void homogeneousg(float *HomG, complex *cshot, complex *Refl, int nx, int nt, int nxs, int nts, float dt, float *xsyn, int Nsyn, float *xrcv, float *xsrc, float fxs2, float fxs, float dxs, float dxsrc, float dx, int ixa, int ixb, int ntfft, int nw, int nw_low, int nw_high, int mode, int reci, int nshots, int *ixpossyn, int npossyn, float *pmin, float *f1min, float *f1plus, float *f2p, float *G_d, int *muteW, int smooth, int shift, int above, int pad, int nt0, int *synpos, int verbose)
{
int i, j, l, ret;
int iter, niter, ix, nfreq;
float *iRN, *rtrace;
complex *Fop, *ctrace, *chom;
double t0, t2, tfft;
tfft = 0.0;
ret = 0;
t0 = wallclock_time();
nfreq = ntfft/2+1;
#pragma omp parallel default(shared) \
private(i,j,ctrace,chom,rtrace)
{
ctrace = (complex *)calloc(nfreq,sizeof(complex));
chom = (complex *)calloc(nfreq,sizeof(complex));
rtrace = (float *)calloc(ntfft,sizeof(float));
#pragma omp for
for (l = 0; l < Nsyn; l++) {
if (verbose > 2) vmess("Creating Homogeneous G at location %d out of %d",l+1,Nsyn);
/* Construct the image */
for (i = 0; i < nxs; i++) {
rc1fft(&f2p[l*nxs*ntfft+i*ntfft],ctrace,nt,-1);
for (j = 0; j < nfreq; j++) {
chom[j].r += 2*(ctrace[j].r*cshot[i*nfreq+j].r - ctrace[j].i*cshot[i*nfreq+j].i);
}
}
cr1fft(&chom[0],rtrace,nt,1);
for (i = 0; i < ntfft; i++) {
HomG[i*Nsyn+synpos[l]] = rtrace[i];
}
/*for (i = 0; i < ntfft/2; i++) {
HomG[i*Nsyn+synpos[l]] = rtrace[ntfft/2+i];
}
for (i = ntfft/2; i < ntfft; i++) {
HomG[i*Nsyn+synpos[l]] = rtrace[i-ntfft/2];
}*/
}
free(rtrace);free(chom);free(ctrace);
}
//free(Gmin);
t2 = wallclock_time();
if (verbose) {
vmess("Total Homogeneous G time = %.3f", t2-t0);
}
return;
}
int main(int argc, char **argv)
{
modPar mod;
recPar rec;
srcPar src;
shotPar shot;
rayPar ray;
float *velocity, *slowness, *smooth, *trueslow, **inter;
double t0, t1, t2, tinit, tray, tio;
size_t size;
int nw, n1, ix, iz, ir, ixshot, izshot;
int nt, ntfft, nfreq, ig;
int irec, sbox, ipos, nrx, nrz, nr;
fcoord coordsx, coordgx, Time;
icoord grid, isrc;
float Jr, *ampl, *time, *ttime, *ttime_p, cp_average, *wavelet, dw, dt;
float dxrcv, dzrcv, rdelay, tr, dt_tmp;
segy hdr;
char filetime[1024], fileamp[1024], *method, *file_rcvtime, *file_src;
size_t nwrite, nread;
int verbose;
complex *cmute, *cwav;
FILE *fpt, *fpa, *fpwav, *fprcv;
t0= wallclock_time();
initargs(argc,argv);
requestdoc(0);
if(!getparint("verbose",&verbose)) verbose=0;
if(!getparint("sbox", &sbox)) sbox = 1;
if(!getparstring("method", &method)) method="jesper";
if (!getparfloat("rec_delay",&rdelay)) rdelay=0.0;
getParameters(&mod, &rec, &src, &shot, &ray, verbose);
/* read file_src if file_rcvtime is defined */
if (!getparstring("file_rcvtime",&file_rcvtime)) file_rcvtime=NULL;
if (file_rcvtime != NULL) {
