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, 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,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 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);
}
/*------------------------------------------------------------*/
wexcip3d wexcip_init(wexcub3d cub,
int nhx_,
int nhy_,
int nhz_,
int nht_,
int nhx2_,
int nhy2_,
int nhz2_,
int nht2_,
int nc_,
float dht_,
float oht_,
sf_file Fc,
int eic)
/*< initialize I.C. >*/
{
int iw, ic, ihx, ihy, ihz, iht, icx, icy, icz;
float ht, w;
/*------------------------------------------------------------*/
wexcip3d cip;
cip = (wexcip3d) sf_alloc(1,sizeof(*cip));
/*------------------------------------------------------------*/
/* allocate wavefields storage */
cip->ws = sf_complexalloc3(cub->amx.n,cub->amy.n,cub->az.n);
cip->wr = sf_complexalloc3(cub->amx.n,cub->amy.n,cub->az.n);
cip->ci = sf_complexalloc3(cub->amx.n,cub->amy.n,cub->az.n);
if(eic) {
cip->nhx = nhx_;
cip->nhy = nhy_;
cip->nhz = nhz_;
cip->nht = nht_;
cip->nhx2= nhx2_;
cip->nhy2= nhy2_;
cip->nhz2= nhz2_;
cip->nht2= nht2_;
cip->nc = nc_;
cip->oht = oht_;
cip->dht = dht_;
/*------------------------------------------------------------*/
/* precompute phase for time delay */
cip->tt = sf_complexalloc2(cip->nht2,cub->aw.n);
for(iw=0; iw<cub->aw.n; iw++) {
w = -( cub->aw.o+iw*cub->aw.d );
for(iht=0; iht<cip->nht2; iht++) {
ht = cip->oht + (iht+0)*cip->dht;
cip->tt[iw][iht] = sf_cmplx(cosf(2*w*ht),sinf(2*w*ht));
}
}
/*------------------------------------------------------------*/
/* allocate image storage */
cip->ei = sf_complexalloc5(cip->nhx2,cip->nhy2,cip->nhz2,cip->nht2,cip->nc);
cip->di = sf_complexalloc5(cip->nhx2,cip->nhy2,cip->nhz2,cip->nht2,cip->nc);
for(ic=0; ic<cip->nc; ic++) {
for(iht=0; iht<cip->nht2; iht++) {
for(ihz=0; ihz<cip->nhz2; ihz++) {
for(ihy=0; ihy<cip->nhy2; ihy++) {
for(ihx=0; ihx<cip->nhx2; ihx++) {
cip->ei[ic][iht][ihz][ihy][ihx] = sf_cmplx(0.0,0.0);
cip->di[ic][iht][ihz][ihy][ihx] = sf_cmplx(0.0,0.0);
}
}
}
}
}
/*------------------------------------------------------------*/
/* CIP coordinates */
cip->cc= (pt3d*) sf_alloc(cip->nc,sizeof(*cip->cc));
pt3dread1(Fc,cip->cc,cip->nc,3); /* read coordinates */
cip->mcxall=sf_intalloc2(cip->nhx2,cip->nc);
cip->pcxall=sf_intalloc2(cip->nhx2,cip->nc);
cip->mcyall=sf_intalloc2(cip->nhy2,cip->nc);
cip->pcyall=sf_intalloc2(cip->nhy2,cip->nc);
cip->mczall=sf_intalloc2(cip->nhz2,cip->nc);
cip->pczall=sf_intalloc2(cip->nhz2,cip->nc);
cip->ccin=sf_intalloc(cip->nc);
cip->cxmin = cub->amx.o + cip->nhx *cub->amx.d;
cip->cxmax = cub->amx.o + (cub->amx.n-1-cip->nhx)*cub->amx.d;
cip->cymin = cub->amy.o + cip->nhy *cub->amy.d;
cip->cymax = cub->amy.o + (cub->amy.n-1-cip->nhy)*cub->amy.d;
cip->czmin = cub->az.o + cip->nhz *cub->az.d;
cip->czmax = cub->az.o + (cub->az.n -1-cip->nhz)*cub->az.d;
