c33);
free(**c44); free(*c44); free(c44);
free(**c55); free(*c55); free(c55);
free(**c66); free(*c66); free(c66);
free(**c12); free(*c12); free(c12);
free(**c13); free(*c13); free(c13);
free(**c23); free(*c23); free(c23);
free(**umz); free(*umz); free(umz);
free(**uoz); free(*uoz); free(uoz);
free(**upz); free(*upz); free(upz);
free(**uaz); free(*uaz); free(uaz);
free(**umx); free(*umx); free(umx);
free(**uox); free(*uox); free(uox);
free(**upx); free(*upx); free(upx);
free(**uax); free(*uax); free(uax);
free(**umy); free(*umy); free(umy);
free(**uoy); free(*uoy); free(uoy);
free(**upy); free(*upy); free(upy);
free(**uay); free(*uay); free(uay);
free(**tzz); free(*tzz); free(tzz);
free(**txx); free(*txx); free(txx);
free(**tyy); free(*tyy); free(tyy);
free(**txy); free(*txy); free(txy);
free(**tyz); free(*tyz); free(tyz);
free(**tzx); free(*tzx); free(tzx);
if (snap) {
free(**uc); free(*uc); free(uc);
}
if(opot) {
free(**qp); free(*qp); free(qp);
free(**qsx); free(*qsx); free(qsx);
free(**qsy); free(*qsy); free(qsy);
free(**qsz); free(*qsz); free(qsz);
}
/*------------------------------------------------------------*/
exit (0);
}
int main(int argc, char *argv[])
{
bool wantwf, verb;
int ix, iz, is, it, wfit, im, ik, i, j, itau;
int ns, nx, nz, nt, wfnt, rnx, rnz, nzx, rnzx, vnx, ntau, htau, nds;
int scalet, snap, snapshot, fnx, fnz, fnzx, nk, nb;
int rectx, rectz, repeat, gpz, n, m, pad1, trunc, spx, spz;
float dt, t0, z0, dz, x0, dx, s0, ds, wfdt, srctrunc;
float dtau, tau0, tau;
int nr, ndr, nr0;
char *path1, *path2, number[5], *left, *right;
double tstart, tend;
struct timeval tim;
/*wavenumber domain tapering*/
int taper;
float *ktp;
float ktmp,kx_trs,kz_trs,thresh;
float dkx,dkz,kx0,kz0;
float kx,kz;
int nkz;
sf_complex c, **lt, **rt;
sf_complex *ww, **dd, ***dd3;
float ***img1, **img2, ***mig1, **mig2;
float *rr, **ccr, **sill, ***fwf, ***bwf;
sf_complex *cwave, *cwavem, **wave, *curr;
sf_axis at, ax, az, atau;
sf_file Fdat, Fsrc, Fimg1, Fimg2;
sf_file Ffwf, Fbwf, Fvel;
sf_file Fleft, Fright;
int cpuid, numprocs, nth, nspad, iturn;
float *sendbuf, *recvbuf;
sf_complex *sendbufc, *recvbufc;
MPI_Comm comm=MPI_COMM_WORLD;
MPI_Init(&argc, &argv);
MPI_Comm_rank(comm, &cpuid);
MPI_Comm_size(comm, &numprocs);
sf_init(argc, argv);
#ifdef _OPENMP
#pragma omp parallel
{
nth=omp_get_num_threads();
}
sf_warning(">>> Using %d threads <<<", nth);
#endif
gettimeofday(&tim, NULL);
tstart=tim.tv_sec+(tim.tv_usec/1000000.0);
if (!sf_getint("taper",&taper)) taper=0; /* tapering in the frequency domain */
if (!sf_getfloat("thresh",&thresh)) thresh=0.92; /* tapering threshold */
if(!sf_getbool("wantwf", &wantwf)) wantwf=false;
if(!sf_getbool("verb", &verb)) verb=false;
if(!sf_getint("pad1", &pad1)) pad1=1;
/* padding factor on the first axis */
if(!sf_getint("nb", &nb)) sf_error("Need nb= ");
if(!sf_getfloat("srctrunc", &srctrunc)) srctrunc=0.4;
if(!sf_getint("rectx", &rectx)) rectx=2;
if(!sf_getint("rectz", &rectz)) rectz=2;
if(!sf_getint("repeat", &repeat)) repeat=2;
if(!sf_getint("scalet", &scalet)) scalet=1;
if(!sf_getint("snap", &snap)) snap=100;
/* interval of the output wavefield */
if(!sf_getint("snapshot", &snapshot)) snapshot=0;
/* print out the wavefield snapshots of this shot */
if(!sf_getint("nds", &nds)) sf_error("Need nds=!");
/* source and receiver positions */
if(!sf_getint("gpz", &gpz)) sf_error("Need gpz=");
if(!sf_getint("spx", &spx)) sf_error("Need spx=");
if(!sf_getint("spz", &spz)) sf_error("Need spz=");
/* tau parameters */
if(!sf_getint("ntau", &ntau)) sf_error("Need ntau=");
if(!sf_getfloat("dtau", &dtau)) sf_error("Need dtau=");
if(!sf_getfloat("tau0", &tau0)) sf_error("Need tau0=");
/* geometry parameters */
if(!sf_getint("rnx", &rnx)) sf_error("Need rnx=");
if(!sf_getint("ndr", &ndr)) ndr=1;
if(!sf_getint("nr0", &nr0)) nr0=0;
/* input/output files */
Fdat=sf_input("--input");
Fimg1=sf_output("--output");
Fimg2=sf_output("Fimg2");
Fsrc=sf_input("Fsrc");
Fvel=sf_input("Fpadvel");
if(wantwf){
Ffwf=sf_output("Ffwf");
Fbwf=sf_output("Fbwf");
}
at=sf_iaxa(Fsrc, 1); nt=sf_n(at); dt=sf_d(at); t0=sf_o(at);
ax=sf_iaxa(Fvel, 2); vnx=sf_n(ax); dx=sf_d(ax); x0=sf_o(ax);
az=sf_iaxa(Fvel, 1); rnz=sf_n(az); dz=sf_d(az); z0=sf_o(az);
if(!sf_histint(Fdat, "n2", &nr)) sf_error("Need n2= in input!");
if(!sf_histint(Fdat, "n3", &ns)) sf_error("Need n3= in input!");
if(!sf_histfloat(Fdat, "d3", &ds)) sf_error("Need d3= in input!");
if(!sf_histfloat(Fdat, "o3", &s0)) sf_error("Need o3= in input!");
wfnt=(nt-1)/scalet+1;
wfdt=dt*scalet;
/* double check the geometry parameters */
if(nds != (int)(ds/dx)) sf_error("Need ds/dx= %d", nds);
//sf_warning("s0=%g, x0+(rnx-1)*dx/2=%g", s0, x0+(rnx-1)*dx/2);
//if(s0 != x0+(rnx-1)*dx/2) sf_error("Wrong origin information!");
if(vnx != nds*(ns-1)+rnx) sf_error("Wrong dimension in x axis!");
/* set up the output files */
atau=sf_iaxa(Fsrc, 1);
sf_setn(atau, ntau);
sf_setd(atau, dtau);
sf_seto(atau, tau0);
sf_setlabel(atau, "Tau");
sf_setunit(atau, "s");
sf_oaxa(Fimg1, az, 1);
sf_oaxa(Fimg1, ax, 2);
sf_oaxa(Fimg1, atau, 3);
sf_oaxa(Fimg2, az, 1);
sf_oaxa(Fimg2, ax, 2);
sf_putint(Fimg2, "n3", 1);
sf_settype(Fimg1, SF_FLOAT);
sf_settype(Fimg2, SF_FLOAT);
if(wantwf){
sf_setn(ax, rnx);
sf_seto(ax, -(rnx-1)*dx/2.0);
sf_oaxa(Ffwf, az, 1);
sf_oaxa(Ffwf, ax, 2);
sf_putint(Ffwf, "n3", (wfnt-1)/snap+1);
sf_putfloat(Ffwf, "d3", snap*wfdt);
sf_putfloat(Ffwf, "o3", t0);
sf_putstring(Ffwf, "label3", "Time");
sf_putstring(Ffwf, "unit3", "s");
sf_settype(Ffwf, SF_FLOAT);
sf_oaxa(Fbwf, az, 1);
sf_oaxa(Fbwf, ax, 2);
sf_putint(Fbwf, "n3", (wfnt-1)/snap+1);
sf_putfloat(Fbwf, "d3", -snap*wfdt);
sf_putfloat(Fbwf, "o3", (wfnt-1)*wfdt);
sf_putstring(Fbwf, "label3", "Time");
sf_putstring(Fbwf, "unit3", "s");
sf_settype(Fbwf, SF_FLOAT);
}
nx=rnx+2*nb; nz=rnz+2*nb;
nzx=nx*nz; rnzx=rnz*rnx;
nk=cfft2_init(pad1, nz, nx, &fnz, &fnx);
fnzx=fnz*fnx;
if(ns%numprocs==0) nspad=ns;
else nspad=(ns/numprocs+1)*numprocs;
/* print axies parameters for double check */
sf_warning("cpuid=%d, numprocs=%d, nspad=%d", cpuid, numprocs, nspad);
sf_warning("nt=%d, dt=%g, scalet=%d, wfnt=%d, wfdt=%g",nt, dt, scalet, wfnt, wfdt);
sf_warning("vnx=%d, nx=%d, dx=%g, nb=%d, rnx=%d", vnx, nx, dx, nb, rnx);
sf_warning("nr=%d, ndr=%d, nr0=%g", nr, ndr, nr0);
sf_warning("nz=%d, rnz=%d, dz=%g, z0=%g", nz, rnz, dz, z0);
sf_warning("spx=%d, spz=%d, gpz=%d", spx, spz, gpz);
sf_warning("ns=%d, ds=%g, s0=%g", ns, ds, s0);
sf_warning("ntau=%d, dtau=%g, tau0=%g", ntau, dtau, tau0);
sf_warning("nzx=%d, fnzx=%d, nk=%d", nzx, fnzx, nk);
/* allocate storage and read data */
ww=sf_complexalloc(nt);
sf_complexread(ww, nt, Fsrc);
sf_fileclose(Fsrc);
gpz=gpz+nb;
spz=spz+nb;
spx=spx+nb;
nr0=nr0+nb;
trunc=srctrunc/dt+0.5;
dd=sf_complexalloc2(nt, nr);
if(cpuid==0) dd3=sf_complexalloc3(nt, nr, numprocs);
rr=sf_floatalloc(nzx);
reflgen(nz, nx, spz, spx, rectz, rectx, repeat, rr);
fwf=sf_floatalloc3(rnz, rnx, wfnt);
bwf=sf_floatalloc3(rnz, rnx, wfnt);
img1=sf_floatalloc3(rnz, vnx, ntau);
img2=sf_floatalloc2(rnz, vnx);
mig1=sf_floatalloc3(rnz, rnx, ntau);
mig2=sf_floatalloc2(rnz, rnx);
ccr=sf_floatalloc2(rnz, rnx);
sill=sf_floatalloc2(rnz, rnx);
curr=sf_complexalloc(fnzx);
cwave=sf_complexalloc(nk);
cwavem=sf_complexalloc(nk);
icfft2_allocate(cwavem);
if (taper!=0) {
dkz = 1./(fnz*dz); kz0 = -0.5/dz;
dkx = 1./(fnx*dx); kx0 = -0.5/dx;
nkz = fnz;
sf_warning("dkz=%f,dkx=%f,kz0=%f,kx0=%f",dkz,dkx,kz0,kx0);
sf_warning("nk=%d,nkz=%d,nkx=%d",nk,nkz,fnx);
kx_trs = thresh*fabs(0.5/dx);
kz_trs = thresh*fabs(0.5/dz);
sf_warning("Applying kz tapering below %f",kz_trs);
sf_warning("Applying kx tapering below %f",kx_trs);
ktp = sf_floatalloc(nk);
/* constructing the tapering op */
for (ix=0; ix < fnx; ix++) {
kx = kx0+ix*dkx;
for (iz=0; iz < nkz; iz++) {
kz = kz0+iz*dkz;
ktmp = 1.;
if (fabs(kx) > kx_trs)
ktmp *= powf((2*kx_trs - fabs(kx))/(kx_trs),2);
if (fabs(kz) > kz_trs)
ktmp *= powf((2*kz_trs - fabs(kz))/(kz_trs),2);
ktp[iz+ix*nkz] = ktmp;
}
}
}
/* initialize image tables that would be used for summing images */
#ifdef _OPENMP
#pragma omp parallel for private(ix, iz, itau)
#endif
for(ix=0; ix<vnx; ix++){
for(iz=0; iz<rnz; iz++){
img2[ix][iz]=0.;
for(itau=0; itau<ntau; itau++){
img1[itau][ix][iz]=0.;
}
}
}
path1=sf_getstring("path1");
path2=sf_getstring("path2");
if(path1==NULL) path1="./mat/left";
if(path2==NULL) path2="./mat/right";
/* shot loop */
for (iturn=0; iturn*numprocs<nspad; iturn++){
is=iturn*numprocs+cpuid;
/* read data */
if(cpuid==0){
sf_seek(Fdat, ((off_t) is)*((off_t) nr)*((off_t) nt)*sizeof(float complex), SEEK_SET);
if((iturn+1)*numprocs<=ns){
sf_complexread(dd3[0][0], nr*nt*numprocs, Fdat);
}else{
sf_complexread(dd3[0][0], nr*nt*(ns-iturn*numprocs), Fdat);
for(is=ns; is<nspad; is++)
for(ix=0; ix<nr; ix++)
for(it=0; it<nt; it++)
dd3[is-iturn*numprocs][ix][it]=sf_cmplx(0.,0.);
is=iturn*numprocs;
}
sendbufc=dd3[0][0];
recvbufc=dd[0];
}else{
sendbufc=NULL;
recvbufc=dd[0];
}
MPI_Scatter(sendbufc, nt*nr, MPI_COMPLEX, recvbufc, nt*nr, MPI_COMPLEX, 0, comm);
if(is<ns){ /* effective shot loop */
/* construct the names of left and right matrices */
left=sf_charalloc(strlen(path1));
right=sf_charalloc(strlen(path2));
strcpy(left, path1);
strcpy(right, path2);
sprintf(number, "%d", is+1);
strcat(left, number);
strcat(right, number);
Fleft=sf_input(left);
Fright=sf_input(right);
if(!sf_histint(Fleft, "n1", &n) || n != nzx) sf_error("Need n1=%d in Fleft", nzx);
if(!sf_histint(Fleft, "n2", &m)) sf_error("No n2 in Fleft");
if(!sf_histint(Fright, "n1", &n) || n != m) sf_error("Need n1=%d in Fright", m);
if(!sf_histint(Fright, "n2", &n) || n != nk) sf_error("Need n2=%d in Fright", nk);
/* allocate storage for each shot migration */
lt=sf_complexalloc2(nzx, m);
rt=sf_complexalloc2(m, nk);
sf_complexread(lt[0], nzx*m, Fleft);
sf_complexread(rt[0], m*nk, Fright);
sf_fileclose(Fleft);
sf_fileclose(Fright);
/* initialize curr and imaging variables */
#ifdef _OPENMP
#pragma omp parallel for private(iz)
#endif
for(iz=0; iz<fnzx; iz++){
curr[iz]=sf_cmplx(0.,0.);
}
#ifdef _OPENMP
#pragma omp parallel for private(ix, iz, itau)
#endif
for(ix=0; ix<rnx; ix++){
for(iz=0; iz<rnz; iz++){
mig2[ix][iz]=0.;
ccr[ix][iz]=0.;
sill[ix][iz]=0.;
for(itau=0; itau<ntau; itau++){
mig1[itau][ix][iz]=0.;
}
}
}
/* wave */
wave=sf_complexalloc2(fnzx, m);
/* snapshot */
if(wantwf && is==snapshot) wantwf=true;
else wantwf=false;
/* forward propagation */
wfit=0;
for(it=0; it<nt; it++){
if(verb) sf_warning("Forward propagation it=%d/%d",it+1, nt);
cfft2(curr, cwave);
for(im=0; im<m; im++){
#ifdef _OPENMP
#pragma omp parallel for private(ik)
#endif
for(ik=0; ik<nk; ik++){
#ifdef SF_HAS_COMPLEX_H
cwavem[ik]=cwave[ik]*rt[ik][im];
#else
cwavem[ik]=sf_cmul(cwave[ik],rt[ik][im]);
#endif
}
icfft2(wave[im],cwavem);
}
#ifdef _OPENMP
#pragma omp parallel for private(ix, iz, i, j, im, c) shared(curr, it)
#endif
for(ix=0; ix<nx; ix++){
for(iz=0; iz<nz; iz++){
i=iz+ix*nz;
j=iz+ix*fnz;
if(it<trunc){
#ifdef SF_HAS_COMPLEX_H
c=ww[it]*rr[i];
#else
c=sf_crmul(ww[it],rr[i]);
#endif
}else{
c=sf_cmplx(0.,0.);
}
// c += curr[j];
for(im=0; im<m; im++){
#ifdef SF_HAS_COMPLEX_H
c += lt[im][i]*wave[im][j];
#else
c += sf_cmul(lt[im][i], wave[im][j]);
#endif
}
curr[j]=c;
}
}
if (taper!=0) {
if (it%taper == 0) {
cfft2(curr,cwave);
for (ik = 0; ik < nk; ik++) {
#ifdef SF_HAS_COMPLEX_H
cwavem[ik] = cwave[ik]*ktp[ik];
#else
cwavem[ik] = sf_crmul(cwave[ik],ktp[ik]);
#endif
}
icfft2(curr,cwavem);
}
}
if(it%scalet==0){
#ifdef _OPENMP
#pragma omp parallel for private(ix, iz)
#endif
for(ix=0; ix<rnx; ix++){
for(iz=0; iz<rnz; iz++){
fwf[wfit][ix][iz]=crealf(curr[(ix+nb)*fnz+(iz+nb)]);
}
}
wfit++;
}
} //end of it
/* check wfnt */
if(wfit != wfnt) sf_error("At this point, wfit should be equal to wfnt");
/* backward propagation starts from here... */
#ifdef _OPENMP
#pragma omp parallel for private(iz)
#endif
for(iz=0; iz<fnzx; iz++){
curr[iz]=sf_cmplx(0.,0.);
}
wfit=wfnt-1;
for(it=nt-1; it>=0; it--){
if(verb) sf_warning("Backward propagation it=%d/%d",it+1, nt);
#ifdef _OPENMP
#pragma omp parallel for private(ix)
#endif
for(ix=0; ix<nr; ix++){
curr[(nr0+ix*ndr)*fnz+gpz]+=dd[ix][it];
}
cfft2(curr, cwave);
for(im=0; im<m; im++){
#ifdef _OPENMP
#pragma omp parallel for private(ik)
#endif
for(ik=0; ik<nk; ik++){
#ifdef SF_HAS_COMPLEX_H
cwavem[ik]=cwave[ik]*conjf(rt[ik][im]);
#else
cwavem[ik]=sf_cmul(cwave[ik],conjf(rt[ik][im]));
#endif
}
icfft2(wave[im],cwavem);
}