if (!getparstring("file_src",&file_src)) file_src=NULL;
if (!getparfloat("dt",&dt)) dt=0.004;
if (file_src != NULL ) {
fpwav = fopen( file_src, "r" );
assert( fpwav != NULL);
nread = fread( &hdr, 1, TRCBYTES, fpwav );
assert(nread == TRCBYTES);
ntfft = optncr(MAX(hdr.ns, rec.nt));
wavelet = (float *)calloc(ntfft,sizeof(float));
/* read first trace */
nread = fread(wavelet, sizeof(float), hdr.ns, fpwav);
assert (nread == hdr.ns);
fclose(fpwav);
}
else {
ntfft = optncr(rec.nt);
wavelet = (float *)calloc(ntfft,sizeof(float));
wavelet[0] = 1.0;
}
nfreq = ntfft/2+1;
cwav = (complex *)calloc(nfreq,sizeof(complex));
cmute = (complex *)calloc(nfreq,sizeof(complex));
rc1fft(wavelet,cwav,ntfft,-1);
dw = 2*M_PI/(ntfft*dt);
}
/* allocate arrays for model parameters: the different schemes use different arrays */
n1 = mod.nz;
if(!strcmp(method,"fd")) nw = 0;
else nw = ray.smoothwindow;
velocity = (float *)calloc(mod.nx*mod.nz,sizeof(float));
slowness = (float *)calloc((mod.nx+2*nw)*(mod.nz+2*nw),sizeof(float));
trueslow = (float *)calloc(mod.nx*mod.nz,sizeof(float));
if(!strcmp(method,"fd")) {
ttime = (float *)calloc(mod.nx*mod.nz,sizeof(float));
}
/* read velocity and density files */
readModel(mod, velocity, slowness, nw);
/* allocate arrays for wavefield and receiver arrays */
size = shot.n*rec.n;
time = (float *)calloc(size,sizeof(float));
ampl = (float *)calloc(size,sizeof(float));
/* Sinking source and receiver arrays:
If P-velocity==0 the source and receiver
postions are placed deeper until the P-velocity changes.
Setting the option rec.sinkvel only sinks the receiver position
(not the source) and uses the velocity
of the first receiver to sink through to the next layer. */
/* sink receivers to value different than sinkvel */
for (ir=0; ir<rec.n; ir++) {
iz = rec.z[ir];
ix = rec.x[ir];
while(velocity[(ix)*n1+iz] == rec.sinkvel) iz++;
rec.z[ir]=iz+rec.sinkdepth;
rec.zr[ir]=rec.zr[ir]+(rec.z[ir]-iz)*mod.dz;
// rec.zr[ir]=rec.z[ir]*mod.dz;
if (verbose>3) vmess("receiver position %d at grid[ix=%d, iz=%d] = (x=%f z=%f)", ir, ix, rec.z[ir], rec.xr[ir]+mod.x0, rec.zr[ir]+mod.z0);
}
vmess(" - method for ray-tracing = %s", method);
/*
*/
/* sink sources to value different than zero */
for (izshot=0; izshot<shot.nz; izshot++) {
for (ixshot=0; ixshot<shot.nx; ixshot++) {
iz = shot.z[izshot];
ix = shot.x[ixshot];
while(velocity[(ix)*n1+iz] == 0.0) iz++;
shot.z[izshot]=iz+src.sinkdepth;
}
}
if (verbose>3) writeSrcRecPos(&mod, &rec, &src, &shot);
/* smooth slowness grid */
grid.x = mod.nx;
grid.z = mod.nz;
grid.y = 1;
if ( nw != 0 ) { /* smooth slowness */
smooth = (float *)calloc(grid.x*grid.z,sizeof(float));
applyMovingAverageFilter(slowness, grid, nw, 2, smooth);
memcpy(slowness,smooth,grid.x*grid.z*sizeof(float));
free(smooth);
}
/* prepare output file and headers */