for(ic=0; ic<cip->nc; ic++) {
cip->ccin[ic]=(cip->cc[ic].x>=cip->cxmin && cip->cc[ic].x<=cip->cxmax &&
cip->cc[ic].y>=cip->cymin && cip->cc[ic].y<=cip->cymax &&
cip->cc[ic].z>=cip->czmin && cip->cc[ic].z<=cip->czmax)?1:0;
if(cip->ccin[ic]) {
icx = 0.5+(cip->cc[ic].x-cub->amx.o)/cub->amx.d;
for(ihx=-cip->nhx; ihx<cip->nhx+1; ihx++) {
cip->mcxall[ic][cip->nhx+ihx] = icx-ihx;
cip->pcxall[ic][cip->nhx+ihx] = icx+ihx;
}
icy = 0.5+(cip->cc[ic].y-cub->amy.o)/cub->amy.d;
for(ihy=-cip->nhy; ihy<cip->nhy+1; ihy++) {
cip->mcyall[ic][cip->nhy+ihy] = icy-ihy;
cip->pcyall[ic][cip->nhy+ihy] = icy+ihy;
}
icz = 0.5+(cip->cc[ic].z-cub->az.o)/cub->az.d;
for(ihz=-cip->nhz; ihz<cip->nhz+1; ihz++) {
cip->mczall[ic][cip->nhz+ihz] = icz-ihz;
cip->pczall[ic][cip->nhz+ihz] = icz+ihz;
}
for(ihx=-cip->nhx; ihx<cip->nhx+1; ihx++) {
for(ihy=-cip->nhy; ihy<cip->nhy+1; ihy++) {
for(ihz=-cip->nhz; ihz<cip->nhz+1; ihz++) {
// sf_warning("ihx=%d,ihy=%d,ihz=%d,mcx=%d,pcx=%d,mcy=%d,pcy=%d,mcz=%d,pcz=%d",ihx,ihy,ihz,cip->mcxall[ic][cip->nhx+ihx],cip->pcxall[ic][cip->nhx+ihx],cip->mcyall[ic][cip->nhy+ihy],cip->pcyall[ic][cip->nhy+ihy],cip->mczall[ic][cip->nhz+ihz],cip->pczall[ic][cip->nhz+ihz]);
}
}
}
}
} /* loop over nc */
}
return cip;
}
/*------------------------------------------------------------*/
int main(int argc, char* argv[])
{
bool verb; /* verbosity flag */
bool pos; /* direction of spraying */
bool adj; /* adjoint operator flag */
bool wflcausal, oprcausal; /* causal wfl?, opr? */
sf_file Fopr, Fwfl, Fimg, Fcip; /* I/O files */
float ****opr=NULL,****wfl=NULL,*****img=NULL;
int itO,itW;
sf_axis az,ax,ay,at,ac,aa; /* wfld axes */
int nz,nx,ny,nt,nc;
int iz,ix,iy,it,ic;
sf_axis ahx, ahy, ahz, aht; /* EIC axes */
int nhx, nhy, nhz, nht;
int ihx, ihy, ihz, iht;
float dhx, dhy, dhz, dht;
pt3d *cc=NULL;
bool *ccin=NULL;
float cxmin,czmin,cymin;
float cxmax,czmax,cymax;
int icx, icz, icy;
int mcx, mcz, mcy, mct;
int pcx, pcz, pcy, pct;
int **mcxall, **pcxall;
int **mcyall, **pcyall;
int **mczall, **pczall;
int *mctall, *pctall;
int lht,fht; /* last buffer index */
float scale; /* time summation scaling */
int nslice; /* wavefield slice size */
bool gaus; /* gaussian taper */
float gsx,gsy,gsz,gst; /* std dev */
/*------------------------------------------------------------*/
sf_init(argc,argv);
#ifdef _OPENMP
omp_init();
#endif
if(! sf_getbool( "verb",&verb )) verb=false; /* verbosity flag */
if(! sf_getbool( "positive",&pos )) pos=true; /* if positive sprays opr to positive shits, else, sprays to negative shifts */
if(! sf_getbool( "adj",&adj )) adj=false; /* adjoint flag */
if(! sf_getbool("wflcausal",&wflcausal)) wflcausal=false; /* causal wfl? */
if(! sf_getbool("oprcausal",&oprcausal)) oprcausal=false; /* causal opr? */
/*------------------------------------------------------------*/