#ifdef _OPENMP
#pragma omp parallel for private(ix, iz, i, j, im, c) shared(curr, it)
#endif
for(ix=0; ix<nx; ix++){
for(iz=0; iz<nz; iz++){
i=iz+ix*nz;
j=iz+ix*fnz;
// c=curr[j];
c=sf_cmplx(0.,0.);
for(im=0; im<m; im++){
#ifdef SF_HAS_COMPLEX_H
c += conjf(lt[im][i])*wave[im][j];
#else
c += sf_cmul(conjf(lt[im][i]), wave[im][j]);
#endif
}
curr[j]=c;
}
}
if (taper!=0) {
if (it%taper == 0) {
cfft2(curr,cwave);
for (ik = 0; ik < nk; ik++) {
#ifdef SF_HAS_COMPLEX_H
cwavem[ik] = cwave[ik]*ktp[ik];
#else
cwavem[ik] = sf_crmul(cwave[ik],ktp[ik]);
#endif
}
icfft2(curr,cwavem);
}
}
if(it%scalet==0){
#ifdef _OPENMP
#pragma omp parallel for private(ix, iz)
#endif
for(ix=0; ix<rnx; ix++){
for(iz=0; iz<rnz; iz++){
bwf[wfit][ix][iz]=crealf(curr[(ix+nb)*fnz+(iz+nb)]);
ccr[ix][iz] += fwf[wfit][ix][iz]*bwf[wfit][ix][iz];
sill[ix][iz] += fwf[wfit][ix][iz]*fwf[wfit][ix][iz];
}
}
wfit--;
}
} //end of it
if(wfit != -1) sf_error("Check program! The final wfit should be -1!");
/* free storage */
free(*rt); free(rt);
free(*lt); free(lt);
free(*wave); free(wave);
free(left); free(right);
/* normalized image */
#ifdef _OPENMP
#pragma omp parallel for private(ix, iz)
#endif
for (ix=0; ix<rnx; ix++){
for(iz=0; iz<rnz; iz++){
mig2[ix][iz]=ccr[ix][iz]/(sill[ix][iz]+SF_EPS);
// sill[ix][iz]=0.;
}
}
/* time-shift imaging condition */
for(itau=0; itau<ntau; itau++){
//sf_warning("itau/ntau=%d/%d", itau+1, ntau);
tau=itau*dtau+tau0;
htau=tau/wfdt;
for(it=abs(htau); it<wfnt-abs(htau); it++){
#ifdef _OPENMP
#pragma omp parallel for private(ix, iz)
#endif
for(ix=0; ix<rnx; ix++){
for(iz=0; iz<rnz; iz++){
mig1[itau][ix][iz]+=fwf[it+htau][ix][iz]*bwf[it-htau][ix][iz];
// sill[ix][iz]+=fwf[it+htau][ix][iz]*fwf[it+htau][ix][iz];
} // end of iz
} // end of ix
} // end of it
//#ifdef _OPENMP
//#pragma omp parallel for private(ix, iz)
//#endif
/* source illumination */
// for(ix=0; ix<rnx; ix++){
// for(iz=0; iz<rnz; iz++){
// mig1[itau][ix][iz] = mig1[itau][ix][iz]/(sill[ix][iz]+SF_EPS);
// }
// }
} //end of itau
/* output wavefield snapshot */
if(wantwf){
for(it=0; it<wfnt; it++){
if(it%snap==0){
sf_floatwrite(fwf[it][0], rnzx, Ffwf);
sf_floatwrite(bwf[wfnt-1-it][0], rnzx, Fbwf);
}
}
sf_fileclose(Ffwf);
sf_fileclose(Fbwf);
}
/* add all the shot images that are on the same node */
#ifdef _OPENMP
#pragma omp parallel for private(itau, ix, iz)
#endif
for(itau=0; itau<ntau; itau++){
for(ix=0; ix<rnx; ix++){
for(iz=0; iz<rnz; iz++){
img1[itau][ix+is*nds][iz] += mig1[itau][ix][iz];
}
}
}
#ifdef _OPENMP
#pragma omp parallel for private(ix, iz)
#endif
for(ix=0; ix<rnx; ix++){
for(iz=0; iz<rnz; iz++){
img2[ix+is*nds][iz] += mig2[ix][iz];
}
}
} // end of is<ns
} // end of iturn
////////////////end of ishot
MPI_Barrier(comm);
cfft2_finalize();
sf_fileclose(Fdat);
free(ww); free(rr);
free(*dd); free(dd);
if(cpuid==0) {free(**dd3); free(*dd3); free(dd3);}
free(cwave); free(cwavem); free(curr);
free(*ccr); free(ccr);
free(*sill); free(sill);
free(**fwf); free(*fwf); free(fwf);
free(**bwf); free(*bwf); free(bwf);
free(**mig1); free(*mig1); free(mig1);
free(*mig2); free(mig2);
/* sum image */
if(cpuid==0){
sendbuf=(float *)MPI_IN_PLACE;
recvbuf=img1[0][0];
}else{
sendbuf=img1[0][0];
recvbuf=NULL;
}
MPI_Reduce(sendbuf, recvbuf, ntau*vnx*rnz, MPI_FLOAT, MPI_SUM, 0, comm);
if(cpuid==0){
sendbuf=MPI_IN_PLACE;
recvbuf=img2[0];
}else{
sendbuf=img2[0];
recvbuf=NULL;
}
MPI_Reduce(sendbuf, recvbuf, vnx*rnz, MPI_FLOAT, MPI_SUM, 0, comm);
/* output image */
if(cpuid==0){
sf_floatwrite(img1[0][0], ntau*vnx*rnz, Fimg1);
sf_floatwrite(img2[0], vnx*rnz, Fimg2);
}
MPI_Barrier(comm);
sf_fileclose(Fimg1);
sf_fileclose(Fimg2);
free(**img1); free(*img1); free(img1);
free(*img2); free(img2);
gettimeofday(&tim, NULL);
tend=tim.tv_sec+(tim.tv_usec/1000000.0);
sf_warning(">> The computing time is %.3lf minutes <<", (tend-tstart)/60.);
MPI_Finalize();
exit(0);
}
/* main function */
int main(int argc, char* argv[])
{
clock_t tstart,tend;
double duration;
/*flags*/
bool verb, adj; /* migration(adjoint) flag */
bool wantwf; /* outputs wavefield snapshots */
bool wantrecord; /* actually means "need record" */
bool illum; /* source illumination flag */
bool roll; /* survey strategy */
bool fm; /* forward modeling */
/*I/O*/
sf_file Fvel;
sf_file left, right, leftb, rightb;
sf_file Fsrc, Frcd/*source and record*/;
sf_file Ftmpwf;
sf_file Fimg;
sf_file mask;
/*axis*/
sf_axis at, ax, az, as;
/*grid index variables*/
int nx, nz, nt, wfnt;
int nzx, nx2, nz2, n2, m2, pad1, nk;
int ix, iz, it, is;
int nxb, nzb;
int snpint, wfint;
float dt, dx, dz, wfdt;
float ox, oz;
/*source/geophone location*/
int spx, spz;
int gpz,gpx,gpl; /*geophone depth/x-crd/length*/
/*Model*/
sf_complex **lt, **rt;
sf_complex **ltb, **rtb;
/*Data*/
sf_complex ***wavefld;
sf_complex ***record, **tmprec, **img, **imgsum;
float **sill;
/*source*/
sf_complex *ww;
float **rr;
int **kill=NULL;
int rectz,rectx,repeat; /*smoothing parameters*/
float trunc;
int sht0,shtbgn,shtend,shtnum,shtint;
/*abc boundary*/
int top,bot,lft,rht;
/*tmp*/
int tmpint;
/*parameter structs*/
geopar geop;
mpipar mpip;
/*MPI*/
int rank, nodes;
sf_complex *sendbuf, *recvbuf;
MPI_Init(&argc, &argv);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &nodes);
sf_init(argc, argv);
if(rank==0) sf_warning("nodes=%d",nodes);
if (!sf_getbool("verb", &verb)) verb=false; /*verbosity*/
if (!sf_getbool("adj", &adj)) adj=true; /*migration*/
if (!sf_getbool("wantwf", &wantwf)) wantwf=false; /*output forward and backward wavefield*/
if (!sf_getbool("wantrecord", &wantrecord)) wantrecord=true; /*if n, using record data generated by this program */
if (!sf_getbool("illum", &illum)) illum=false; /*if n, no source illumination applied */
if (!sf_getbool("roll", &roll)) roll=true; /*if n, receiver is independent of source location and gpl=nx*/
if (!sf_getbool("fm", &fm)) fm=false; /* if n, Born modelling */
if (!sf_getbool("incom", &incom)) incom=false; /* if n, use complete data */
/* source/receiver info */
if (!sf_getint("shtbgn", &shtbgn)) sf_error("Need shot starting location on grid!");
if (!sf_getint("sht0", &sht0)) sht0=shtbgn; /*actual shot origin on grid*/
if (!sf_getint("shtend", &shtend)) sf_error("Need shot ending location on grid!");
if (!sf_getint("shtint", &shtint)) sf_error("Need shot interval on grid!");
shtnum = (int)((shtend-shtbgn)/shtint) + 1;
if (!sf_getint("spz", &spz)) sf_error("Need source depth!");
if (!sf_getint("gpz", &gpz)) sf_error("Need receiver depth!");
if (roll) if (!sf_getint("gpl", &gpl)) sf_error("Need receiver length");
if (!sf_getint("snapinter", &snpint)) snpint=1; /* snap interval */
if (!sf_getint("wfint", &wfint)) wfint=50; /* snap interval */
/*--- parameters of source ---*/
if (!sf_getfloat("srctrunc", &trunc)) trunc=0.4;
if (!sf_getint("rectz", &rectz)) rectz=2;
if (!sf_getint("rectx", &rectx)) rectx=2;
if (!sf_getint("repeat", &repeat)) repeat=2;
/* abc parameters */
if (!sf_getint("top", &top)) top=40;
if (!sf_getint("bot", &bot)) bot=40;
if (!sf_getint("lft", &lft)) lft=40;
if (!sf_getint("rht", &rht)) rht=40;
/* simultaneous sources parameter */
if (!sf_getint("nsource", &nsource)) nsource=1;
if (!sf_getint("dsource", &dsource)) dsource=0;
if (!sf_getfloat("tdelay", &tdelay)) tdelay=0;
if (!sf_getint("choose", &choose)) choose=nsource;
/*Set I/O file*/
if (adj) { /* migration */
if (wantrecord) {
Frcd = sf_input("input"); /*record from elsewhere*/
Fsrc = sf_input("src"); /*source wavelet*/
} else {
Frcd = sf_output("rec"); /*record produced by forward modeling*/
Fsrc = sf_input("input"); /*source wavelet*/
}
Fimg = sf_output("output");
} else { /* modeling */
Fimg = sf_input("input");
Frcd = sf_output("output");
Fsrc = sf_input("src"); /*source wavelet*/
}
left = sf_input("left");
right = sf_input("right");
leftb = sf_input("leftb");
rightb = sf_input("rightb");
Fvel = sf_input("vel"); /*velocity - just for model dimension*/
if (wantwf) {
Ftmpwf = sf_output("tmpwf");/*wavefield snap*/
}
if (incom) {
mask=sf_input("mask"); /*mask operator*/
}
/*--- Axes parameters ---*/
at = sf_iaxa(Fsrc, 1);
nt = sf_n(at);
dt = sf_d(at);
az = sf_iaxa(Fvel, 1);
nzb = sf_n(az);
dz = sf_d(az);
oz = sf_o(az);
ax = sf_iaxa(Fvel, 2);
nxb = sf_n(ax);
dx = sf_d(ax);
ox = sf_o(ax);
nzx = nzb*nxb;
nz = nzb - top - bot;
nx = nxb - lft - rht;
if (!roll) gpl = nx; /* global survey setting */
/* wavefield axis */
wfnt = (int)(nt-1)/snpint+1;
wfdt = dt*snpint;
ndelay=tdelay/dt;
/* propagator matrices */
if (!sf_getint("pad1",&pad1)) pad1=1; /* padding factor on the first axis */
nz2 = kiss_fft_next_fast_size(nzb*pad1);
nx2 = kiss_fft_next_fast_size(nxb);
nk = nz2*nx2; /*wavenumber*/
if (!sf_histint(left,"n1",&n2) || n2 != nzx) sf_error("Need n1=%d in left",nzx);
if (!sf_histint(left,"n2",&m2)) sf_error("Need n2= in left");
if (!sf_histint(right,"n1",&n2) || n2 != m2) sf_error("Need n1=%d in right",m2);
if (!sf_histint(right,"n2",&n2) || n2 != nk) sf_error("Need n2=%d in right",nk);
/*check record data*/
if (adj && wantrecord) {
sf_histint(Frcd,"n1", &tmpint);
if (tmpint != nt ) sf_error("Error parameter n1 in record!");
sf_histint(Frcd,"n2", &tmpint);
if (tmpint != gpl ) sf_error("Error parameter n2 in record!");
sf_histint(Frcd,"n3", &tmpint);
if (tmpint != shtnum ) sf_error("Error parameter n3 in record!");
}
/*allocate memory*/
ww=sf_complexalloc(nt);
rr=sf_floatalloc2(nzx,nsource);
lt = sf_complexalloc2(nzx,m2);
rt = sf_complexalloc2(m2,nk);
ltb = sf_complexalloc2(nzx,m2);
rtb = sf_complexalloc2(m2,nk);
geop = (geopar) sf_alloc(1, sizeof(*geop));
mpip = (mpipar) sf_alloc(1, sizeof(*mpip));
tmprec = sf_complexalloc2(nt, gpl);
record = sf_complexalloc3(nt, gpl, shtnum);
if (incom) {
kill = sf_intalloc2(gpl, shtnum);
sf_intread(kill[0],gpl*shtnum,mask);
}
wavefld = sf_complexalloc3(nz, nx, wfnt);
if (illum ) sill = sf_floatalloc2(nz, nx);
else sill = NULL;
img = sf_complexalloc2(nz, nx);
if (adj) {
imgsum = sf_complexalloc2(nz, nx);
#ifdef _OPENMP
#pragma omp parallel for private(ix,iz)
#endif
for (ix=0; ix<nx; ix++)
for (iz=0; iz<nz; iz++) {
imgsum[ix][iz] = sf_cmplx(0.,0.);
}
}
/*read from files*/
sf_complexread(ww,nt,Fsrc);
sf_complexread(lt[0],nzx*m2,left);
sf_complexread(rt[0],m2*nk,right);
sf_complexread(ltb[0],nzx*m2,leftb);
sf_complexread(rtb[0],m2*nk,rightb);
if(!adj) sf_complexread(img[0],nx*nz,Fimg);
if(adj && wantrecord) {
sf_complexread(record[0][0], shtnum*gpl*nt, Frcd);
if(incom) {
for (is=0; is<shtnum; is++)
for (ix=0; ix<gpl; ix++)
if(kill[is][ix]==0)
for (it=0; it<nt; it++)
record[is][ix][it]=sf_cmplx(0.,0.);
}
} else {
#ifdef _OPENMP
#pragma omp parallel for private(is,ix,it)
#endif
for (is=0; is<shtnum; is++)
for (ix=0; ix<gpl; ix++)
for (it=0; it<nt; it++)
record[is][ix][it] = sf_cmplx(0.,0.);
}
/*close RSF files*/
sf_fileclose(Fsrc);
sf_fileclose(left);
sf_fileclose(right);
sf_fileclose(leftb);
sf_fileclose(rightb);
/*load constant geopar elements*/
mpip->cpuid=rank;
mpip->numprocs=nodes;
/*load constant geopar elements*/
geop->nx = nx;
geop->nz = nz;
geop->nxb = nxb;
geop->nzb = nzb;
geop->dx = dx;
geop->dz = dz;
geop->ox = ox;
geop->oz = oz;
geop->snpint = snpint;
geop->spz = spz;
geop->gpz = gpz;
geop->gpl = gpl;
geop->top = top;
geop->bot = bot;
geop->lft = lft;
geop->rht = rht;
geop->nt = nt;
geop->dt = dt;
geop->trunc = trunc;
geop->shtnum = shtnum;
/* output RSF files */
if (rank==0) {
sf_setn(ax, gpl);
sf_setn(az, nz);
as = sf_iaxa(Fvel, 2);
sf_setn(as,shtnum);
sf_setd(as,shtint*dx);
sf_seto(as,shtbgn*dx+ox);
if (adj) { /* migration */
if(!wantrecord) {
sf_oaxa(Frcd, at, 1);
sf_oaxa(Frcd, ax, 2);
sf_oaxa(Frcd, as, 3);
sf_settype(Frcd,SF_COMPLEX);
}
sf_setn(ax, nx);
/*write image*/
sf_oaxa(Fimg, az, 1);
sf_oaxa(Fimg, ax, 2);
sf_settype(Fimg,SF_COMPLEX);
} else { /* modeling */
sf_oaxa(Frcd, at, 1);
sf_oaxa(Frcd, ax, 2);
sf_oaxa(Frcd, as ,3);
sf_settype(Frcd,SF_COMPLEX);
}
if (wantwf) {
sf_setn(ax, nx);
/*write temp wavefield */
sf_setn(at, (wfnt-1)/wfint+1);
sf_setd(at, wfdt*wfint);
sf_oaxa(Ftmpwf, az, 1);
sf_oaxa(Ftmpwf, ax, 2);
sf_oaxa(Ftmpwf, at, 3);
sf_settype(Ftmpwf,SF_COMPLEX);
}
}
tstart = clock();
for (is=rank; is<shtnum; is+=nodes) {
spx = shtbgn + shtint*is;
if (roll)
gpx = spx - (int)(gpl/2);
else
gpx = 0;
geop->spx = spx;
geop->gpx = gpx;
if (verb) {
sf_warning("============================");
sf_warning("processing shot #%d", is);
sf_warning("nx=%d nz=%d nt=%d", geop->nx, geop->nz, geop->nt);
sf_warning("nxb=%d nzb=%d ", geop->nxb, geop->nzb);