strcpy(filetime, rec.file_rcv);
name_ext(filetime, "_time");
fpt = fopen(filetime, "w");
assert(fpt != NULL);
if (ray.geomspread) {
strcpy(fileamp, rec.file_rcv);
name_ext(fileamp, "_amp");
fpa = fopen(fileamp, "w");
assert(fpa != NULL);
}
if (file_rcvtime != NULL) {
fprcv = fopen(file_rcvtime, "w");
assert(fprcv != NULL);
}
memset(&hdr,0,sizeof(hdr));
hdr.scalco = -1000;
hdr.scalel = -1000;
hdr.trid = 0;
t1=wallclock_time();
tinit = t1-t0;
tray=0.0;
tio=0.0;
/* Outer loop over number of shots */
for (izshot=0; izshot<shot.nz; izshot++) {
for (ixshot=0; ixshot<shot.nx; ixshot++) {
t2=wallclock_time();
if (verbose) {
vmess("Modeling source %d at gridpoints ix=%d iz=%d", (izshot*shot.n)+ixshot, shot.x[ixshot], shot.z[izshot]);
vmess(" which are actual positions x=%.2f z=%.2f", mod.x0+mod.dx*shot.x[ixshot], mod.z0+mod.dz*shot.z[izshot]);
vmess("Receivers at gridpoint x-range ix=%d - %d", rec.x[0], rec.x[rec.n-1]);
vmess(" which are actual positions x=%.2f - %.2f", mod.x0+rec.xr[0], mod.x0+rec.xr[rec.n-1]);
vmess("Receivers at gridpoint z-range iz=%d - %d", rec.z[0], rec.z[rec.n-1]);
vmess(" which are actual positions z=%.2f - %.2f", mod.z0+rec.zr[0], mod.z0+rec.zr[rec.n-1]);
}
coordsx.x = shot.x[ixshot]*mod.dx;
coordsx.z = shot.z[izshot]*mod.dz;
coordsx.y = 0;
t1=wallclock_time();
tio += t1-t2;
if (!strcmp(method,"jesper")) {
#pragma omp parallel for default(shared) \
private (coordgx,irec,Time,Jr)
for (irec=0; irec<rec.n; irec++) {
coordgx.x=rec.xr[irec];
coordgx.z=rec.zr[irec];
coordgx.y = 0;
getWaveParameter(slowness, grid, mod.dx, coordsx, coordgx, ray, &Time, &Jr);
time[((izshot*shot.nx)+ixshot)*rec.n + irec] = Time.x + Time.y + 0.5*Time.z;
ampl[((izshot*shot.nx)+ixshot)*rec.n + irec] = Jr;
if (verbose>4) vmess("JS: shot=%f,%f receiver at %f,%f T0=%f T1=%f T2=%f Jr=%f",coordsx.x, coordsx.z, coordgx.x, coordgx.z, Time.x, Time.y, Time.z, Jr);
}
}
else if(!strcmp(method,"fd")) {
int mzrcv;
isrc.x = shot.x[ixshot];
isrc.y = 0;
isrc.z = shot.z[izshot];
mzrcv = 0;
for (irec = 0; irec < rec.n; irec++) mzrcv = MAX(rec.z[irec], mzrcv);
vidale(ttime,slowness,&isrc,grid,mod.dx,sbox, mzrcv);
for (irec=0; irec<rec.n; irec++) {
coordgx.x=mod.x0+rec.xr[irec];
coordgx.z=mod.z0+rec.zr[irec];
coordgx.y = 0;
ipos = ((izshot*shot.nx)+ixshot)*rec.n + irec;
time[ipos] = ttime[rec.z[irec]*mod.nx+rec.x[irec]];
/* compute average velocity between source and receiver */
nrx = (rec.x[irec]-isrc.x);
nrz = (rec.z[irec]-isrc.z);
nr = abs(nrx) + abs(nrz);
cp_average = 0.0;
for (ir=0; ir<nr; ir++) {
ix = isrc.x + floor((ir*nrx)/nr);
iz = isrc.z + floor((ir*nrz)/nr);
//fprintf(stderr,"ir=%d ix=%d iz=%d velocity=%f\n", ir, ix, iz, velocity[ix*mod.nz+iz]);
cp_average += velocity[ix*mod.nz+iz];
}
cp_average = cp_average/((float)nr);