Fopr = sf_input ("opr" ); /* operator */
az=sf_iaxa(Fopr,1); if(verb) sf_raxa(az); nz = sf_n(az);
ax=sf_iaxa(Fopr,2); if(verb) sf_raxa(ax); nx = sf_n(ax);
ay=sf_iaxa(Fopr,3); if(verb) sf_raxa(ay); ny = sf_n(ay);
at=sf_iaxa(Fopr,4); if(verb) sf_raxa(at); nt = sf_n(at);
scale = 1./nt; /* time summation scaling */
nslice = nz*nx*ny*sizeof(float); /* wavefield slice */
Fcip = sf_input ("cip" ); /* CIP coordinates */
ac = sf_iaxa(Fcip,2);
sf_setlabel(ac,"c"); sf_setunit(ac,""); if(verb) sf_raxa(ac); nc = sf_n(ac);
/*------------------------------------------------------------*/
/* setup output */
if(adj) {
Fimg = sf_input ("in"); /* read img */
ahz=sf_iaxa(Fimg,1); nhz=(sf_n(ahz)-1)/2; if(verb) sf_raxa(ahz);
ahx=sf_iaxa(Fimg,2); nhx=(sf_n(ahx)-1)/2; if(verb) sf_raxa(ahx);
ahy=sf_iaxa(Fimg,3); nhy=(sf_n(ahy)-1)/2; if(verb) sf_raxa(ahy);
aht=sf_iaxa(Fimg,4); nht=(sf_n(aht)-1)/2; if(verb) sf_raxa(aht);
aa=sf_maxa(1,0,1); sf_setlabel(aa,""); sf_setunit(aa,"");
/* set output axes */
Fwfl = sf_output("out"); /* write wfl */
sf_oaxa(Fwfl,az,1);
sf_oaxa(Fwfl,ax,2);
sf_oaxa(Fwfl,ay,3);
sf_oaxa(Fwfl,at,4);
sf_oaxa(Fwfl,aa,5);
} else {
Fwfl = sf_input ( "in"); /* read wfl */
if(! sf_getint("nhz",&nhz)) nhz=0; /* z lags */
dhz=2*sf_d(az);
ahz=sf_maxa(2*nhz+1,-nhz*dhz,dhz); sf_setlabel(ahz,"hz"); sf_setunit(ahz,"");
if(verb) sf_raxa(ahz);
if(! sf_getint("nhx",&nhx)) nhx=0; /* x lags */
dhx=2*sf_d(ax);
ahx=sf_maxa(2*nhx+1,-nhx*dhx,dhx); sf_setlabel(ahx,"hx"); sf_setunit(ahx,"");
if(verb) sf_raxa(ahx);
if(! sf_getint("nhy",&nhy)) nhy=0; /* y lags */
dhy=2*sf_d(ay);
ahy=sf_maxa(2*nhy+1,-nhy*dhy,dhy); sf_setlabel(ahy,"hy"); sf_setunit(ahy,"");
if(verb) sf_raxa(ahy);
if(! sf_getint("nht",&nht)) nht=0; /* t lags */
dht=2*sf_d(at);
aht=sf_maxa(2*nht+1,-nht*dht,dht); sf_setlabel(aht,"ht"); sf_setunit(aht,"");
if(verb) sf_raxa(aht);
Fimg = sf_output("out"); /* write img */
sf_oaxa(Fimg,ahz,1);
sf_oaxa(Fimg,ahx,2);
sf_oaxa(Fimg,ahy,3);
sf_oaxa(Fimg,aht,4);
sf_oaxa(Fimg, ac,5);
}
/*------------------------------------------------------------*/
if(! sf_getbool("gaus",&gaus)) gaus=false; /* Gaussian taper */
if(gaus) {
if(! sf_getfloat("gsx",&gsx)) gsx=0.25*sf_n(ahx)*sf_d(ahx); gsx=(nhx==0)?1:1./(2*gsx*gsx);
if(! sf_getfloat("gsy",&gsy)) gsy=0.25*sf_n(ahy)*sf_d(ahy); gsy=(nhy==0)?1:1./(2*gsy*gsy);
if(! sf_getfloat("gsz",&gsz)) gsz=0.25*sf_n(ahz)*sf_d(ahz); gsz=(nhz==0)?1:1./(2*gsz*gsz);
if(! sf_getfloat("gst",&gst)) gst=0.25*sf_n(aht)*sf_d(aht); gst=(nht==0)?1:1./(2*gst*gst);
}
/*------------------------------------------------------------*/
/* allocate arrays */
opr=sf_floatalloc4(nz,nx,ny,sf_n(aht));
wfl=sf_floatalloc4(nz,nx,ny,sf_n(aht));
img=sf_floatalloc5(sf_n(ahz),sf_n(ahx),sf_n(ahy),sf_n(aht),sf_n(ac));
/*------------------------------------------------------------*/
/* CIP coordinates */
cc= (pt3d*) sf_alloc(nc,sizeof(*cc));