sf_warning("dx=%f dz=%f dt=%f", geop->dx, geop->dz, geop->dt);
sf_warning("top=%d bot=%d lft=%d rht=%d", geop->top, geop->bot, geop->lft, geop->rht);
sf_warning("rectz=%d rectx=%d repeat=%d srctrunc=%f",rectz,rectx,repeat,geop->trunc);
sf_warning("spz=%d spx=%d gpz=%d gpx=%d gpl=%d", spz, spx, gpz, gpx, gpl);
sf_warning("snpint=%d wfdt=%f wfnt=%d ", snpint, wfdt, wfnt);
sf_warning("sht0=%d shtbgn=%d shtend=%d shtnum=%d", sht0, shtbgn, shtend, shtnum);
if (roll) sf_warning("Rolling survey!");
else sf_warning("Global survey (gpl=nx)!");
if (illum) sf_warning("Using source illumination!");
else sf_warning("No source illumination!");
sf_warning("============================");
}
/*generate reflectivity map*/
reflgen(nzb, nxb, spz+top, spx+lft, rectz, rectx, repeat, rr);
lrosfor2(wavefld, sill, tmprec, verb, lt, rt, m2, geop, ww, rr, pad1, illum);
if(adj && wantrecord)
#ifdef _OPENMP
#pragma omp parallel for private(ix,it)
#endif
for (ix=0; ix<gpl; ix++)
for (it=0; it<nt; it++)
tmprec[ix][it] = record[is][ix][it];
if(!fm) {
lrosback2(img, wavefld, sill, tmprec, adj, verb, wantwf, ltb, rtb, m2, geop, pad1, illum);
}
if (adj) {
#ifdef _OPENMP
#pragma omp parallel for private(ix,iz)
#endif
for (ix=0; ix<nx; ix++)
for (iz=0; iz<nz; iz++)
imgsum[ix][iz] += img[ix][iz];
}
if (!adj || !wantrecord)
#ifdef _OPENMP
#pragma omp parallel for private(ix,it)
#endif
for (ix=0; ix<gpl; ix++)
for (it=0; it<nt; it++)
record[is][ix][it] = tmprec[ix][it];
if (wantwf && is==0)
for (it=0; it<wfnt; it++) {
if (it%wfint == 0) {
sf_complexwrite(wavefld[it][0], nx*nz, Ftmpwf);
}
}
} /*shot iteration*/
MPI_Barrier(MPI_COMM_WORLD);
/*write record/image*/
if (adj) {
if (rank==0) {
sendbuf = (sf_complex *) MPI_IN_PLACE;
recvbuf = imgsum[0];
} else {
sendbuf = imgsum[0];
recvbuf = NULL;
}
MPI_Reduce(sendbuf, recvbuf, nx*nz, MPI_COMPLEX, MPI_SUM, 0, MPI_COMM_WORLD);
if (rank==0)
sf_complexwrite(imgsum[0], nx*nz, Fimg);
}
if (!adj || !wantrecord) {
if (rank==0) {
sendbuf = (sf_complex *) MPI_IN_PLACE;
recvbuf = record[0][0];
} else {
sendbuf = record[0][0];
recvbuf = NULL;
}
MPI_Reduce(sendbuf, recvbuf, shtnum*gpl*nt, MPI_COMPLEX, MPI_SUM, 0, MPI_COMM_WORLD);
if (rank==0) {
if(incom) {
for (is=0; is<shtnum; is++)
for (ix=0; ix<gpl; ix++)
if(kill[is][ix]==0)
for (it=0; it<nt; it++)
record[is][ix][it]=sf_cmplx(0.,0.);
}
sf_complexwrite(record[0][0], shtnum*gpl*nt, Frcd);
}
}
/*free memory*/
free(ww);
free(rr);
free(*lt);
free(lt);
free(*rt);
free(rt);
free(*ltb);
free(ltb);
free(*rtb);
free(rtb);
free(geop);
free(mpip);
free(*tmprec);
free(tmprec);
free(**record);
free(*record);
free(record);
free(**wavefld);
free(*wavefld);
free(wavefld);
if (illum) {
free(*sill);
free(sill);
}
free(*img);
free(img);
if (adj) {
free(*imgsum);
free(imgsum);
}
if (incom) {
free(* kill);
free(kill);
}
tend = clock();
duration=(double)(tend-tstart)/CLOCKS_PER_SEC;
sf_warning(">> The CPU time of single shot migration is: %f seconds << ", duration);
MPI_Finalize();
exit(0);
}
int main(int argc, char* argv[])
{
bool verb,fsrf,snap,expl,dabc,sout,uses;
int jsnap,ntsnap,jdata;
char *atype;
#ifdef _OPENMP
int ompnth=1;
#endif
/* I/O files */
sf_file Fwav=NULL; /* wavelet */
sf_file Fsou=NULL; /* sources */
sf_file Frec=NULL; /* receivers */
sf_file Fvel=NULL; /* velocity */
sf_file Fang=NULL; /* angles */
sf_file Fdat=NULL; /* data */
sf_file Fwfl=NULL; /* wavefield */
/* cube axes */
sf_axis at,az,ax;
sf_axis as,ar;
int nt,nz,nx,ns,nr,nb;
int it,iz,ix;
float dt,dz,dx,dt2;
/* FDM structure */
fdm2d fdm=NULL;
abcone2d abc=NULL;
sponge spo=NULL;
/* I/O arrays */
float *ww=NULL; /* wavelet */
pt2d *ss=NULL; /* sources */
pt2d *rr=NULL; /* receivers */
float *dd=NULL; /* data */
float **tt=NULL;
float **vp=NULL; /* velocity */
float **vpn=NULL;
float **vpz=NULL;
float **vpx=NULL;
float **vsz=NULL;
float **tht=NULL,**sit=NULL,**cot=NULL;
float st,ct;
float **pm=NULL,**po=NULL,**pp=NULL,**pa=NULL,**pt=NULL; /* main wavefield */
float **qm=NULL,**qo=NULL,**qp=NULL,**qa=NULL,**qt=NULL; /* auxiliary wavefield */
float **sf=NULL; /* "stress" field */
/* linear inteppolation weights/indices */
lint2d cs,cr;
/* FD operator size */
float cox,cax,cbx,c1x,c2x;
float coz,caz,cbz,c1z,c2z;
/* wavefield cut params */
sf_axis acz=NULL,acx=NULL;
int nqz,nqx;
float oqz,oqx;
float dqz,dqx;
float **pc=NULL;
float H1p,H2p,H1q,H2q;
/*------------------------------------------------------------*/
/* init RSF */
sf_init(argc,argv);
/* select anisotropy model */
if (NULL == (atype = sf_getstring("atype"))) atype = "i";
switch(atype[0]) {
case 't':
sf_warning("TTI model");
break;
case 'v':
sf_warning("VTI model");
break;
case 'i':
default:
sf_warning("ISO model");
break;
}
/*------------------------------------------------------------*/
/* OMP parameters */
#ifdef _OPENMP
ompnth=omp_init();
#endif
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
if(! sf_getbool("verb",&verb)) verb=false; /* verbosity */
if(! sf_getbool("snap",&snap)) snap=false; /* wavefield snapshots */
if(! sf_getbool("free",&fsrf)) fsrf=false; /* free surface */
if(! sf_getbool("expl",&expl)) expl=false; /* "exploding reflector" */
if(! sf_getbool("dabc",&dabc)) dabc=false; /* absorbing BC */
if(! sf_getbool("sout",&sout)) sout=false; /* stress output */
if(! sf_getbool("uses",&uses)) uses=false; /* use vsz */
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* I/O files */
Fwav = sf_input ("in" ); /* wavelet */
Fvel = sf_input ("vel"); /* velocity */
Fsou = sf_input ("sou"); /* sources */
Frec = sf_input ("rec"); /* receivers */
Fang = sf_input ("ang"); /* angles */
Fwfl = sf_output("wfl"); /* wavefield */
Fdat = sf_output("out"); /* data */
/*------------------------------------------------------------*/
/* axes */
at = sf_iaxa(Fwav,2); sf_setlabel(at,"t"); if(verb) sf_raxa(at); /* time */
az = sf_iaxa(Fvel,1); sf_setlabel(az,"z"); if(verb) sf_raxa(az); /* depth */
ax = sf_iaxa(Fvel,2); sf_setlabel(ax,"x"); if(verb) sf_raxa(ax); /* space */
as = sf_iaxa(Fsou,2); sf_setlabel(as,"s"); if(verb) sf_raxa(as); /* sources */
ar = sf_iaxa(Frec,2); sf_setlabel(ar,"r"); if(verb) sf_raxa(ar); /* receivers */
nt = sf_n(at); dt = sf_d(at);
nz = sf_n(az); dz = sf_d(az);
nx = sf_n(ax); dx = sf_d(ax);
ns = sf_n(as);
nr = sf_n(ar);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* other execution parameters */
if(! sf_getint("jdata",&jdata)) jdata=1;
if(snap) { /* save wavefield every *jsnap* time steps */
if(! sf_getint("jsnap",&jsnap)) jsnap=nt;
}
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* setup output data header */
sf_oaxa(Fdat,ar,1);
sf_setn(at,nt/jdata);
sf_setd(at,dt*jdata);
sf_oaxa(Fdat,at,2);
/* setup output wavefield header */
if(snap) {
if(!sf_getint ("nqz",&nqz)) nqz=sf_n(az);
if(!sf_getint ("nqx",&nqx)) nqx=sf_n(ax);
if(!sf_getfloat("oqz",&oqz)) oqz=sf_o(az);
if(!sf_getfloat("oqx",&oqx)) oqx=sf_o(ax);
dqz=sf_d(az);
dqx=sf_d(ax);
acz = sf_maxa(nqz,oqz,dqz); sf_raxa(acz);
acx = sf_maxa(nqx,oqx,dqx); sf_raxa(acx);
/* check if the imaging window fits in the wavefield domain */
pc=sf_floatalloc2(sf_n(acz),sf_n(acx));
ntsnap=0;
for(it=0; it<nt; it++) {
if(it%jsnap==0) ntsnap++;
}
sf_setn(at, ntsnap);
sf_setd(at,dt*jsnap);
if(verb) sf_raxa(at);
sf_oaxa(Fwfl,acz,1);
sf_oaxa(Fwfl,acx,2);
sf_oaxa(Fwfl,at, 3);
}
/*------------------------------------------------------------*/
/* expand domain for FD operators and ABC */
if( !sf_getint("nb",&nb) || nb<NOP) nb=NOP;
fdm=fdutil_init(verb,fsrf,az,ax,nb,1);
sf_setn(az,fdm->nzpad); sf_seto(az,fdm->ozpad); if(verb) sf_raxa(az);
sf_setn(ax,fdm->nxpad); sf_seto(ax,fdm->oxpad); if(verb) sf_raxa(ax);
/*------------------------------------------------------------*/
if(expl) {
ww = sf_floatalloc( 1);
} else {
ww = sf_floatalloc(ns);
}
dd = sf_floatalloc(nr);
/*------------------------------------------------------------*/
/* setup source/receiver coordinates */
ss = (pt2d*) sf_alloc(ns,sizeof(*ss));
rr = (pt2d*) sf_alloc(nr,sizeof(*rr));
pt2dread1(Fsou,ss,ns,2); /* read (x,z) coordinates */
pt2dread1(Frec,rr,nr,2); /* read (x,z) coordinates */
cs = lint2d_make(ns,ss,fdm);
cr = lint2d_make(nr,rr,fdm);
/*------------------------------------------------------------*/
/* setup FD coefficients */
cox = C0 / (dx*dx);
cax = CA / (dx*dx);
cbx = CB / (dx*dx);
c1x = C1 / dx;
c2x = C2 / dx;
coz = C0 / (dz*dz);
caz = CA / (dz*dz);
cbz = CB / (dz*dz);
c1z = C1 / dz;
c2z = C2 / dz;
/* precompute dt^2*/
dt2 = dt*dt;
/*------------------------------------------------------------*/
tt = sf_floatalloc2(nz,nx);
/*------------------------------------------------------------*/
/* input velocity */
vp =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
vpz =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sf_floatread(tt[0],nz*nx,Fvel );
expand(tt,vpz,fdm); /* VPz */
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
vp [ix][iz] = vpz[ix][iz];
vpz[ix][iz] = vpz[ix][iz] * vpz[ix][iz];
}
}
if(fsrf) { /* free surface */
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nb; iz++) {
vpz[ix][iz]=0;
}
}
}
if(atype[0] != 'i') {
vpn =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sf_floatread(tt[0],nz*nx,Fvel );
expand(tt,vpn,fdm); /* VPn */
vpx =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sf_floatread(tt[0],nz*nx,Fvel );
expand(tt,vpx,fdm); /* VPx */
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
vpn[ix][iz] = vpn[ix][iz] * vpn[ix][iz];
vpx[ix][iz] = vpx[ix][iz] * vpx[ix][iz];
}
}
if(fsrf) { /* free surface */
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nb; iz++) {
vpn[ix][iz]=0;
vpx[ix][iz]=0;
}
}
}
if(uses) {
vsz =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sf_floatread(tt[0],nz*nx,Fvel );
expand(tt,vsz,fdm); /* VSz */
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
vsz[ix][iz] = vsz[ix][iz] * vsz[ix][iz];
}
}
}
}
/*------------------------------------------------------------*/
if( atype[0]=='t') {
/* input tilt angle */
tht =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sit =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
cot =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sf_floatread(tt[0],nz*nx,Fang);
expand(tt,tht,fdm);
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
tht[ix][iz] *= SF_PI/180.;
sit[ix][iz] = sinf(tht[ix][iz]);
cot[ix][iz] = cosf(tht[ix][iz]);
}
}
free(*tht); free(tht);
}
/*------------------------------------------------------------*/
free(*tt); free(tt);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* allocate wavefield arrays */
pm=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
po=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
pp=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
pa=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
pm[ix][iz]=0;
po[ix][iz]=0;
pp[ix][iz]=0;
pa[ix][iz]=0;
}
}
if(atype[0] != 'i') {
qm=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
qo=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
qp=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
qa=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
qm[ix][iz]=0;
qo[ix][iz]=0;
qp[ix][iz]=0;
qa[ix][iz]=0;
}
}
if(sout) sf=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
}
/*------------------------------------------------------------*/
if(dabc) {
abc = abcone2d_make(NOP,dt,vp,fsrf,fdm); /* one-way */
spo = sponge_make(fdm->nb); /* sponge */
}
/*------------------------------------------------------------*/
/*
* MAIN LOOP
*/
/*------------------------------------------------------------*/
if(verb) fprintf(stderr,"\n");
for (it=0; it<nt; it++) {
if(verb) fprintf(stderr,"\b\b\b\b\b%d",it);
/* compute acceleration */
switch(atype[0]) {
case 't':
if(uses) {
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iz,H1p,H2p,H1q,H2q,st,ct) \
shared(fdm,pa,po,qa,qo, \
cox,cax,cbx,c1x,c2x, \
coz,caz,cbz,c1z,c2z, \
vpn,vpz,vpx,vsz, \
sit,cot)
#endif
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
st=sit[ix][iz];
ct=cot[ix][iz];
H1p = H1(po,ix,iz, \
st,ct, \
cox,cax,cbx,c1x,c2x, \
coz,caz,cbz,c1z,c2z);
H2p = H2(po,ix,iz, \
st,ct, \
cox,cax,cbx,c1x,c2x, \
coz,caz,cbz,c1z,c2z);
H1q = H1(qo,ix,iz, \
st,ct, \
cox,cax,cbx,c1x,c2x, \
coz,caz,cbz,c1z,c2z);
H2q = H2(qo,ix,iz, \
st,ct, \
cox,cax,cbx,c1x,c2x, \
coz,caz,cbz,c1z,c2z);
/* p - main field */
pa[ix][iz] =
H1p * vsz[ix][iz] +
H2p * vpx[ix][iz] +
H1q * vpz[ix][iz] -
H1q * vsz[ix][iz];
/* q - auxiliary field */
qa[ix][iz] =
H2p * vpn[ix][iz] -