ampl[ipos] = sqrt(time[ipos]*cp_average);
if (verbose>4) vmess("FD: shot=%f,%f receiver at %f(%d),%f(%d) T=%e V=%f Ampl=%f",coordsx.x, coordsx.z, coordgx.x, rec.x[irec], coordgx.z, rec.z[irec], time[ipos], cp_average, ampl[ipos]);
}
}
t2=wallclock_time();
tray += t2-t1;
hdr.sx = 1000*(mod.x0+mod.dx*shot.x[ixshot]);
hdr.sdepth = 1000*(mod.z0+mod.dz*shot.z[izshot]);
hdr.selev = (int)(-1000.0*(mod.z0+mod.dz*shot.z[izshot]));
hdr.fldr = ((izshot*shot.nx)+ixshot)+1;
hdr.tracl = ((izshot*shot.nx)+ixshot)+1;
hdr.tracf = ((izshot*shot.nx)+ixshot)+1;
hdr.ntr = shot.n;
hdr.dt = (unsigned short)1;
hdr.trwf = shot.n;
hdr.ns = rec.n;
//hdr.d1 = (rec.x[1]-rec.x[0])*mod.dx; // discrete
hdr.d1 = (rec.xr[1]-rec.xr[0]);
hdr.f1 = mod.x0+rec.x[0]*mod.dx;
hdr.d2 = (shot.x[MIN(shot.n-1,1)]-shot.x[0])*mod.dx;
hdr.f2 = mod.x0+shot.x[0]*mod.dx;
dt_tmp = (fabs(hdr.d1*((float)hdr.scalco)));
hdr.dt = (unsigned short)dt_tmp;
nwrite = fwrite( &hdr, 1, TRCBYTES, fpt);
assert(nwrite == TRCBYTES);
nwrite = fwrite( &time[((izshot*shot.nx)+ixshot)*rec.n], sizeof(float), rec.n, fpt);
assert(nwrite == rec.n);
fflush(fpt);
if (ray.geomspread) {
nwrite = fwrite( &hdr, 1, TRCBYTES, fpa);
assert(nwrite == TRCBYTES);
nwrite = fwrite( &l[((izshot*shot.nx)+ixshot)*rec.n], sizeof(float), rec.n, fpa);
assert(nwrite == rec.n);
fflush(fpa);
}
if (file_rcvtime != NULL) {
hdr.ns = rec.nt;
hdr.trwf = rec.n;
hdr.ntr = ((izshot*shot.nx)+ixshot+1)*rec.n;
hdr.dt = dt*1000000;
hdr.d1 = dt;
hdr.f1 = 0.0;
hdr.d2 = (rec.xr[1]-rec.xr[0]);
hdr.f2 = mod.x0+rec.x[0]*mod.dx;
for (irec=0; irec<rec.n; irec++) {
ipos = ((izshot*shot.nx)+ixshot)*rec.n + irec;
hdr.tracf = irec+1;
hdr.tracl = ((izshot*shot.nx)+ixshot*shot.nz)+irec+1;
hdr.gx = 1000*(mod.x0+rec.xr[irec]);
hdr.offset = (rec.xr[irec]-shot.x[ixshot]*mod.dx);
hdr.gelev = (int)(-1000*(mod.z0+rec.zr[irec]));
tr = time[ipos]+rdelay;
for (ig=0; ig<nfreq; ig++) {
cmute[ig].r = (cwav[ig].r*cos(ig*dw*tr-M_PI/4.0)-cwav[ig].i*sin(ig*dw*tr-M_PI/4.0))/(ntfft*ampl[ipos]);
cmute[ig].i = (cwav[ig].i*cos(ig*dw*tr-M_PI/4.0)+cwav[ig].r*sin(ig*dw*tr-M_PI/4.0))/(ntfft*ampl[ipos]);
}
cr1fft(cmute,wavelet,ntfft,-1);
nwrite = fwrite( &hdr, 1, TRCBYTES, fprcv);
nwrite = fwrite( wavelet, sizeof(float), rec.nt, fprcv );
}
}
t1=wallclock_time();
tio += t1-t2;
} /* end of ixshot loop */
} /* end of loop over number of shots */
fclose(fpt);
if (file_rcvtime != NULL) fclose(fprcv);
if (ray.geomspread) fclose(fpa);
t1= wallclock_time();
if (verbose) {
vmess("*******************************************");
vmess("************* runtime info ****************");
vmess("*******************************************");
vmess("Total compute time ray-tracing = %.2f s.", t1-t0);
vmess(" - intializing arrays and model = %.3f", tinit);
vmess(" - ray tracing = %.3f", tray);