pt3dread1(Fcip,cc,nc,3);
mcxall=sf_intalloc2(sf_n(ahx),sf_n(ac));
pcxall=sf_intalloc2(sf_n(ahx),sf_n(ac));
mcyall=sf_intalloc2(sf_n(ahy),sf_n(ac));
pcyall=sf_intalloc2(sf_n(ahy),sf_n(ac));
mczall=sf_intalloc2(sf_n(ahz),sf_n(ac));
pczall=sf_intalloc2(sf_n(ahz),sf_n(ac));
ccin=sf_boolalloc(sf_n(ac));
cxmin = sf_o(ax) + nhx *sf_d(ax);
cxmax = sf_o(ax) + (sf_n(ax)-1-nhx)*sf_d(ax);
cymin = sf_o(ay) + nhy *sf_d(ay);
cymax = sf_o(ay) + (sf_n(ay)-1-nhy)*sf_d(ay);
czmin = sf_o(az) + nhz *sf_d(az);
czmax = sf_o(az) + (sf_n(az)-1-nhz)*sf_d(az);
for(ic=0;ic<nc;ic++) {
ccin[ic]=(cc[ic].x>=cxmin && cc[ic].x<=cxmax &&
cc[ic].y>=cymin && cc[ic].y<=cymax &&
cc[ic].z>=czmin && cc[ic].z<=czmax)?true:false;
if(ccin[ic]) {
icx = 0.5+(cc[ic].x-sf_o(ax))/sf_d(ax);
for(ihx=-nhx; ihx<nhx+1; ihx++) {
mcxall[ic][nhx+ihx] = icx-ihx;
pcxall[ic][nhx+ihx] = icx+ihx;
}
icy = 0.5+(cc[ic].y-sf_o(ay))/sf_d(ay);
for(ihy=-nhy; ihy<nhy+1; ihy++) {
mcyall[ic][nhy+ihy] = icy-ihy;
pcyall[ic][nhy+ihy] = icy+ihy;
}
icz = 0.5+(cc[ic].z-sf_o(az))/sf_d(az);
for(ihz=-nhz; ihz<nhz+1; ihz++) {
mczall[ic][nhz+ihz] = icz-ihz;
pczall[ic][nhz+ihz] = icz+ihz;
}
}
}
mctall=sf_intalloc(sf_n(aht));
pctall=sf_intalloc(sf_n(aht));
for (iht=0; iht<sf_n(aht); iht++) {
mctall[iht]= iht;
pctall[iht]=sf_n(aht)-1-iht;
}
if(adj) { /* ADJIONT OPERATOR */
for(iht=0;iht<sf_n(aht);iht++)
CICLOOP( wfl[iht][iy][ix][iz]=0; ); /* zero wfl */
for(it=0;it<nt;it++) sf_floatwrite(wfl[0][0][0],nz*nx*ny,Fwfl); /* reserve wfl */
sf_seek(Fwfl,0,SEEK_SET); /* seek back */
sf_floatread(img[0][0][0][0],sf_n(ac)*sf_n(ahy)*sf_n(ahx)*sf_n(ahz)*sf_n(aht),Fimg); /* read img */
; applyScaling (img,ac,aht,ahx,ahy,ahz,scale); /* scaling */
if(gaus) applyGaussian(img,ac,aht,ahx,ahy,ahz,gst,gsx,gsy,gsz); /* Gaussian */
lht=0; itO=-999999; itW=-999999;
for(it=-nht;it<nt+nht;it++) { if(verb) fprintf(stderr,"\b\b\b\b\b\b%04d",it);
fht=(lht+1) % sf_n(aht);
if(it<nt-nht) {
itO = it + nht;
if( !oprcausal ) sf_seek(Fopr,(off_t)(nt-1-itO)*nslice,SEEK_SET);
else sf_seek(Fopr,(off_t) itO *nslice,SEEK_SET);
sf_floatread(opr[ lht ][0][0],nz*nx*ny,Fopr);
}
for(iht=0;iht<sf_n(aht);iht++) {
mctall[iht] = (mctall[iht]+1) % sf_n(aht); /* cycle iht index */
pctall[iht] = (pctall[iht]+1) % sf_n(aht);
}
if(it>=0+nht &&
it<nt-nht) {
#ifdef _OPENMP
#pragma omp parallel for schedule(dynamic) \
private(ic, ihx,ihy,ihz,iht,mcx, mcy, mcz, mct, pcx, pcy, pcz, pct) \
shared (nc,ccin,ahx,ahy,ahz,aht,mcxall,mcyall,mczall,mctall,pcxall,pcyall,pczall,pctall)
#endif
for(ic=0;ic<nc;ic++){ if(ccin[ic]) { /* sum over c only! */
if(pos){
EICLOOP( wfl [mct][mcy][mcx][mcz] +=
opr [pct][pcy][pcx][pcz] *
img[ic][iht][ihy][ihx][ihz]; );
}else{
EICLOOP( wfl [pct][pcy][pcx][pcz] +=
opr [mct][mcy][mcx][mcz] *
img[ic][iht][ihy][ihx][ihz]; );
}
}
}
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);
}
}