H2p * vsz[ix][iz] +
H1q * vpz[ix][iz] +
H2q * vsz[ix][iz];
}
}
} else {
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iz,H2p,H1q,st,ct) \
shared(fdm,pa,po,qa,qo, \
cox,cax,cbx,c1x,c2x, \
coz,caz,cbz,c1z,c2z, \
vpn,vpz,vpx, \
sit,cot)
#endif
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
st=sit[ix][iz];
ct=cot[ix][iz];
H2p = H2(po,ix,iz, \
st,ct, \
cox,cax,cbx,c1x,c2x, \
coz,caz,cbz,c1z,c2z);
H1q = H1(qo,ix,iz, \
st,ct, \
cox,cax,cbx,c1x,c2x, \
coz,caz,cbz,c1z,c2z);
/* p - main field */
pa[ix][iz] =
H2p * vpx[ix][iz] +
H1q * vpz[ix][iz];
/* q - auxiliary field */
qa[ix][iz] =
H2p * vpn[ix][iz] +
H1q * vpz[ix][iz];
}
}
}
break;
case 'v':
if(uses) {
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iz,H1p,H2p,H1q,H2q) \
shared(fdm,pa,po,qa,qo, \
cox,cax,cbx, \
coz,caz,cbz, \
vpn,vpz,vpx,vsz)
#endif
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
H1p = Dzz(po,ix,iz,coz,caz,cbz);
H1q = Dzz(qo,ix,iz,coz,caz,cbz);
H2p = Dxx(po,ix,iz,cox,cax,cbx);
H2q = Dxx(qo,ix,iz,cox,cax,cbx);
/* p - main field */
pa[ix][iz] =
H1p * vsz[ix][iz] +
H2p * vpx[ix][iz] +
H1q * vpz[ix][iz] -
H1q * vsz[ix][iz];
/* q - auxiliary field */
qa[ix][iz] =
H2p * vpn[ix][iz] -
H2p * vsz[ix][iz] +
H1q * vpz[ix][iz] +
H2q * vsz[ix][iz];
}
}
} else {
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iz,H2p,H1q) \
shared(fdm,pa,po,qa,qo, \
cox,cax,cbx, \
coz,caz,cbz, \
vpn,vpx,vpz)
#endif
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
H1q = Dzz(qo,ix,iz,coz,caz,cbz);
H2p = Dxx(po,ix,iz,cox,cax,cbx);
/* p - main field */
pa[ix][iz] =
H2p * vpx[ix][iz] +
H1q * vpz[ix][iz];
/* q - auxiliary field */
qa[ix][iz] =
H2p * vpn[ix][iz] +
H1q * vpz[ix][iz];
}
}
}
break;
case 'i':
default:
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iz) \
shared(fdm,pa,po, \
cox,cax,cbx, \
coz,caz,cbz, \
vpz)
#endif
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
pa[ix][iz] = ( Dxx(po,ix,iz,cox,cax,cbx) +
Dzz(po,ix,iz,coz,caz,cbz) ) * vpz[ix][iz];
}
}
break;
}
/* inject acceleration source */
if(expl) {
sf_floatread(ww, 1,Fwav);
; lint2d_inject1(pa,ww[0],cs);
if(atype[0] != 'i') lint2d_inject1(qa,ww[0],cs);
} else {
sf_floatread(ww,ns,Fwav);
; lint2d_inject(pa,ww,cs);
if(atype[0] != 'i') lint2d_inject(qa,ww,cs);
}
/* step forward in time */
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iz) \
shared(fdm,pa,po,pm,pp,dt2)
#endif
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
pp[ix][iz] = 2*po[ix][iz]
- pm[ix][iz]
+ pa[ix][iz] * dt2;
}
}
/* circulate wavefield arrays */
pt=pm;
pm=po;
po=pp;
pp=pt;
if(atype[0] != 'i') {
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iz) \
shared(fdm,qa,qo,qm,qp,dt2)
#endif
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
qp[ix][iz] = 2*qo[ix][iz]
- qm[ix][iz]
+ qa[ix][iz] * dt2;
}
}
/* circulate wavefield arrays */
qt=qm;
qm=qo;
qo=qp;
qp=qt;
}
/* one-way abc apply */
if(dabc) {
abcone2d_apply(po,pm,NOP,abc,fdm);
sponge2d_apply(pm,spo,fdm);
sponge2d_apply(po,spo,fdm);
if(atype[0] != 'i') {
abcone2d_apply(qo,qm,NOP,abc,fdm);
sponge2d_apply(qm,spo,fdm);
sponge2d_apply(qo,spo,fdm);
}
}
/* compute stress */
if(sout && (atype[0] != 'i')) {
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iz) \
shared(fdm,po,qo,sf)
#endif
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
sf[ix][iz] = po[ix][iz] + qo[ix][iz];
}
}
}
/* extract data at receivers */
if(sout && (atype[0] != 'i')) {lint2d_extract(sf,dd,cr);
} else { lint2d_extract(po,dd,cr);}
if(it%jdata==0) sf_floatwrite(dd,nr,Fdat);
/* extract wavefield in the "box" */
if(snap && it%jsnap==0) {
if(sout && (atype[0] != 'i')) {cut2d(sf,pc,fdm,acz,acx);
} else { cut2d(po,pc,fdm,acz,acx);}
sf_floatwrite(pc[0],sf_n(acz)*sf_n(acx),Fwfl);
}
}
if(verb) fprintf(stderr,"\n");
/*------------------------------------------------------------*/
/* deallocate arrays */
free(*pm); free(pm);
free(*pp); free(pp);
free(*po); free(po);
free(*pa); free(pa);
free(*pc); free(pc);
free(*vp); free(vp);
free(*vpz); free(vpz);
if(atype[0] != 'i') {
free(*qm); free(qm);
free(*qp); free(qp);
free(*qo); free(qo);
free(*qa); free(qa);
free(*vpn); free(vpn);
free(*vpx); free(vpx);
if(uses){ free(*vsz); free(vsz); }
if(sout){ free(*sf); free(sf); }
}
if(atype[0] == 't') {
free(*sit); free(sit);
free(*cot); free(cot);
}
free(ww);
free(ss);
free(rr);
free(dd);
/*------------------------------------------------------------*/
exit (0);
}
int main(int argc, char* argv[])
{
bool verb; /* verbosity flag */
bool abc; /* absorbing boundary conditions flag */
bool free; /* free surface flag*/
bool snap; /* wavefield snapshots flag */
bool dens;
int jsnap;/* save wavefield every *jsnap* time steps */
/* I/O files */
sf_file Fw,Fs,Fr;
sf_file Fd,Fu;
sf_file Fv=NULL; /* velocity */
sf_file Fe=NULL; /* density */
/* cube axes */
sf_axis at,az,ax,as,ar;
int it,iz,ix,is,ir, iop;
int nt,nz,nx,ns,nr,nz2,nx2;
float z0,dz,x0,dx,idx,idz,dt,dt2;
/* arrays */
pt2d *ss, *rr; /* source/receiver locations */
float *ww=NULL; /* wavelet */
float *dd=NULL; /* data */
float **vv=NULL; /* velocity */
float **ee=NULL; /* density */
float *fzs,*fxs, *fzr,*fxr;
int *jzs,*jxs, *jzr,*jxr;
float *ws00,*ws01,*ws10,*ws11;
float *wr00,*wr01,*wr10,*wr11;
float **um,**uo,**up,**ud,**vp,**ro,**tt,**ut;
float *bzl,*bzh,*bxl,*bxh; /* boundary */
int nop=2; /* Laplacian operator size */
float c0, c1, c2; /* Laplacian operator coefficients */
float co,c1x,c2x,c1z,c2z;
int nbz,nbx; /* boundary size */
float tz, tx; /* sponge boundary decay coefficients */
float dp;
float ws; /* injected data */
int ompchunk; /* OpenMP data chunk size */
/*------------------------------------------------------------*/
/* init RSF */
sf_init(argc,argv);
if(! sf_getint("ompchunk",&ompchunk)) ompchunk=1;
if(! sf_getbool("verb",&verb)) verb=false;
if(! sf_getbool( "abc",&abc )) abc=false;
if(! sf_getbool("snap",&snap)) snap=false;
if(! sf_getbool("free",&free)) free=false;
if(! sf_getbool("dens",&dens)) dens=false;
Fw = sf_input ("in" ); /* wavelet */
Fv = sf_input ("vel"); /* velocity */
Fs = sf_input ("sou"); /* sources */
Fr = sf_input ("rec"); /* receivers */
Fu = sf_output("wfl"); /* wavefield */
Fd = sf_output("out"); /* data */
if(dens) Fe = sf_input("den"); /* density */
/* read axes*/
at=sf_iaxa(Fw,1); sf_setlabel(at,"t"); if(verb) sf_raxa(at); /* time */
az=sf_iaxa(Fv,1); sf_setlabel(az,"z"); if(verb) sf_raxa(az); /* depth */
ax=sf_iaxa(Fv,2); sf_setlabel(ax,"x"); if(verb) sf_raxa(ax); /* space */
as=sf_iaxa(Fs,2); sf_setlabel(as,"s"); if(verb) sf_raxa(as); /* source */
ar=sf_iaxa(Fr,2); sf_setlabel(ar,"r"); if(verb) sf_raxa(ar); /* receiver */
nt=sf_n(at); dt=sf_d(at);
nz=sf_n(az);
nx=sf_n(ax);
ns=sf_n(as);
nr=sf_n(ar);
/* configure wavefield snapshots */
if(snap) {
if(! sf_getint("jsnap",&jsnap)) jsnap=nt;
}
/*------------------------------------------------------------*/
/* expand domain for absorbing boundary conditions */
if(abc) {
if(! sf_getint("nbz",&nbz)) nbz=nop; if(nbz<nop) nbz=nop;
if(! sf_getint("nbx",&nbx)) nbx=nop; if(nbx<nop) nbx=nop;
if(! sf_getfloat("tz",&tz)) tz=0.025;
if(! sf_getfloat("tx",&tx)) tx=0.025;
} else {
nbz=nop;
nbx=nop;
}
/* expanded domain ( az+2 nz, ax+2 nx ) */
nz2=nz+2*nbz; dz=sf_d(az); z0=sf_o(az)-nbz*dz;
nx2=nx+2*nbx; dx=sf_d(ax); x0=sf_o(ax)-nbx*dx;
sf_setn(az,nz2); sf_seto(az,z0); if(verb) sf_raxa(az);
sf_setn(ax,nx2); sf_seto(ax,x0); if(verb) sf_raxa(ax);
/*------------------------------------------------------------*/
/* setup output data header */
sf_oaxa(Fd,ar,1);
sf_oaxa(Fd,at,2);
/* setup output wavefield header */
if(snap) {
sf_setn(at,nt/jsnap);
sf_setd(at,dt*jsnap);
sf_oaxa(Fu,az,1);
sf_oaxa(Fu,ax,2);
sf_oaxa(Fu,at,3);
}
/* Laplacian coefficients */
c0=-30./12.;
c1=+16./12.;
c2=- 1./12.;
dt2 = dt*dt;
idz = 1/dz;
idx = 1/dx;
co = c0 * (idx*idx+idz*idz);
c1x= c1 * idx*idx;
c2x= c2 * idx*idx;
c1z= c1 * idz*idz;
c2z= c2 * idz*idz;
/*------------------------------------------------------------*/
/* allocate arrays */
ww=sf_floatalloc (nt); sf_floatread(ww ,nt ,Fw);
vv=sf_floatalloc2(nz,nx); sf_floatread(vv[0],nz*nx,Fv);
ee=sf_floatalloc2(nz,nx);
if(dens) {
sf_floatread(ee[0],nz*nx,Fe);
} else {
for (iz=0; iz<nz; iz++) {
for (ix=0; ix<nx; ix++) {
ee[ix][iz]=1;
}
}
}
/* allocate source/receiver point arrays */
ss = (pt2d*) sf_alloc(ns,sizeof(*ss));
rr = (pt2d*) sf_alloc(nr,sizeof(*rr));
pt2dread1(Fs,ss,ns,3); /* read 3 elements (x,z,v) */
pt2dread1(Fr,rr,nr,2); /* read 2 elements (x,z) */
dd=sf_floatalloc(nr);
for(ir=0;ir<nr;ir++) {
dd[ir]=0;
}
jzs=sf_intalloc(ns); fzs=sf_floatalloc(ns);
jzr=sf_intalloc(nr); fzr=sf_floatalloc(nr);
jxs=sf_intalloc(ns); fxs=sf_floatalloc(ns);
jxr=sf_intalloc(nr); fxr=sf_floatalloc(nr);
ws00 = sf_floatalloc(ns); wr00 = sf_floatalloc(nr);
ws01 = sf_floatalloc(ns); wr01 = sf_floatalloc(nr);
ws10 = sf_floatalloc(ns); wr10 = sf_floatalloc(nr);
ws11 = sf_floatalloc(ns); wr11 = sf_floatalloc(nr);
/*------------------------------------------------------------*/
for (is=0;is<ns;is++) {
if(ss[is].z >= z0 &&
ss[is].z < z0 + (nz2-1)*dz &&
ss[is].x >= x0 &&
ss[is].x < x0 + (nx2-1)*dx) {
jzs[is] = (int)( (ss[is].z-z0)/dz);
fzs[is] = (ss[is].z-z0)/dz - jzs[is];
jxs[is] = (int)( (ss[is].x-x0)/dx);
fxs[is] = (ss[is].x-x0)/dx - jxs[is];
} else {
jzs[is] = 0; jxs[is] = 0;
fzs[is] = 1; fxs[is] = 0;
ss[is].v= 0;
}
ws00[is] = (1-fzs[is])*(1-fxs[is]);
ws01[is] = ( fzs[is])*(1-fxs[is]);
ws10[is] = (1-fzs[is])*( fxs[is]);
ws11[is] = ( fzs[is])*( fxs[is]);
}
for (ir=0;ir<nr;ir++) {
if(rr[ir].z >= z0 &&
rr[ir].z < z0 + (nz2-1)*dz &&
rr[ir].x >= x0 &&
rr[ir].x < x0 + (nx2-1)*dx) {
jzr[ir] = (int)( (rr[ir].z-z0)/dz);
fzr[ir] = (rr[ir].z-z0)/dz - jzr[ir];
jxr[ir] = (int)( (rr[ir].x-x0)/dx);
fxr[ir] = (rr[ir].x-x0)/dx - jxr[ir];
rr[ir].v=1;
} else {
jzr[ir] = 0;
fzr[ir] = 1;
rr[ir].v= 0;
}
wr00[ir] = (1-fzr[ir])*(1-fxr[ir]);
wr01[ir] = ( fzr[ir])*(1-fxr[ir]);
wr10[ir] = (1-fzr[ir])*( fxr[ir]);
wr11[ir] = ( fzr[ir])*( fxr[ir]);
}
/*------------------------------------------------------------*/
/* allocate temporary arrays */
um=sf_floatalloc2(nz2,nx2);
uo=sf_floatalloc2(nz2,nx2);
up=sf_floatalloc2(nz2,nx2);
ud=sf_floatalloc2(nz2,nx2);
tt=sf_floatalloc2(nz2,nx2);
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(iz,ix) shared(nx2,nz2,um,uo,up,ud,tt)
#endif
for (iz=0; iz<nz2; iz++) {
for (ix=0; ix<nx2; ix++) {
um[ix][iz]=0;
uo[ix][iz]=0;
up[ix][iz]=0;
ud[ix][iz]=0;
tt[ix][iz]=1;
}
}
/*------------------------------------------------------------*/
/* velocity in the expanded domain (vp=vv^2)*/
vp=sf_floatalloc2(nz2,nx2);
ro=sf_floatalloc2(nz2,nx2);
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(iz,ix) shared(nz,nx,vp,ro,vv,ee)
#endif
for (iz=0; iz<nz; iz++) {
for (ix=0; ix<nx; ix++) {
vp[nbx+ix][nbz+iz] = vv[ix][iz] * vv[ix][iz];
ro[nbx+ix][nbz+iz] = ee[ix][iz];
}
}
/* fill boundaries */
for (iz=0; iz<nbz; iz++) {
for (ix=0; ix<nx2; ix++) {
vp[ix][ iz ] = vp[ix][ nbz ];
vp[ix][nz2-iz-1] = vp[ix][nz2-nbz-1];
ro[ix][ iz ] = ro[ix][ nbz ];
ro[ix][nz2-iz-1] = ro[ix][nz2-nbz-1];
}
}
for (iz=0; iz<nz2; iz++) {
for (ix=0; ix<nbx; ix++) {
vp[ ix ][iz] = vp[ nbx ][iz];
vp[nx2-ix-1][iz] = vp[nx2-nbx-1][iz];
ro[ ix ][iz] = ro[ nbx ][iz];
ro[nx2-ix-1][iz] = ro[nx2-nbx-1][iz];
}
}
/*------------------------------------------------------------*/
/* free surface */
if(abc && free) {
for (iz=0; iz<nbz; iz++) {
for (ix=0; ix<nx2; ix++) {
vp[ix][iz]=0;
}
}
}
/*------------------------------------------------------------*/
/* sponge ABC setup */
if(abc) {
for (iz=0; iz<nbz; iz++) {
for (ix=0; ix<nx2; ix++) {
tt[ix][ iz ] = exp( - (tz*(nbz-iz))*(tz*(nbz-iz)) );
tt[ix][nz2-iz-1] = tt[ix][iz];
}
}
for (iz=0; iz<nz2; iz++) {
for (ix=0; ix<nbx; ix++) {
tt[ ix ][iz] = exp( - (tx*(nbx-ix))*(tx*(nbx-ix)) );
tt[nx2-ix-1][iz] = tt[ix][iz];
}
}
}
/* one-way ABC setup */
bzl=sf_floatalloc(nx2);
bzh=sf_floatalloc(nx2);
bxl=sf_floatalloc(nz2);
bxh=sf_floatalloc(nz2);
for (ix=0;ix<nx2;ix++) {
dp = vp[ix][ nop ] *dt/dz; bzl[ix] = (1-dp)/(1+dp);
dp = vp[ix][nz2-nop-1] *dt/dz; bzh[ix] = (1-dp)/(1+dp);
}
for (iz=0;iz<nz2;iz++) {
dp = vp[ nop ][iz] *dt/dx; bxl[iz] = (1-dp)/(1+dp);
dp = vp[nx2-nop-1][iz] *dt/dx; bxh[iz] = (1-dp)/(1+dp);
}
/*------------------------------------------------------------*/
/*
* MAIN LOOP
*/
if(verb) fprintf(stderr,"\n");
for (it=0; it<nt; it++) {
if(verb) fprintf(stderr,"\b\b\b\b\b%d",it);
if(dens) { /* variable density */
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(iz,ix) shared(nop,nx2,nz2,ud,uo,ro,co,c1x,c1z,c2x,c2z,idx,idz)