vmess(" - writing data to file = %.3f", tio);
}
/* free arrays */
initargs(argc,argv); /* this will free the arg arrays declared */
free(velocity);
free(slowness);
return 0;
}
void synthesis(complex *Refl, complex *Fop, float *Top, float *iRN, int nx, int nt, int nxs, int nts, float dt, float *xsyn, int Nsyn, float *xrcv, float *xsrc, float fxs2, float fxs, float dxs, float dxsrc, float dx, int ixa, int ixb, int ntfft, int nw, int nw_low, int nw_high, int mode, int reci, int nshots, int *ixpossyn, int npossyn, double *tfft, int *first, int verbose)
{
int nfreq, size, iox, inx;
float scl;
int i, j, l, m, iw, ix, k;
float *rtrace, idxs;
complex *sum, *ctrace;
int npe;
static int *ixrcv;
static double t0, t1, t;
size = nxs*nts;
nfreq = ntfft/2+1;
/* scale factor 1/N for backward FFT,
* scale dt for correlation/convolution along time,
* scale dx (or dxsrc) for integration over receiver (or shot) coordinates */
scl = 1.0*dt/((float)ntfft);
#ifdef _OPENMP
npe = omp_get_max_threads();
/* parallelisation is over number of virtual source positions (Nsyn) */
if (npe > Nsyn) {
vmess("Number of OpenMP threads set to %d (was %d)", Nsyn, npe);
omp_set_num_threads(Nsyn);
}
#endif
t0 = wallclock_time();
/* reset output data to zero */
memset(&iRN[0], 0, Nsyn*nxs*nts*sizeof(float));
idxs = 1.0/dxs;
if (ixrcv == NULL) {
ixrcv = (int *)malloc(nshots*nx*sizeof(int));
}
for (k=0; k<nshots; k++) {
for (i = 0; i < nx; i++) {
ixrcv[k*nx+i] = NINT((xrcv[k*nx+i]-fxs)*idxs);
}
}
ctrace = (complex *)calloc(ntfft,sizeof(complex));
if (!*first) {
/* transform muted Ni (Top) to frequency domain, input for next iteration */
for (l = 0; l < Nsyn; l++) {
/* set Fop to zero, so new operator can be defined within ixpossyn points */
//memset(&Fop[l*nxs*nw].r, 0, nxs*nw*2*sizeof(float));
bzero(&Fop[l*nxs*nw].r, nxs*nw*2*sizeof(float));
for (i = 0; i < npossyn; i++) {
rc1fft(&Top[l*size+i*nts],ctrace,ntfft,-1);
ix = ixpossyn[i];
for (iw=0; iw<nw; iw++) {
Fop[l*nxs*nw+iw*nxs+ix].r = ctrace[nw_low+iw].r;
Fop[l*nxs*nw+iw*nxs+ix].i = mode*ctrace[nw_low+iw].i;
}
}
}
}
else { /* only for first call to synthesis */
/* transform G_d to frequency domain, over all nxs traces */
*first=0;
for (l = 0; l < Nsyn; l++) {
/* set Fop to zero, so new operator can be defined within all ix points */
//memset(&Fop[l*nxs*nw].r, 0, nxs*nw*2*sizeof(float));
bzero(&Fop[l*nxs*nw].r, nxs*nw*2*sizeof(float));
for (i = 0; i < nxs; i++) {
rc1fft(&Top[l*size+i*nts],ctrace,ntfft,-1);
for (iw=0; iw<nw; iw++) {
Fop[l*nxs*nw+iw*nxs+i].r = ctrace[nw_low+iw].r;
Fop[l*nxs*nw+iw*nxs+i].i = mode*ctrace[nw_low+iw].i;
}
}
}
}
free(ctrace);
t1 = wallclock_time();
*tfft += t1 - t0;
for (k=0; k<nshots; k++) {
/* if (verbose>=3) {
vmess("source position: %.2f ixpossyn=%d", xsrc[k], ixpossyn[k]);