#endif
for( ix=nop; ix<nx2-nop; ix++) {
for(iz=nop; iz<nz2-nop; iz++) {
/* 4th order Laplacian operator */
ud[ix][iz] =
co * uo[ix ][iz ] +
c1x*(uo[ix-1][iz ] + uo[ix+1][iz ]) +
c2x*(uo[ix-2][iz ] + uo[ix+2][iz ]) +
c1z*(uo[ix ][iz-1] + uo[ix ][iz+1]) +
c2z*(uo[ix ][iz-2] + uo[ix ][iz+2]);
/* density terms */
ud[ix][iz] -= (
D1(uo,ix,iz,idz) * D1(ro,ix,iz,idz) +
D2(uo,ix,iz,idx) * D2(ro,ix,iz,idx) ) / ro[ix][iz];
}
}
} else { /* constant density */
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(iz,ix) shared(nop,nx2,nz2,ud,uo,co,c1x,c1z,c2x,c2z)
#endif
for( ix=nop; ix<nx2-nop; ix++) {
for(iz=nop; iz<nz2-nop; iz++) {
/* 4th order Laplacian operator */
ud[ix][iz] =
co * uo[ix ][iz ] +
c1x*(uo[ix-1][iz ] + uo[ix+1][iz ]) +
c2x*(uo[ix-2][iz ] + uo[ix+2][iz ]) +
c1z*(uo[ix ][iz-1] + uo[ix ][iz+1]) +
c2z*(uo[ix ][iz-2] + uo[ix ][iz+2]);
}
}
}
/* inject wavelet */
for (is=0;is<ns;is++) {
ws = ww[it] * ss[is].v;
ud[ jxs[is] ][ jzs[is] ] -= ws * ws00[is];
ud[ jxs[is] ][ jzs[is]+1] -= ws * ws01[is];
ud[ jxs[is]+1][ jzs[is] ] -= ws * ws10[is];
ud[ jxs[is]+1][ jzs[is]+1] -= ws * ws11[is];
}
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(ix,iz) shared(nx2,nz2,ud,uo,um,up,vp,dt2)
#endif
for( ix=0; ix<nx2; ix++) {
for(iz=0; iz<nz2; iz++) {
/* time step and velocity scale*/
up[ix][iz] = 2*uo[ix][iz] -
um[ix][iz] +
ud[ix][iz] * vp[ix][iz] * dt2;
}
}
/* circulate arrays */
ut=um;
um=uo;
uo=up;
up=ut;
/* one-way ABC apply */
if(abc) {
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(ix,iz,iop) shared(nx2,nz2,nop,uo,um,bzl,bzh)
#endif
for(ix=0;ix<nx2;ix++) {
for(iop=0;iop<nop;iop++) {
iz = nop-iop;
uo [ix][iz ]
= um[ix][iz+1]
+(um[ix][iz ]
- uo[ix][iz+1]) * bzl[ix];
iz = nz2-nop+iop-1;
uo [ix][iz ]
= um[ix][iz-1]
+(um[ix][iz ]
- uo[ix][iz-1]) * bzh[ix];
}
}
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,1) private(ix,iz,iop) shared(nx2,nz2,nop,uo,um,bzl,bzh)
#endif
for(iop=0;iop<nop;iop++) {
for(iz=0;iz<nz2;iz++) {
ix = nop-iop;
uo [ix ][iz]
= um[ix+1][iz]
+(um[ix ][iz]
- uo[ix+1][iz]) * bxl[iz];
ix = nx2-nop+iop-1;
uo [ix ][iz]
= um[ix-1][iz]
+(um[ix ][iz]
- uo[ix-1][iz]) * bxh[iz];
}
}
}
/* sponge ABC apply */
if(abc) {
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,ompchunk) private(ix,iz) shared(nx2,nz2,uo,um,ud,tt)
#endif
for( ix=0; ix<nx2; ix++) {
for(iz=0; iz<nz2; iz++) {
uo[ix][iz] *= tt[ix][iz];
um[ix][iz] *= tt[ix][iz];
ud[ix][iz] *= tt[ix][iz];
}
}
}
/* write wavefield */
if(snap && it%jsnap==0) {
sf_floatwrite(uo[0],nz2*nx2,Fu);
}
/* collect data */
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic,1) private(ir) shared(dd,rr,uo,jzr,wr00,wr01,wr10,wr11)
#endif
for (ir=0;ir<nr;ir++) {
dd[ir] =
uo[ jxr[ir] ][ jzr[ir] ] * wr00[ir] +
uo[ jxr[ir] ][ jzr[ir]+1] * wr01[ir] +
uo[ jxr[ir]+1][ jzr[ir] ] * wr10[ir] +
uo[ jxr[ir]+1][ jzr[ir]+1] * wr11[ir];
dd[ir] *= rr[ir].v;
}
/* write data */
sf_floatwrite(dd,nr,Fd);
}
if(verb) fprintf(stderr,"\n");
exit (0);
}
int main(int argc, char* argv[])
{
bool verb,fsrf,snap,expl,dabc;
int jsnap,ntsnap,jdata;
/* OMP parameters */
#ifdef _OPENMP
int ompnth;
#endif
/* I/O files */
sf_file Fwav=NULL; /* wavelet */
sf_file Fsou=NULL; /* sources */
sf_file Frec=NULL; /* receivers */
sf_file Fvel=NULL; /* velocity */
sf_file Fdat=NULL; /* data */
sf_file Fwfl=NULL; /* wavefield */
/* cube axes */
sf_axis at,az,ax;
sf_axis as,ar;
int nt,nz,nx,ns,nr,nb;
int it,iz,ix;
float dt,dz,dx,idz,idx,dt2;
/* FDM structure */
fdm2d fdm=NULL;
abcone2d abc=NULL;
sponge spo=NULL;
/* I/O arrays */
float *ww=NULL; /* wavelet */
pt2d *ss=NULL; /* sources */
pt2d *rr=NULL; /* receivers */
float *dd=NULL; /* data */
float **tt=NULL;
float **vp=NULL; /* velocity */
float **vp2=NULL;
float **vv2=NULL;
float **vh2=NULL;
float **rm,**ro,**rp,**ra,**rt; /* main wavefield */
float **qm,**qo,**qp,**qa,**qt; /* auxiliary wavefield */
/* linear interpolation weights/indices */
lint2d cs,cr;
/* FD operator size */
float cox,coz,cax,cbx,caz,cbz;
/* wavefield cut params */
sf_axis acz=NULL,acx=NULL;
int nqz,nqx;
float oqz,oqx;
float dqz,dqx;
float **qc=NULL;
float H2q,H1r;
/*------------------------------------------------------------*/
/* init RSF */
sf_init(argc,argv);
/*------------------------------------------------------------*/
/* OMP parameters */
#ifdef _OPENMP
ompnth=omp_init();
#endif
/*------------------------------------------------------------*/
if(! sf_getbool("verb",&verb)) verb=false; /* verbosity flag */
if(! sf_getbool("snap",&snap)) snap=false; /* wavefield snapshots flag */
if(! sf_getbool("free",&fsrf)) fsrf=false; /* free surface flag */
if(! sf_getbool("expl",&expl)) expl=false; /* "exploding reflector" */
if(! sf_getbool("dabc",&dabc)) dabc=false; /* absorbing BC */
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* I/O files */
Fwav = sf_input ("in" ); /* wavelet */
Fvel = sf_input ("vel"); /* velocity */
Fsou = sf_input ("sou"); /* sources */
Frec = sf_input ("rec"); /* receivers */
Fwfl = sf_output("wfl"); /* wavefield */
Fdat = sf_output("out"); /* data */
/*------------------------------------------------------------*/
/* axes */
at = sf_iaxa(Fwav,2); sf_setlabel(at,"t"); if(verb) sf_raxa(at); /* time */
az = sf_iaxa(Fvel,1); sf_setlabel(az,"z"); if(verb) sf_raxa(az); /* depth */
ax = sf_iaxa(Fvel,2); sf_setlabel(ax,"x"); if(verb) sf_raxa(ax); /* space */
as = sf_iaxa(Fsou,2); sf_setlabel(as,"s"); if(verb) sf_raxa(as); /* sources */
ar = sf_iaxa(Frec,2); sf_setlabel(ar,"r"); if(verb) sf_raxa(ar); /* receivers */
nt = sf_n(at); dt = sf_d(at);
nz = sf_n(az); dz = sf_d(az);
nx = sf_n(ax); dx = sf_d(ax);
ns = sf_n(as);
nr = sf_n(ar);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* other execution parameters */
if(! sf_getint("jdata",&jdata)) jdata=1;
if(snap) { /* save wavefield every *jsnap* time steps */
if(! sf_getint("jsnap",&jsnap)) jsnap=nt;
}
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* expand domain for FD operators and ABC */
if( !sf_getint("nb",&nb) || nb<NOP) nb=NOP;
fdm=fdutil_init(verb,fsrf,az,ax,nb,1);
sf_setn(az,fdm->nzpad); sf_seto(az,fdm->ozpad); if(verb) sf_raxa(az);
sf_setn(ax,fdm->nxpad); sf_seto(ax,fdm->oxpad); if(verb) sf_raxa(ax);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* setup output data header */
sf_oaxa(Fdat,ar,1);
sf_setn(at,nt/jdata);
sf_setd(at,dt*jdata);
sf_oaxa(Fdat,at,2);
/* setup output wavefield header */
if(snap) {
if(!sf_getint ("nqz",&nqz)) nqz=sf_n(az);
if(!sf_getint ("nqx",&nqx)) nqx=sf_n(ax);
if(!sf_getfloat("oqz",&oqz)) oqz=sf_o(az);
if(!sf_getfloat("oqx",&oqx)) oqx=sf_o(ax);
dqz=sf_d(az);
dqx=sf_d(ax);
acz = sf_maxa(nqz,oqz,dqz); sf_raxa(acz);
acx = sf_maxa(nqx,oqx,dqx); sf_raxa(acx);
/* check if the imaging window fits in the wavefield domain */
qc=sf_floatalloc2(sf_n(acz),sf_n(acx));
ntsnap=0;
for(it=0; it<nt; it++) {
if(it%jsnap==0) ntsnap++;
}
sf_setn(at, ntsnap);
sf_setd(at,dt*jsnap);
if(verb) sf_raxa(at);
sf_oaxa(Fwfl,acz,1);
sf_oaxa(Fwfl,acx,2);
sf_oaxa(Fwfl,at, 3);
}
if(expl) {
ww = sf_floatalloc( 1);
} else {
ww = sf_floatalloc(ns);
}
dd = sf_floatalloc(nr);
/*------------------------------------------------------------*/
/* setup source/receiver coordinates */
ss = (pt2d*) sf_alloc(ns,sizeof(*ss));
rr = (pt2d*) sf_alloc(nr,sizeof(*rr));
pt2dread1(Fsou,ss,ns,2); /* read (x,z) coordinates */
pt2dread1(Frec,rr,nr,2); /* read (x,z) coordinates */
cs = lint2d_make(ns,ss,fdm);
cr = lint2d_make(nr,rr,fdm);
/*------------------------------------------------------------*/
/* setup FD coefficients */
idz = 1/dz;
idx = 1/dx;
cox= C0 * (idx*idx);
cax= CA * idx*idx;
cbx= CB * idx*idx;
coz= C0 * (idz*idz);
caz= CA * idz*idz;
cbz= CB * idz*idz;
/* precompute dt^2*/
dt2 = dt*dt;
/*------------------------------------------------------------*/
tt = sf_floatalloc2(nz,nx);
/*------------------------------------------------------------*/
/* input velocity */
vp =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
vp2 =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
vv2 =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
vh2 =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
sf_floatread(tt[0],nz*nx,Fvel ); expand(tt,vv2,fdm); /* vertical v */
sf_floatread(tt[0],nz*nx,Fvel ); expand(tt,vp2,fdm); /* NMO v */
sf_floatread(tt[0],nz*nx,Fvel ); expand(tt,vh2,fdm); /* horizontal v */
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
vp2[ix][iz] = vp2[ix][iz] * vp2[ix][iz];
vv2[ix][iz] = vv2[ix][iz] * vv2[ix][iz];
vh2[ix][iz] = vh2[ix][iz] * vh2[ix][iz];
}
}
if(fsrf) { /* free surface */
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nb; iz++) {
vp2[ix][iz]=0;
vv2[ix][iz]=0;
vh2[ix][iz]=0;
}
}
}
/*------------------------------------------------------------*/
free(*tt); free(tt);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* allocate wavefield arrays */
qm=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
qo=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
qp=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
qa=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
rm=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
ro=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
rp=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
ra=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
qm[ix][iz]=0;
qo[ix][iz]=0;
qp[ix][iz]=0;
qa[ix][iz]=0;
rm[ix][iz]=0;
ro[ix][iz]=0;
rp[ix][iz]=0;
ra[ix][iz]=0;
}
}
/*------------------------------------------------------------*/
if(dabc) {
/* one-way abc setup */
abc = abcone2d_make(NOP,dt,vp,fsrf,fdm);
/* sponge abc setup */
spo = sponge_make(fdm->nb);
}
/*------------------------------------------------------------*/
/*
* MAIN LOOP
*/
/*------------------------------------------------------------*/
if(verb) fprintf(stderr,"\n");
for (it=0; it<nt; it++) {
if(verb) fprintf(stderr,"\b\b\b\b\b%d",it);
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iz,H2q,H1r) \
shared(fdm,ra,ro,qa,qo,cox,coz,cax,caz,cbx,cbz,idx,idz,vp2)
#endif
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
H2q = Dxx(qo,ix,iz,cox,cax,cbx);
H1r = Dzz(ro,ix,iz,coz,caz,cbz);
/* main field - q */
qa[ix][iz] = H2q * vh2[ix][iz] + H1r * vv2[ix][iz] ;
/* auxiliary field - r */
ra[ix][iz] = H2q * vp2[ix][iz] + H1r * vv2[ix][iz] ;
}
}
/* inject acceleration source */
if(expl) {
sf_floatread(ww, 1,Fwav);
lint2d_inject1(ra,ww[0],cs);
lint2d_inject1(qa,ww[0],cs);
} else {
sf_floatread(ww,ns,Fwav);
lint2d_inject(ra,ww,cs);
lint2d_inject(qa,ww,cs);
}
/* step forward in time */
#ifdef _OPENMP
#pragma omp parallel for \
schedule(dynamic,fdm->ompchunk) \
private(ix,iz) \
shared(fdm,ra,ro,rm,rp,qa,qo,qm,qp,dt2)
#endif
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
qp[ix][iz] = 2*qo[ix][iz]
- qm[ix][iz]
+ qa[ix][iz] * dt2;
rp[ix][iz] = 2*ro[ix][iz]
- rm[ix][iz]
+ ra[ix][iz] * dt2;
}
}
/* circulate wavefield arrays */
qt=qm;
qm=qo;
qo=qp;
qp=qt;
rt=rm;
rm=ro;
ro=rp;
rp=rt;
if(dabc) {
/* one-way abc apply */
abcone2d_apply(qo,qm,NOP,abc,fdm);
sponge2d_apply(qm,spo,fdm);
sponge2d_apply(qo,spo,fdm);
sponge2d_apply(qp,spo,fdm);
/* one-way abc apply */
abcone2d_apply(ro,rm,NOP,abc,fdm);
sponge2d_apply(rm,spo,fdm);
sponge2d_apply(ro,spo,fdm);
sponge2d_apply(rp,spo,fdm);
}
/* extract data */
lint2d_extract(qo,dd,cr);
if(snap && it%jsnap==0) {
cut2d(qo,qc,fdm,acz,acx);
sf_floatwrite(qc[0],sf_n(acz)*sf_n(acx),Fwfl);
}
if( it%jdata==0)
sf_floatwrite(dd,nr,Fdat);
}
if(verb) fprintf(stderr,"\n");
/*------------------------------------------------------------*/
/* deallocate arrays */
free(*rm); free(rm);
free(*rp); free(rp);
free(*ro); free(ro);
free(*ra); free(ra);
free(*qm); free(qm);
free(*qp); free(qp);
free(*qo); free(qo);
free(*qa); free(qa);
free(*qc); free(qc);
free(*vp); free(vp);
free(*vp2); free(vp2);
free(*vh2); free(vh2);
free(*vv2); free(vv2);
free(ww);
free(ss);
free(rr);
free(dd);
exit (0);
}
int main(int argc, char* argv[])
{
int jt, jtm, is, st; /* index */
int nt, n1, n2, ns, ng, m[SF_MAX_DIM], nm; /* dimensions */
float ot, o1, d1; /* original */
sf_file in, dat, wfl, vel, sgrid, ggrid; /* I/O files */
float **wvlt;
sf_axis ax;
HVel hv;
HCoord hs, hg;
bool verb;
sf_init(argc,argv);
in = sf_input ("in");
/* source wavelet \n
\t nt X 1: regular shot gather
\t nt X ns: simultaneous source shotting
*/
vel = sf_input ("vel"); /* velocity field */
sgrid = sf_input ("sgrid"); /* source grid */
ggrid = sf_input ("ggrid"); /* geophone grid */