vmess("receiver positions: %.2f <--> %.2f", xrcv[k*nx+0], xrcv[k*nx+nx-1]);
}
*/
if ((NINT(xsrc[k]-fxs2) > 0) || (NINT(xrcv[k*nx+nx-1]-fxs2) > 0) ||
(NINT(xrcv[k*nx+nx-1]-fxs) < 0) || (NINT(xsrc[k]-fxs) < 0) ||
(NINT(xrcv[k*nx+0]-fxs) < 0) || (NINT(xrcv[k*nx+0]-fxs2) > 0) ) {
vwarn("source/receiver positions are outside synthesis model");
vwarn("integration calculation is stopped at gather %d", k);
vmess("xsrc = %.2f xrcv_1 = %.2f xrvc_N = %.2f", xsrc[k], xrcv[k*nx+0], xrcv[k*nx+nx-1]);
break;
}
iox = 0; inx = nx;
/*================ SYNTHESIS ================*/
#pragma omp parallel default(none) \
shared(iRN, dx, npe, nw, verbose) \
shared(Refl, Nsyn, reci, xrcv, xsrc, xsyn, fxs, nxs, dxs) \
shared(nx, ixa, ixb, dxsrc, iox, inx, k, nfreq, nw_low, nw_high) \
shared(Fop, size, nts, ntfft, scl, ixrcv, stderr) \
private(l, ix, j, m, i, sum, rtrace)
{ /* start of parallel region */
sum = (complex *)malloc(nfreq*sizeof(complex));
rtrace = (float *)calloc(ntfft,sizeof(float));
#pragma omp for schedule(guided,1)
for (l = 0; l < Nsyn; l++) {
ix = k;
/* multiply R with Fop and sum over nx */
memset(&sum[0].r,0,nfreq*2*sizeof(float));
//for (j = 0; j < nfreq; j++) sum[j].r = sum[j].i = 0.0;
for (j = nw_low, m = 0; j <= nw_high; j++, m++) {
for (i = iox; i < inx; i++) {
sum[j].r += Refl[k*nw*nx+m*nx+i].r*Fop[l*nw*nxs+m*nxs+ixrcv[k*nx+i]].r -
Refl[k*nw*nx+m*nx+i].i*Fop[l*nw*nxs+m*nxs+ixrcv[k*nx+i]].i;
sum[j].i += Refl[k*nw*nx+m*nx+i].i*Fop[l*nw*nxs+m*nxs+ixrcv[k*nx+i]].r +
Refl[k*nw*nx+m*nx+i].r*Fop[l*nw*nxs+m*nxs+ixrcv[k*nx+i]].i;
}
}
/* transfrom result back to time domain */
cr1fft(sum, rtrace, ntfft, 1);
/* dx = receiver distance */
for (j = 0; j < nts; j++)
iRN[l*size+ix*nts+j] += rtrace[j]*scl*dx;
} /* end of parallel Nsyn loop */
free(sum);
free(rtrace);
#pragma omp single
{
#ifdef _OPENMP
npe = omp_get_num_threads();
#endif
}
} /* end of parallel region */
if (verbose>3) vmess("*** Shot gather %d processed ***", k);
} /* end of nshots (k) loop */
t = wallclock_time() - t0;
if (verbose) {
vmess("OMP: parallel region = %f seconds (%d threads)", t, npe);
}
return;
}
int main (int argc, char **argv)
{
FILE *fp_out, *fp_f1plus, *fp_f1min;
FILE *fp_gmin, *fp_gplus, *fp_f2, *fp_pmin;
int i, j, l, ret, nshots, Nsyn, nt, nx, nts, nxs, ngath;
int size, n1, n2, ntap, tap, di, ntraces, nb, ib;
int nw, nw_low, nw_high, nfreq, *xnx, *xnxsyn, *synpos;
int reci, mode, ixa, ixb, n2out, verbose, ntfft;
int iter, niter, niterh, tracf, *muteW, pad, nt0, ampest, *hmuteW, *hxnxsyn;
int hw, smooth, above, shift, *ixpossyn, npossyn, ix, first=1;
float fmin, fmax, *tapersh, *tapersy, fxf, dxf, fxs2, *xsrc, *xrcv, *zsyn, *zsrc, *xrcvsyn;
float *hzsyn, *hxsyn, *hxrcvsyn, *hG_d, xloc, zloc, *HomG;
double t0, t1, t2, t3, tsyn, tread, tfft, tcopy, energyNi, *J;