dat = sf_output("out"); /* seismic data */
if(sf_getstring("wfl")!=NULL)
wfl = sf_output("wfl"); /* wavefield movie file */
else wfl=NULL;
if (!sf_getint("jt",&jt)) jt=1;
/* time interval in observation system */
if (!sf_getint("jtm",&jtm)) jtm=100;
/* time interval of wave movie */
if (!sf_getfloat("ot", &ot)) ot = 0.0;
/* time delay */
if (!sf_getbool("verb", &verb)) verb = false;
/* verbosity */
/* velocity and observation system */
hv = obs_vel(vel);
m[0] = sf_n(hv->z); nm = 1;
if(hv->nd >= 2) {m[1] = sf_n(hv->x); nm=2;}
if(hv->nd >= 3) {m[2] = sf_n(hv->y); nm=3;}
hs = obs_coord(sgrid, m, nm);
hg = obs_coord(ggrid, m, nm);
ns = sf_n(hs->a2);
ng = sf_n(hg->a2);
/* waveform */
ax = sf_iaxa(in, 1);
n1 = sf_n(ax); o1 = sf_o(ax); d1 = sf_d(ax);
if(!sf_histint(in, "n2", &n2) || n2 != ns) n2=1;
wvlt = sf_floatalloc2(n1, n2);
sf_floatread(wvlt[0], n1*n2, in);
if(ot<o1) ot=o1;
st = (ot-o1)/d1;
nt = (n1-st+1)/jt;
sf_setn(ax, nt);
sf_setd(ax, d1*jt);
sf_seto(ax, ot);
sf_oaxa(dat, ax, 1);
sf_oaxa(dat, hg->a2, 2);
if(n2==1) sf_oaxa(dat, hs->a2, 3);
if(wfl!=NULL)
{
sf_oaxa(wfl, hv->z, 1);
if(hv->nd >= 2) sf_oaxa(wfl, hv->x, 2);
if(hv->nd >= 3) sf_oaxa(wfl, hv->y, 3);
sf_setn(ax, (n1-st+1)/jtm);
sf_setd(ax, d1*jtm);
sf_seto(ax, ot);
sf_oaxa(wfl, ax, hv->nd+1);
if(n2==1) sf_oaxa(wfl, hs->a2, hv->nd+2);
}
wavmod_init(hv, d1, n1, st, jt, jtm, hg->p, ng, verb);
if(n2==1)
for (is=0; is < ns; is++)
{
wavmod_shot(dat, wfl, 1, hs->p+is, wvlt);
if(verb) sf_warning("shot %d of %d", is, ns/n2);
}
else
wavmod_shot(dat, wfl, ns, hs->p, wvlt);
wavmod_close();
free(wvlt[0]);
free(wvlt);
return (0);
}
int main(int argc, char* argv[])
{
bool verb,fsrf,snap,dabc;
int jsnap,ntsnap;
int jdata;
/* I/O files */
sf_file Fwav=NULL; /* wavelet */
sf_file Fsou=NULL; /* sources */
sf_file Frec=NULL; /* receivers */
sf_file Fvel=NULL; /* velocity */
sf_file Fref=NULL; /* reflectivity */
sf_file Fden=NULL; /* density */
sf_file Fdat=NULL; /* data (background) */
sf_file Fwfl=NULL; /* wavefield (background) */
sf_file Flid=NULL; /* data (scattered) */
sf_file Fliw=NULL; /* wavefield (scattered) */
/* I/O arrays */
float *ww=NULL; /* wavelet */
pt3d *ss=NULL; /* sources */
pt3d *rr=NULL; /* receivers */
float ***vpin=NULL; /* velocity */
float ***roin=NULL; /* density */
float ***rfin=NULL; /* reflectivity */
float ***vp=NULL; /* velocity in expanded domain */
float ***ro=NULL; /* density in expanded domain */
float ***iro=NULL; /* buoyancy in the expanded domain */
float ***rf=NULL; /* reflectivity in expanded domain */
float *bdd=NULL; /* data (background) */
float *sdd=NULL; /* data (scattered) */
float ***vt=NULL; /* temporary vp*vp * dt*dt */
float ***fsrfbck=NULL; /* ghost cells for free surface BC */
float ***fsrfsct=NULL; /* ghost cells for free surface BC */
float ***bum,***buo,***bup,***bua,***buat,***but; /* wavefield: um = U @ t-1; uo = U @ t; up = U @ t+1 */
float ***sum,***suo,***sup,***sua,***suat,***sut; /* wavefield: um = U @ t-1; uo = U @ t; up = U @ t+1 */
/* cube axes */
sf_axis at,a1,a2,a3,as,ar;
int nt,n1,n2,n3,ns,nr,nb;
int it,i1,i2,i3;
float dt,d1,d2,d3,id1,id2,id3,dt2;
/* linear interpolation weights/indices */
lint3d cs,cr;
fdm3d fdm;
abcone3d abc; /* abc */
sponge spo;
/* FD coefficients */
float c1x,c1y,c1z,
c2x,c2y,c2z,
c3x,c3y,c3z;
int ompchunk;
#ifdef _OPENMP
int ompnth,ompath;
#endif
sf_axis ac1=NULL,ac2=NULL,ac3=NULL;
int nqz,nqx,nqy;
float oqz,oqx,oqy;
float dqz,dqx,dqy;
float ***uc=NULL;
/* for benchmarking */
clock_t start_t, end_t;
float total_t;
/*------------------------------------------------------------*/
/* init RSF */
sf_init(argc,argv);
if(! sf_getint("ompchunk",&ompchunk)) ompchunk=1;
/* OpenMP data chunk size */
#ifdef _OPENMP
if(! sf_getint("ompnth", &ompnth)) ompnth=0;
/* OpenMP available threads */
#pragma omp parallel
ompath=omp_get_num_threads();
if(ompnth<1) ompnth=ompath;
omp_set_num_threads(ompnth);
sf_warning("using %d threads of a total of %d",ompnth,ompath);
#endif
if(! sf_getbool("verb",&verb)) verb=false; /* verbosity flag */
if(! sf_getbool("snap",&snap)) snap=false; /* wavefield snapshots flag */
if(! sf_getbool("dabc",&dabc)) dabc=false; /* Absorbing BC */
if(! sf_getbool("free",&fsrf)) fsrf=false; /* free surface flag */
Fwav = sf_input ("in" ); /* wavelet */
Fsou = sf_input ("sou"); /* sources */
Frec = sf_input ("rec"); /* receivers */
Fvel = sf_input ("vel"); /* velocity */
Fden = sf_input ("den"); /* density */
Fref = sf_input ("ref"); /* reflectivity */
Fwfl = sf_output("wfl"); /* wavefield */
Fdat = sf_output("out"); /* data */
Fliw = sf_output("liw"); /* wavefield (scattered) */
Flid = sf_output("lid"); /* data (scattered) */
/* axes */
at = sf_iaxa(Fwav,2);
sf_setlabel(at,"t");
if(verb) sf_raxa(at); /* time */
as = sf_iaxa(Fsou,2);
sf_setlabel(as,"s");
if(verb) sf_raxa(as); /* sources */
ar = sf_iaxa(Frec,2);
sf_setlabel(ar,"r");
if(verb) sf_raxa(ar); /* receivers */
a1 = sf_iaxa(Fvel,1);
sf_setlabel(a1,"z");
if(verb) sf_raxa(a1); /* z */
a2 = sf_iaxa(Fvel,2);
sf_setlabel(a2,"x");
if(verb) sf_raxa(a2); /* x */
a3 = sf_iaxa(Fvel,3);
sf_setlabel(a3,"y");
if(verb) sf_raxa(a3); /* y */
nt = sf_n(at);
dt = sf_d(at);
ns = sf_n(as);
nr = sf_n(ar);
n1 = sf_n(a1);
d1 = sf_d(a1);
n2 = sf_n(a2);
d2 = sf_d(a2);
n3 = sf_n(a3);
d3 = sf_d(a3);
if(! sf_getint("jdata",&jdata)) jdata=1;
if(snap) { /* save wavefield every *jsnap* time steps */
if(! sf_getint("jsnap",&jsnap)) jsnap=nt;
}
/*------------------------------------------------------------*/
/* expand domain for FD operators and ABC */
if( !sf_getint("nb",&nb) || nb<NOP) nb=NOP;
fdm=fdutil3d_init(verb,fsrf,a1,a2,a3,nb,ompchunk);
sf_setn(a1,fdm->nzpad);
sf_seto(a1,fdm->ozpad);
if(verb) sf_raxa(a1);
sf_setn(a2,fdm->nxpad);
sf_seto(a2,fdm->oxpad);
if(verb) sf_raxa(a2);
sf_setn(a3,fdm->nypad);
sf_seto(a3,fdm->oypad);
if(verb) sf_raxa(a3);
/*------------------------------------------------------------*/
/* setup output data header */
sf_oaxa(Fdat,ar,1);
sf_oaxa(Flid,ar,1);
sf_setn(at,nt/jdata);
sf_setd(at,dt*jdata);
sf_oaxa(Fdat,at,2);
sf_oaxa(Flid,at,2);
/* setup output wavefield header */
if(snap) {
if(!sf_getint ("nqz",&nqz)) nqz=sf_n(a1);
if(!sf_getint ("nqx",&nqx)) nqx=sf_n(a2);
if(!sf_getint ("nqy",&nqy)) nqy=sf_n(a3);
if(!sf_getfloat("oqz",&oqz)) oqz=sf_o(a1);
if(!sf_getfloat("oqx",&oqx)) oqx=sf_o(a2);
if(!sf_getfloat("oqy",&oqy)) oqy=sf_o(a3);
dqz=sf_d(a1);
dqx=sf_d(a2);
dqy=sf_d(a3);
ac1 = sf_maxa(nqz,oqz,dqz);
ac2 = sf_maxa(nqx,oqx,dqx);
ac3 = sf_maxa(nqy,oqy,dqy);
/* check if the imaging window fits in the wavefield domain */
uc=sf_floatalloc3(sf_n(ac1),sf_n(ac2),sf_n(ac3));
ntsnap=0;
for(it=0; it<nt; it++) {
if(it%jsnap==0) ntsnap++;
}
sf_setn(at, ntsnap);
sf_setd(at,dt*jsnap);
if(verb) sf_raxa(at);
/* sf_setn(at,nt/jsnap);
sf_setd(at,dt*jsnap); */
sf_oaxa(Fwfl,ac1,1);
sf_oaxa(Fwfl,ac2,2);
sf_oaxa(Fwfl,ac3,3);
sf_oaxa(Fwfl,at, 4);
sf_oaxa(Fliw,ac1,1);
sf_oaxa(Fliw,ac2,2);
sf_oaxa(Fliw,ac3,3);
sf_oaxa(Fliw,at, 4);
}
/* source wavelet array allocation */
ww = sf_floatalloc(ns);
/* data array allocation*/
bdd = sf_floatalloc(nr);
sdd = sf_floatalloc(nr);
/*------------------------------------------------------------*/
/* setup source/receiver coordinates */
ss = (pt3d*) sf_alloc(ns,sizeof(*ss));
rr = (pt3d*) sf_alloc(nr,sizeof(*rr));
pt3dread1(Fsou,ss,ns,3); /* read (x,y,z) coordinates */
pt3dread1(Frec,rr,nr,3); /* read (x,y,z) coordinates */
cs = lint3d_make(ns,ss,fdm);
cr = lint3d_make(nr,rr,fdm);
fdbell3d_init(1);
/*------------------------------------------------------------*/
/* setup FD coefficients */
dt2 = dt*dt;
id1 = 1/d1;
id2 = 1/d2;
id3 = 1/d3;
c1x = C1*id2;
c1y = C1*id3;
c1z = C1*id1;
c2x = C2*id2;
c2y = C2*id3;
c2z = C2*id1;
c3x = C3*id2;
c3y = C3*id3;
c3z = C3*id1;
/*------------------------------------------------------------*/
/* input density */
roin = sf_floatalloc3(n1, n2, n3);
ro = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
iro = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
sf_floatread(roin[0][0],n1*n2*n3,Fden);
expand3d(roin,ro,fdm);
/* inverse density to avoid computation on the fly */
/*
there is 1 shell for i1=0 || i2=0 || i3=0 that is zero,
no big deal but better to fix it
*/
for (i3=1; i3<fdm->nypad; i3++) {
for (i2=1; i2<fdm->nxpad; i2++) {
for (i1=1; i1<fdm->nzpad; i1++) {
iro[i3][i2][i1] = 6./( 3*ro[i3 ][i2 ][i1 ] +
ro[i3 ][i2 ][i1-1] +
ro[i3 ][i2-1][i1 ] +
ro[i3-1][i2 ][i1 ] );
}
}
}
free(**roin);
free(*roin);
free(roin);
/*------------------------------------------------------------*/
/* input velocity */
vpin = sf_floatalloc3(n1, n2, n3);
vp = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
vt = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
sf_floatread(vpin[0][0],n1*n2*n3,Fvel);
expand3d(vpin,vp,fdm);
free(**vpin);
free(*vpin);
free(vpin);
/*------------------------------------------------------------*/
/* input reflectivity */
rfin = sf_floatalloc3(n1, n2, n3);
rf = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
sf_floatread(rfin[0][0],n1*n2*n3,Fref);
expand3d(rfin,rf,fdm);
free(**rfin);
free(*rfin);
free(rfin);
for (i3=0; i3<fdm->nypad; i3++) {
for (i2=0; i2<fdm->nxpad; i2++) {
for (i1=0; i1<fdm->nzpad; i1++) {
vt[i3][i2][i1] = vp[i3][i2][i1] * vp[i3][i2][i1] * dt2;
}
}
}
/* free surface */
if(fsrf) {
fsrfbck = sf_floatalloc3(4*NOP, fdm->nxpad, fdm->nypad);
fsrfsct = sf_floatalloc3(4*NOP, fdm->nxpad, fdm->nypad);
}
/*------------------------------------------------------------*/
/* allocate wavefield arrays */
bum = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
buo = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
bup = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
bua = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
buat = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
sum = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
suo = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
sup = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
sua = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
suat = sf_floatalloc3(fdm->nzpad, fdm->nxpad, fdm->nypad);
for (i3=0; i3<fdm->nypad; i3++) {
for (i2=0; i2<fdm->nxpad; i2++) {
for (i1=0; i1<fdm->nzpad; i1++) {
bum[i3][i2][i1]=0;
buo[i3][i2][i1]=0;
bup[i3][i2][i1]=0;
bua[i3][i2][i1]=0;
sum[i3][i2][i1]=0;
suo[i3][i2][i1]=0;
sup[i3][i2][i1]=0;
sua[i3][i2][i1]=0;
}
}
}
/*------------------------------------------------------------*/
/* one-way abc setup */
abc = abcone3d_make(NOP,dt,vp,fsrf,fdm);
/* sponge abc setup */
spo = sponge_make(fdm->nb);
free(**vp);
free(*vp);
free(vp);
/*--------------------------------------------------------------*/
/* */
/* MAIN LOOP */
/* */
/*--------------------------------------------------------------*/
if(verb) fprintf(stderr,"\nFORWARD BORN ACOUSTIC VARIABLE-DENSITY WAVE EXTRAPOLATION \n");
/* extrapolation */
start_t = clock();
for (it=0; it<nt; it++) {
if(verb) fprintf(stderr,"%d/%d \r",it,nt);
#ifdef _OPENMP
#pragma omp parallel private(i3,i2,i1)
#endif
{
if (fsrf) {
/* free surface */
#ifdef _OPENMP
#pragma omp for schedule(dynamic,fdm->ompchunk)
#endif
for (i3=0; i3<fdm->nypad; i3++) {
for (i2=0; i2<fdm->nxpad; i2++) {
for (i1=nb; i1<nb+2*NOP; i1++) {
fsrfbck[i3][i2][2*NOP+(i1-nb) ] = buo[i3][i2][i1];
fsrfbck[i3][i2][2*NOP-(i1-nb)-1] = -buo[i3][i2][i1];
fsrfsct[i3][i2][2*NOP+(i1-nb) ] = suo[i3][i2][i1];
fsrfsct[i3][i2][2*NOP-(i1-nb)-1] = -suo[i3][i2][i1];
}
}
}
}
// spatial derivatives z
#ifdef _OPENMP
#pragma omp for schedule(dynamic,fdm->ompchunk)
#endif
for (i3=NOP; i3<fdm->nypad-NOP; i3++) {
for (i2=NOP; i2<fdm->nxpad-NOP; i2++) {
for (i1=NOP; i1<fdm->nzpad-NOP; i1++) {
// gather
buat[i3][i2][i1] = iro[i3][i2][i1]*(
c3z*(buo[i3][i2][i1+2] - buo[i3][i2][i1-3]) +
c2z*(buo[i3][i2][i1+1] - buo[i3][i2][i1-2]) +
c1z*(buo[i3][i2][i1 ] - buo[i3][i2][i1-1])
);
suat[i3][i2][i1] = iro[i3][i2][i1]*(
c3z*(suo[i3][i2][i1+2] - suo[i3][i2][i1-3]) +
c2z*(suo[i3][i2][i1+1] - suo[i3][i2][i1-2]) +
c1z*(suo[i3][i2][i1 ] - suo[i3][i2][i1-1])
);
}
}
}
if (fsrf) {
// free surface
#ifdef _OPENMP
#pragma omp for schedule(dynamic,fdm->ompchunk)
#endif
for (i3=NOP; i3<fdm->nypad-NOP; i3++) {
for (i2=NOP; i2<fdm->nxpad-NOP; i2++) {
for (i1=nb-NOP; i1<nb+NOP; i1++) {
buat[i3][i2][i1] = iro[i3][i2][i1]*(
c3z*(fsrfbck[i3][i2][2*NOP+(i1-nb)+2] - fsrfbck[i3][i2][2*NOP+(i1-nb)-3]) +
c2z*(fsrfbck[i3][i2][2*NOP+(i1-nb)+1] - fsrfbck[i3][i2][2*NOP+(i1-nb)-2]) +
c1z*(fsrfbck[i3][i2][2*NOP+(i1-nb) ] - fsrfbck[i3][i2][2*NOP+(i1-nb)-1])
);
suat[i3][i2][i1] = iro[i3][i2][i1]*(
c3z*(fsrfsct[i3][i2][2*NOP+(i1-nb)+2] - fsrfsct[i3][i2][2*NOP+(i1-nb)-3]) +
c2z*(fsrfsct[i3][i2][2*NOP+(i1-nb)+1] - fsrfsct[i3][i2][2*NOP+(i1-nb)-2]) +
c1z*(fsrfsct[i3][i2][2*NOP+(i1-nb) ] - fsrfsct[i3][i2][2*NOP+(i1-nb)-1])
);
}
}
}
}
#ifdef _OPENMP
#pragma omp for schedule(dynamic,fdm->ompchunk)
#endif
for (i3=NOP; i3<fdm->nypad-NOP; i3++) {
for (i2=NOP; i2<fdm->nxpad-NOP; i2++) {
for (i1=NOP; i1<fdm->nzpad-NOP; i1++) {
// scatter
bua[i3][i2][i1] = c1z*( buat[i3][i2][i1 ] -
buat[i3][i2][i1+1]) +
c2z*( buat[i3][i2][i1-1] -
buat[i3][i2][i1+2]) +
c3z*( buat[i3][i2][i1-2] -
buat[i3][i2][i1+3]);
sua[i3][i2][i1] = c1z*( suat[i3][i2][i1 ] -
suat[i3][i2][i1+1]) +
c2z*( suat[i3][i2][i1-1] -
suat[i3][i2][i1+2]) +
c3z*( suat[i3][i2][i1-2] -
suat[i3][i2][i1+3]);
}
}
}
// spatial derivatives x
#ifdef _OPENMP
#pragma omp for schedule(dynamic,fdm->ompchunk)
#endif
for (i3=NOP; i3<fdm->nypad-NOP; i3++) {
for (i2=NOP; i2<fdm->nxpad-NOP; i2++) {
for (i1=NOP; i1<fdm->nzpad-NOP; i1++) {
// gather
buat[i3][i2][i1] = iro[i3][i2][i1]*(
c3x*(buo[i3][i2+2][i1] - buo[i3][i2-3][i1]) +
c2x*(buo[i3][i2+1][i1] - buo[i3][i2-2][i1]) +
c1x*(buo[i3][i2 ][i1] - buo[i3][i2-1][i1])
);
suat[i3][i2][i1] = iro[i3][i2][i1]*(
c3x*(suo[i3][i2+2][i1] - suo[i3][i2-3][i1]) +
c2x*(suo[i3][i2+1][i1] - suo[i3][i2-2][i1]) +
c1x*(suo[i3][i2 ][i1] - suo[i3][i2-1][i1])
);
}
}
}
#ifdef _OPENMP
#pragma omp for schedule(dynamic,fdm->ompchunk)
#endif
for (i3=NOP; i3<fdm->nypad-NOP; i3++) {
for (i2=NOP; i2<fdm->nxpad-NOP; i2++) {
for (i1=NOP; i1<fdm->nzpad-NOP; i1++) {
// scatter
bua[i3][i2 ][i1] += c1x*(buat[i3][i2 ][i1] -
buat[i3][i2+1][i1]) +
c2x*(buat[i3][i2-1][i1] -
buat[i3][i2+2][i1]) +
c3x*(buat[i3][i2-2][i1] -
buat[i3][i2+3][i1]);
sua[i3][i2 ][i1] += c1x*(suat[i3][i2 ][i1] -
suat[i3][i2+1][i1]) +
c2x*(suat[i3][i2-1][i1] -
suat[i3][i2+2][i1]) +
c3x*(suat[i3][i2-2][i1] -
suat[i3][i2+3][i1]);
}
}
}
// spatial derivatives y
#ifdef _OPENMP
#pragma omp for schedule(dynamic,fdm->ompchunk)
#endif
for (i3=NOP; i3<fdm->nypad-NOP; i3++) {
for (i2=NOP; i2<fdm->nxpad-NOP; i2++) {
for (i1=NOP; i1<fdm->nzpad-NOP; i1++) {
// gather
buat[i3][i2][i1] = iro[i3][i2][i1]*(
c3x*(buo[i3+2][i2][i1] - buo[i3-3][i2][i1]) +
c2x*(buo[i3+1][i2][i1] - buo[i3-2][i2][i1]) +
c1x*(buo[i3 ][i2][i1] - buo[i3-1][i2][i1])
);
suat[i3][i2][i1] = iro[i3][i2][i1]*(
c3x*(suo[i3+2][i2][i1] - suo[i3-3][i2][i1]) +
c2x*(suo[i3+1][i2][i1] - suo[i3-2][i2][i1]) +
c1x*(suo[i3 ][i2][i1] - suo[i3-1][i2][i1])
);
}
}
}
#ifdef _OPENMP
#pragma omp for schedule(dynamic,fdm->ompchunk)
#endif
for (i3=NOP; i3<fdm->nypad-NOP; i3++) {
for (i2=NOP; i2<fdm->nxpad-NOP; i2++) {
for (i1=NOP; i1<fdm->nzpad-NOP; i1++) {
// scatter
bua[i3][i2][i1] += c1y*(buat[i3 ][i2][i1] -
buat[i3+1][i2][i1]) +
c2y*( buat[i3-1][i2][i1] -
buat[i3+2][i2][i1]) +
c3y*( buat[i3-2][i2][i1] -
buat[i3+3][i2][i1]);
sua[i3][i2][i1] += c1y*(suat[i3 ][i2][i1] -
suat[i3+1][i2][i1]) +
c2y*( suat[i3-1][i2][i1] -
suat[i3+2][i2][i1]) +
c3y*( suat[i3-2][i2][i1] -
suat[i3+3][i2][i1]);
}
}
}
/* step forward in time */
#ifdef _OPENMP
#pragma omp for schedule(dynamic,fdm->ompchunk)
#endif
for (i3=NOP; i3<fdm->nypad-NOP; i3++) {
for (i2=NOP; i2<fdm->nxpad-NOP; i2++) {
for (i1=NOP; i1<fdm->nzpad-NOP; i1++) {
bup[i3][i2][i1] = 2*buo[i3][i2][i1]
- bum[i3][i2][i1]
- ro[i3][i2][i1]*vt[i3][i2][i1]*bua[i3][i2][i1];
sup[i3][i2][i1] = 2*suo[i3][i2][i1]
- sum[i3][i2][i1]
- ro[i3][i2][i1]*vt[i3][i2][i1]*sua[i3][i2][i1];
}
}
}
/* single scattering */
#ifdef _OPENMP
#pragma omp for schedule(dynamic,ompchunk)
#endif
for (i3=NOP; i3<fdm->nypad-NOP; i3++) {
for (i2=NOP; i2<fdm->nxpad-NOP; i2++) {
for (i1=NOP; i1<fdm->nzpad-NOP; i1++) {
sup[i3][i2][i1] -= 2*rf[i3][i2][i1]*ro[i3][i2][i1]*bua[i3][i2][i1]*dt2;
}
}
}
} /* end of the parallel section */
/* inject acceleration source */
sf_floatread(ww,ns,Fwav);
lint3d_bell(bup,ww,cs);
/* extract data */
lint3d_extract(bup,bdd,cr);
lint3d_extract(sup,sdd,cr);
if(snap && it%jsnap==0) {
cut3d(bup,uc,fdm,ac1,ac2,ac3);
sf_floatwrite(uc[0][0],sf_n(ac1)*sf_n(ac2)*sf_n(ac3),Fwfl);
cut3d(sup,uc,fdm,ac1,ac2,ac3);
sf_floatwrite(uc[0][0],sf_n(ac1)*sf_n(ac2)*sf_n(ac3),Fliw);
}
if( it%jdata==0) {
sf_floatwrite(bdd,nr,Fdat);
sf_floatwrite(sdd,nr,Flid);
}
/* one-way abc apply*/
if (dabc) {
abcone3d_apply(bup,buo,NOP,abc,fdm);
sponge3d_apply(bup, spo,fdm);
sponge3d_apply(buo, spo,fdm);
abcone3d_apply(sup,suo,NOP,abc,fdm);
sponge3d_apply(sup, spo,fdm);
sponge3d_apply(suo, spo,fdm);
}
/* circulate wavefield arrays */
but=bum;
bum=buo;
buo=bup;
bup=but;
sut=sum;
sum=suo;
suo=sup;
sup=sut;
} /* end time loop */
end_t = clock();
if(verb) fprintf(stderr,"\n");
if (verb) {
total_t = (float)(end_t - start_t) / CLOCKS_PER_SEC;
fprintf(stderr,"Total time taken by CPU: %g\n", total_t );
fprintf(stderr,"Exiting of the program...\n");
}
/*------------------------------------------------------------*/
/* deallocate arrays */
free(**bum);
free(*bum);
free(bum);
free(**buo);
free(*buo);
free(buo);
free(**bup);
free(*bup);
free(bup);
free(**bua);
free(*bua);
free(bua);
free(**buat);
free(*buat);
free(buat);
free(**sum);
free(*sum);
free(sum);
free(**suo);
free(*suo);
free(suo);
free(**sup);
free(*sup);
free(sup);
free(**sua);
free(*sua);
free(sua);
free(**suat);
free(*suat);
free(suat);
if(snap) {
free(**uc);
free(*uc);
free(uc);
}
if (fsrf) {
free(**fsrfbck);
free(*fsrfbck);
free(fsrfbck);
free(**fsrfsct);
free(*fsrfsct);
free(fsrfsct);
}
free(**vt);
free(*vt);
free(vt);
free(**ro);
free(*ro);
free(ro);
free(**iro);
free(*iro);
free(iro);
free(**rf);
free(*rf);
free(rf);
free(ww);
free(ss);
free(rr);
free(bdd);
free(sdd);
if (dabc) {
free(spo);
free(abc);
}
free(fdm);
/* ------------------------------------------------------------------------------------------ */
/* CLOSE FILES AND EXIT */
if (Fwav!=NULL) sf_fileclose(Fwav);
if (Fsou!=NULL) sf_fileclose(Fsou);
if (Frec!=NULL) sf_fileclose(Frec);
if (Fvel!=NULL) sf_fileclose(Fvel);
if (Fden!=NULL) sf_fileclose(Fden);
if (Fref!=NULL) sf_fileclose(Fref);
if (Fdat!=NULL) sf_fileclose(Fdat);
if (Fwfl!=NULL) sf_fileclose(Fwfl);
if (Fliw!=NULL) sf_fileclose(Fliw);
if (Flid!=NULL) sf_fileclose(Flid);
exit (0);
}
int main(int argc, char* argv[])
{
bool verb,fsrf,snap,expl,dabc,abcone,is2D,cfl;
bool ignore_interpolation = false; /* ignore interpolation for receivers - makes code faster, but only works when receivers are on grid points */
int jsnap,ntsnap,jdata;
float fmax, safety;
enum SourceType srctype;
/* I/O files */
sf_file Fwav=NULL; /* wavelet */
sf_file Fsou=NULL; /* sources */
sf_file Frec=NULL; /* receivers */
sf_file Fvel=NULL; /* velocity */
sf_file Fden=NULL; /* density */
sf_file Fdat=NULL; /* data */
sf_file Fwfl=NULL; /* wavefield */
/*set all y variables to be either zero or null to avoid compiler warnings
about being uninitialized */
/* cube axes */
sf_axis at,az,ax,ay=NULL;
sf_axis as,ar;
int nt,nz,nx,ny=0,ns,nr,nb;
int it,iz,ix,iy=0;
float dt,dz,dx,dy=0,idz,idx,idy=0;
/* I/O arrays */
float *ww=NULL; /* wavelet */
float *dd=NULL; /* data */
/* FD operator size */
float co,cax,cbx,cay,cby,caz,cbz;
/* wavefield cut params */
sf_axis acz=NULL,acx=NULL,acy=NULL;
int nqz,nqx,nqy;
float oqz,oqx,oqy;
float dqz,dqx,dqy;
/*------------------------------------------------------------*/
/* init RSF */
sf_init(argc,argv);
/*------------------------------------------------------------*/
/* OMP parameters */
if( !sf_getbool("ignint",&ignore_interpolation)) ignore_interpolation = false;
if(! sf_getbool("verb",&verb)) verb=false; /* verbosity flag */
if(! sf_getbool("snap",&snap)) snap=false; /* wavefield snapshots flag */
if(! sf_getbool("free",&fsrf)) fsrf=false; /* free surface flag */
if(! sf_getbool("expl",&expl)) expl=false; /* "exploding reflector" */
if(! sf_getbool("dabc",&dabc)) dabc=false; /* absorbing BC */
if(! sf_getbool("cfl",&cfl)) cfl=false; /* Use CFL check */
if(! sf_getbool("abcone",&abcone)) abcone=false; /* Use Zero-incident boundary condition*/
int ttype = 0;
if(! sf_getint("srctype",&ttype)) ttype = 0; /* source type, see comments */
if(ttype < 0 || ttype > 1) sf_error("Invalid source type specified");
srctype = ttype;
if (cfl) {
if(! sf_getfloat("fmax",&fmax)) { /* max frequency for cfl check */
sf_error("CFL: Must specify fmax for CFL check");
}
if(! sf_getfloat("safety",&safety) || safety < 0.0) safety= 0.8; /*safety factor for cfl check*/
}
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* I/O files */
Fwav = sf_input ("in" ); /* wavelet */
Fvel = sf_input ("vel"); /* velocity */
Fsou = sf_input ("sou"); /* sources */
Frec = sf_input ("rec"); /* receivers */
Fwfl = sf_output("wfl"); /* wavefield */
Fdat = sf_output("out"); /* data */
Fden = sf_input ("den"); /* density */
/* Determine dimensionality, if 2D then axis 3 has n size of 1 */
sf_axis test = sf_iaxa(Fvel,3);
if(sf_n(test) == 1) is2D = true;
else is2D = false;
/*------------------------------------------------------------*/
/* axes */
at = sf_iaxa(Fwav,2); sf_setlabel(at,"t"); if(verb) sf_raxa(at); /* time */
az = sf_iaxa(Fvel,1); sf_setlabel(az,"z"); if(verb) sf_raxa(az); /* depth */
ax = sf_iaxa(Fvel,2); sf_setlabel(ax,"x"); if(verb) sf_raxa(ax); /* space */
as = sf_iaxa(Fsou,2); sf_setlabel(as,"s"); if(verb) sf_raxa(as); /* sources */
ar = sf_iaxa(Frec,2); sf_setlabel(ar,"r"); if(verb) sf_raxa(ar); /* receivers */
nt = sf_n(at); dt = sf_d(at);
nz = sf_n(az); dz = sf_d(az);
nx = sf_n(ax); dx = sf_d(ax);
ns = sf_n(as);
nr = sf_n(ar);
if(!is2D){ /*If 3D*/
ay=sf_iaxa(Fvel,3); sf_setlabel(ay,"y"); if(verb) sf_raxa(ay); /*space*/
ny=sf_n(ay); dy=sf_d(ay);
}
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* other execution parameters */
if(! sf_getint("jdata",&jdata)) jdata=1;
if(snap) { /* save wavefield every *jsnap* time steps */
if(! sf_getint("jsnap",&jsnap)) jsnap=nt;
}
/*------------------------------------------------------------*/
if(is2D){
/* Begin 2d code */
/* FDM structure */
fdm2d fdm=NULL;
abcone2d abc=NULL;
sponge spo=NULL;
pt2d *ss=NULL; /* sources */
pt2d *rr=NULL; /* receivers */
float **tt=NULL;
float **ro=NULL; /* density */
float **roz=NULL; /* normalized 1st derivative of density on axis 1 */
float **rox=NULL; /* normalized 1st derivative of density on axis 2 */
float **vp=NULL; /* velocity */
float **vt=NULL; /* temporary vp*vp * dt*dt */
float **um,**uo,**up,**ua,**ut; /* wavefield: um = U @ t-1; uo = U @ t; up = U @ t+1 */
/* linear interpolation weights/indices */
lint2d cs,cr;
float **uc=NULL;
/*------------------------------------------------------------*/
/* expand domain for FD operators and ABC */
if( !sf_getint("nb",&nb) || nb<NOP) nb=NOP;
fdm=fdutil_init(verb,fsrf,az,ax,nb,1);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* setup output data header */
sf_oaxa(Fdat,ar,1);
sf_setn(at,nt/jdata);
sf_setd(at,dt*jdata);
sf_oaxa(Fdat,at,2);
/* setup output wavefield header */
if(snap) {
if(!sf_getint ("nqz",&nqz)) nqz=sf_n(az);
if(!sf_getint ("nqx",&nqx)) nqx=sf_n(ax);
if(!sf_getfloat("oqz",&oqz)) oqz=sf_o(az);
if(!sf_getfloat("oqx",&oqx)) oqx=sf_o(ax);
sf_setn(az,fdm->nzpad); sf_seto(az,fdm->ozpad); if(verb) sf_raxa(az);
sf_setn(ax,fdm->nxpad); sf_seto(ax,fdm->oxpad); if(verb) sf_raxa(ax);
dqz=sf_d(az);
dqx=sf_d(ax);
acz = sf_maxa(nqz,oqz,dqz); sf_raxa(acz);
acx = sf_maxa(nqx,oqx,dqx); sf_raxa(acx);
/* check if the imaging window fits in the wavefield domain */
uc=sf_floatalloc2(sf_n(acz),sf_n(acx));
ntsnap=0;
for(it=0; it<nt; it++) {
if(it%jsnap==0) ntsnap++;
}
sf_setn(at, ntsnap);
sf_setd(at,dt*jsnap);
if(verb) sf_raxa(at);
sf_oaxa(Fwfl,acz,1);
sf_oaxa(Fwfl,acx,2);
sf_oaxa(Fwfl,at, 3);
}
if(expl) {
ww = sf_floatalloc( 1);
} else {
ww = sf_floatalloc(ns);
}
dd = sf_floatalloc(nr);
/*------------------------------------------------------------*/
/* setup source/receiver coordinates */
ss = (pt2d*) sf_alloc(ns,sizeof(*ss));
rr = (pt2d*) sf_alloc(nr,sizeof(*rr));
pt2dread1(Fsou,ss,ns,2); /* read (x,z) coordinates */
pt2dread1(Frec,rr,nr,2); /* read (x,z) coordinates */
cs = lint2d_make(ns,ss,fdm);
cr = lint2d_make(nr,rr,fdm);
/*------------------------------------------------------------*/
/* setup FD coefficients */
idz = 1/dz;
idx = 1/dx;
co = C0 * (idx*idx+idz*idz);
cax= CA * idx*idx;
cbx= CB * idx*idx;
caz= CA * idz*idz;
cbz= CB * idz*idz;
/*------------------------------------------------------------*/
tt = sf_floatalloc2(nz,nx);
ro =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
roz =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
rox =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
vp =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
vt =sf_floatalloc2(fdm->nzpad,fdm->nxpad);
/* input density */
sf_floatread(tt[0],nz*nx,Fden); expand(tt,ro ,fdm);
/* normalized density derivatives */
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
roz[ix][iz] = DZ(ro,ix,iz,idz) / ro[ix][iz];
rox[ix][iz] = DX(ro,ix,iz,idx) / ro[ix][iz];
}
}
free(*ro); free(ro);
/* input velocity */
sf_floatread(tt[0],nz*nx,Fvel ); expand(tt,vp,fdm);
float vpmax = 0.0; float vpmin = 10000000000000000;
/* precompute vp^2 * dt^2 */
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
vt[ix][iz] = vp[ix][iz] * vp[ix][iz] * dt*dt;
if (vp[ix][iz] < vpmin) vpmin = vp[ix][iz];
else if (vp[ix][iz] > vpmax) vpmax = vp[ix][iz];
}
}
if (cfl) cfl_acoustic(vpmin,vpmax,dx,-1.0f,dz,dt,fmax,safety,NUM_INTERVALS);
if(fsrf) { /* free surface */
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nb; iz++) {
vt[ix][iz]=0;
}
}
}
free(*tt); free(tt);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* allocate wavefield arrays */
um=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
uo=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
up=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
ua=sf_floatalloc2(fdm->nzpad,fdm->nxpad);
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
um[ix][iz]=0;
uo[ix][iz]=0;
up[ix][iz]=0;
ua[ix][iz]=0;
}
}
/*------------------------------------------------------------*/
if (abcone) abc = abcone2d_make(NOP,dt,vp,fsrf,fdm);
if(dabc) {
/* one-way abc setup */
/* sponge abc setup */
spo = sponge_make(fdm->nb);
}
/*------------------------------------------------------------*/
/*
* MAIN LOOP
*/
/*------------------------------------------------------------*/
if(verb) fprintf(stderr,"\n");
for (it=0; it<nt; it++) {
if(verb) fprintf(stderr,"%d/%d \r",it,nt);
#pragma omp parallel for \
schedule(dynamic) \
private(ix,iz) \
shared(fdm,ua,uo,co,caz,cbz)
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
/* 4th order Laplacian operator */
ua[ix][iz] =
co * uo[ix ][iz ] +
caz*(uo[ix ][iz-1] + uo[ix ][iz+1]) +
cbz*(uo[ix ][iz-2] + uo[ix ][iz+2]) ;
/* density term */
/*ua[ix][iz] -= (
DZ(uo,ix,iz,idz) * roz[ix][iz] +
DX(uo,ix,iz,idx) * rox[ix][iz] );
*/
}
}
#pragma omp parallel for \
schedule(dynamic) \
private(ix,iz) \
shared(fdm,ua,uo,co,cax,cbx)
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
ua[ix][iz] = ua[ix][iz] +
cax*(uo[ix-1][iz ] + uo[ix+1][iz ]) +
cbx*(uo[ix-2][iz ] + uo[ix+2][iz ]);
}
}
/* inject acceleration source */
if (srctype == ACCELERATION){
if(expl) {
sf_floatread(ww, 1,Fwav);
lint2d_inject1(ua,ww[0],cs);
} else {
sf_floatread(ww,ns,Fwav);
lint2d_inject(ua,ww,cs);
}
}
/* step forward in time */
#pragma omp parallel for \
schedule(dynamic) \
private(ix,iz) \
shared(fdm,ua,uo,um,up,vt)
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
up[ix][iz] = 2*uo[ix][iz]
- um[ix][iz]
+ ua[ix][iz] * vt[ix][iz];
}
}
if(srctype == DISPLACEMENT){
if(expl) {
sf_floatread(ww, 1,Fwav);
lint2d_inject1(up,ww[0],cs);
} else {
sf_floatread(ww,ns,Fwav);
lint2d_inject(up,ww,cs);
}
}
/* circulate wavefield arrays */
ut=um;
um=uo;
uo=up;
up=ut;
if (abcone) abcone2d_apply(uo,um,NOP,abc,fdm);
if(dabc) {
/* one-way abc apply */
sponge2d_apply(um,spo,fdm);
sponge2d_apply(uo,spo,fdm);
sponge2d_apply(up,spo,fdm);
}
/* extract data */
if(ignore_interpolation){
cut2d_extract(uo,dd,cr);
} else {
lint2d_extract(uo,dd,cr);
}
if(snap && it%jsnap==0) {
cut2d(uo,uc,fdm,acz,acx);
sf_floatwrite(uc[0],sf_n(acz)*sf_n(acx),Fwfl);
}
if( it%jdata==0)
sf_floatwrite(dd,nr,Fdat);
}
if(verb) fprintf(stderr,"\n");
/*------------------------------------------------------------*/
/* deallocate arrays */
free(*um); free(um);
free(*up); free(up);
free(*uo); free(uo);
free(*ua); free(ua);
if(snap) { free(*uc); free(uc); }
free(*rox); free(rox);
free(*roz); free(roz);
free(*vp); free(vp);
free(*vt); free(vt);
free(ww);
free(ss);
free(rr);
free(dd);
exit (0);
} else {
/* FDM structure */
fdm3d fdm=NULL;
abcone3d abc=NULL;
sponge spo=NULL;
/* I/O arrays */
pt3d *ss=NULL; /* sources */
pt3d *rr=NULL; /* receivers */
/* Non-universal arrays */
float***tt=NULL;
float***ro=NULL; /* density */
float***roz=NULL; /* normalized 1st derivative of density on axis 1 */
float***rox=NULL; /* normalized 1st derivative of density on axis 2 */
float***roy=NULL; /* normalized 1st derivative of density on axis 3 */
float***vp=NULL; /* velocity */
float***vt=NULL; /* temporary vp*vp * dt*dt */
float***um,***uo,***up,***ua,***ut; /* wavefield: um = U @ t-1; uo = U @ t; up = U @ t+1 */
/* linear interpolation weights/indices */
lint3d cs,cr;
/* Wavefield cut params that are not universal */
float ***uc=NULL;
/*------------------------------------------------------------*/
/* expand domain for FD operators and ABC */
if( !sf_getint("nb",&nb) || nb<NOP) nb=NOP;
fdm=fdutil3d_init(verb,fsrf,az,ax,ay,nb,1);
sf_setn(az,fdm->nzpad); sf_seto(az,fdm->ozpad); if(verb) sf_raxa(az);
sf_setn(ax,fdm->nxpad); sf_seto(ax,fdm->oxpad); if(verb) sf_raxa(ax);
sf_setn(ay,fdm->nypad); sf_seto(ay,fdm->oypad); if(verb) sf_raxa(ay);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* setup output data header */
sf_oaxa(Fdat,ar,1);
sf_setn(at,nt/jdata);
sf_setd(at,dt*jdata);
sf_oaxa(Fdat,at,2);
/* setup output wavefield header */
if(snap) {
if(!sf_getint ("nqz",&nqz)) nqz=sf_n(az);
if(!sf_getint ("nqx",&nqx)) nqx=sf_n(ax);
if(!sf_getint ("nqy",&nqy)) nqy=sf_n(ay);
if(!sf_getfloat("oqz",&oqz)) oqz=sf_o(az);
if(!sf_getfloat("oqx",&oqx)) oqx=sf_o(ax);
if(!sf_getfloat("oqy",&oqy)) oqy=sf_o(ay);
dqz=sf_d(az);
dqx=sf_d(ax);
dqy=sf_d(ay);
acz = sf_maxa(nqz,oqz,dqz); sf_raxa(acz);
acx = sf_maxa(nqx,oqx,dqx); sf_raxa(acx);
acy = sf_maxa(nqy,oqy,dqy); sf_raxa(acy);
/* check if the imaging window fits in the wavefield domain */
uc=sf_floatalloc3(sf_n(acz),sf_n(acx),sf_n(acy));
ntsnap=0;
for(it=0; it<nt; it++) {
if(it%jsnap==0) ntsnap++;
}
sf_setn(at, ntsnap);
sf_setd(at,dt*jsnap);
if(verb) sf_raxa(at);
sf_oaxa(Fwfl,acz,1);
sf_oaxa(Fwfl,acx,2);
sf_oaxa(Fwfl,acy,3);
sf_oaxa(Fwfl,at, 4);
}
if(expl) {
ww = sf_floatalloc( 1);
} else {
ww = sf_floatalloc(ns);
}
dd = sf_floatalloc(nr);
/*------------------------------------------------------------*/
/* setup source/receiver coordinates */
ss = (pt3d*) sf_alloc(ns,sizeof(*ss));
rr = (pt3d*) sf_alloc(nr,sizeof(*rr));
pt3dread1(Fsou,ss,ns,3); /* read (x,y,z) coordinates */
pt3dread1(Frec,rr,nr,3); /* read (x,y,z) coordinates */
cs = lint3d_make(ns,ss,fdm);
cr = lint3d_make(nr,rr,fdm);
/*------------------------------------------------------------*/
/* setup FD coefficients */
idz = 1/dz;
idx = 1/dx;
idy = 1/dy;
co = C0 * (idx*idx+idy*idy+idz*idz);
cax= CA * idx*idx;
cbx= CB * idx*idx;
cay= CA * idy*idy;
cby= CB * idy*idy;
caz= CA * idz*idz;
cbz= CB * idz*idz;
/*------------------------------------------------------------*/
tt = sf_floatalloc3(nz,nx,ny);
ro =sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
roz =sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
rox =sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
roy =sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
vp =sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
vt =sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
/* input density */
sf_floatread(tt[0][0],nz*nx*ny,Fden); expand3d(tt,ro ,fdm);
/* normalized density derivatives */
for (iy=NOP; iy<fdm->nypad-NOP; iy++) {
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
roz[iy][ix][iz] = DZ3(ro,ix,iy,iz,idz) / ro[iy][ix][iz];
rox[iy][ix][iz] = DX3(ro,ix,iy,iz,idx) / ro[iy][ix][iz];
roy[iy][ix][iz] = DY3(ro,ix,iy,iz,idy) / ro[iy][ix][iz];
}
}
}
free(**ro); free(*ro); free(ro);
/* input velocity */
sf_floatread(tt[0][0],nz*nx*ny,Fvel ); expand3d(tt,vp,fdm);
/* precompute vp^2 * dt^2 */
float vpmin = 1000000000000; float vpmax = 0.0;
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
float vpt = vp[iy][ix][iz];
vt[iy][ix][iz] = vp[iy][ix][iz] * vp[iy][ix][iz] * dt*dt;
if (vpt > vpmax) vpmax = vpt;
else if (vpt < vpmin) vpmin = vpt;
}
}
}
if (cfl) cfl_acoustic(vpmin,vpmax,dx,dy,dz,dt,fmax,safety,NUM_INTERVALS);
if(fsrf) { /* free surface */
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nb; iz++) {
vt[iy][ix][iz]=0;
}
}
}
}
free(**tt); free(*tt); free(tt);
/*------------------------------------------------------------*/
/*------------------------------------------------------------*/
/* allocate wavefield arrays */
um=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
uo=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
up=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
ua=sf_floatalloc3(fdm->nzpad,fdm->nxpad,fdm->nypad);
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
um[iy][ix][iz]=0;
uo[iy][ix][iz]=0;
up[iy][ix][iz]=0;
ua[iy][ix][iz]=0;
}
}
}
/*------------------------------------------------------------*/
if (abcone) abc = abcone3d_make(NOP,dt,vp,fsrf,fdm);
if(dabc) {
/* one-way abc setup */
/* sponge abc setup */
spo = sponge_make(fdm->nb);
}
/*------------------------------------------------------------*/
/*
* MAIN LOOP
*/
/*------------------------------------------------------------*/
if(verb) fprintf(stderr,"\n");
for (it=0; it<nt; it++) {
if(verb) fprintf(stderr,"%d/%d \r",it,nt);
#pragma omp parallel for \
schedule(dynamic) \
private(ix,iy,iz) \
shared(fdm,ua,uo,co,cax,cay,caz,cbx,cby,cbz,idx,idy,idz)
for (iy=NOP; iy<fdm->nypad-NOP; iy++) {
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
/* 4th order Laplacian operator */
ua[iy][ix][iz] =
co * uo[iy ][ix ][iz ] +
caz*(uo[iy ][ix ][iz-1] + uo[iy ][ix ][iz+1]) +
cbz*(uo[iy ][ix ][iz-2] + uo[iy ][ix ][iz+2]);
/* density term */
/*ua[iy][ix][iz] -= (
DZ3(uo,ix,iy,iz,idz) * roz[iy][ix][iz] +
DX3(uo,ix,iy,iz,idx) * rox[iy][ix][iz] +
DY3(uo,ix,iy,iz,idy) * roy[iy][ix][iz] );
*/
}
}
}
#pragma omp parallel for \
schedule(dynamic) \
private(ix,iy,iz) \
shared(fdm,ua,uo,co,cax,cay,caz,cbx,cby,cbz,idx,idy,idz)
for (iy=NOP; iy<fdm->nypad-NOP; iy++) {
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
ua[iy][ix][iz] = ua[iy][ix][iz] +
cax*(uo[iy ][ix-1][iz ] + uo[iy ][ix+1][iz ]) +
cbx*(uo[iy ][ix-2][iz ] + uo[iy ][ix+2][iz ]) ;
}
}
}
#pragma omp parallel for \
schedule(dynamic) \
private(ix,iy,iz) \
shared(fdm,ua,uo,co,cax,cay,caz,cbx,cby,cbz,idx,idy,idz)
for (iy=NOP; iy<fdm->nypad-NOP; iy++) {
for (ix=NOP; ix<fdm->nxpad-NOP; ix++) {
for(iz=NOP; iz<fdm->nzpad-NOP; iz++) {
ua[iy][ix][iz] = ua[iy][ix][iz] +
cay*(uo[iy-1][ix ][iz ] + uo[iy+1][ix ][iz ]) +
cby*(uo[iy-2][ix ][iz ] + uo[iy+2][ix ][iz ]);
}
}
}
/* inject acceleration source */
if (srctype == ACCELERATION){
if(expl) {
sf_floatread(ww, 1,Fwav);
lint3d_inject1(ua,ww[0],cs);
} else {
sf_floatread(ww,ns,Fwav);
lint3d_inject(ua,ww,cs);
}
}
/* step forward in time */
#pragma omp parallel for \
schedule(static) \
private(ix,iy,iz) \
shared(fdm,ua,uo,um,up,vt)
for (iy=0; iy<fdm->nypad; iy++) {
for (ix=0; ix<fdm->nxpad; ix++) {
for(iz=0; iz<fdm->nzpad; iz++) {
up[iy][ix][iz] = 2*uo[iy][ix][iz]
- um[iy][ix][iz]
+ ua[iy][ix][iz] * vt[iy][ix][iz];
}
}
}
if (srctype == DISPLACEMENT) {
if(expl) {
sf_floatread(ww, 1,Fwav);
lint3d_inject1(up,ww[0],cs);
} else {
sf_floatread(ww,ns,Fwav);
lint3d_inject(up,ww,cs);
}
}
/* circulate wavefield arrays */
ut=um;
um=uo;
uo=up;
up=ut;
if(abcone) abcone3d_apply(uo,um,NOP,abc,fdm);
if(dabc) {
/* one-way abc apply */
sponge3d_apply(um,spo,fdm);
sponge3d_apply(uo,spo,fdm);
sponge3d_apply(up,spo,fdm);
}
/* extract data */
if (ignore_interpolation) {
cut3d_extract(uo,dd,cr);
} else {
lint3d_extract(uo,dd,cr);
}
if(snap && it%jsnap==0) {
cut3d(uo,uc,fdm,acz,acx,acy);
sf_floatwrite(uc[0][0],sf_n(acz)*sf_n(acx)*sf_n(acy),Fwfl);
}
if(it%jdata==0)
sf_floatwrite(dd,nr,Fdat);
}
if(verb) fprintf(stderr,"\n");
/*------------------------------------------------------------*/
/* deallocate arrays */
free(**um); free(*um); free(um);
free(**up); free(*up); free(up);
free(**uo); free(*uo); free(uo);
free(**ua); free(*ua); free(ua);
if (snap) { free(**uc); free(*uc); free(uc); }
free(**rox); free(*rox); free(rox);
free(**roy); free(*roy); free(roy);
free(**roz); free(*roz); free(roz);
free(**vp); free(*vp); free(vp);
free(**vt); free(*vt); free(vt);
free(ss);
free(rr);
free(dd);
free(ww);
exit (0);
}
}