float d1, d2, f1, f2, fxs, ft, fx, *xsyn, dxsrc, Q, f0, *Costdet;
float *green, *f2p, *pmin, *G_d, dt, dx, dxs, scl, mem, *Image, *Image2;
float *f1plus, *f1min, *iRN, *Ni, *trace, *Gmin, *Gplus, *Gm0;
float xmin, xmax, weight, tsq, *Gd, *amp, bstart, bend, db, *bdet, bp, b, bmin;
complex *Refl, *Fop, *cshot;
char *file_tinv, *file_shot, *file_green, *file_iter, *file_wav, *file_ray, *file_amp, *file_img, *file_cp, *file_rays, *file_amps;
char *file_f1plus, *file_f1min, *file_gmin, *file_gplus, *file_f2, *file_pmin, *wavtype, *wavtype2, *file_homg, *file_tinvs;
segy *hdrs_im, *hdrs_homg;
WavePar WP,WPs;
modPar mod;
recPar rec;
srcPar src;
shotPar shot;
rayPar ray;
initargs(argc, argv);
requestdoc(1);
tsyn = tread = tfft = tcopy = 0.0;
t0 = wallclock_time();
if (!getparstring("file_img", &file_img)) file_img = "img.su";
if (!getparstring("file_homg", &file_homg)) file_homg = NULL;
if (!getparstring("file_shot", &file_shot)) file_shot = NULL;
if (!getparstring("file_tinv", &file_tinv)) file_tinv = NULL;
if (!getparstring("file_tinvs", &file_tinvs)) file_tinvs = NULL;
if (!getparstring("file_f1plus", &file_f1plus)) file_f1plus = NULL;
if (!getparstring("file_f1min", &file_f1min)) file_f1min = NULL;
if (!getparstring("file_gplus", &file_gplus)) file_gplus = NULL;
if (!getparstring("file_gmin", &file_gmin)) file_gmin = NULL;
if (!getparstring("file_pplus", &file_f2)) file_f2 = NULL;
if (!getparstring("file_f2", &file_f2)) file_f2 = NULL;
if (!getparstring("file_pmin", &file_pmin)) file_pmin = NULL;
if (!getparstring("file_iter", &file_iter)) file_iter = NULL;
if (!getparstring("file_wav", &file_wav)) file_wav=NULL;
if (!getparstring("file_ray", &file_ray)) file_ray=NULL;
if (!getparstring("file_amp", &file_amp)) file_amp=NULL;
if (!getparstring("file_rays", &file_rays)) file_rays=NULL;
if (!getparstring("file_amps", &file_amps)) file_amps=NULL;
if (!getparstring("file_cp", &file_cp)) file_cp = NULL;
if (!getparint("verbose", &verbose)) verbose = 0;
if (file_tinv == NULL && file_shot == NULL)
verr("file_tinv and file_shot cannot be both input pipe");
if (!getparstring("file_green", &file_green)) {
if (verbose) vwarn("parameter file_green not found, assume pipe");
file_green = NULL;
}
if (!getparfloat("fmin", &fmin)) fmin = 0.0;
if (!getparfloat("fmax", &fmax)) fmax = 70.0;
if (!getparint("ixa", &ixa)) ixa = 0;
if (!getparint("ixb", &ixb)) ixb = ixa;
// if (!getparint("reci", &reci)) reci = 0;
reci=0; // source-receiver reciprocity is not yet fully build into the code
if (!getparfloat("weight", &weight)) weight = 1.0;
if (!getparfloat("tsq", &tsq)) tsq = 0.0;
if (!getparfloat("Q", &Q)) Q = 0.0;
if (!getparfloat("f0", &f0)) f0 = 0.0;
if (!getparint("tap", &tap)) tap = 0;
if (!getparint("ntap", &ntap)) ntap = 0;
if (!getparint("pad", &pad)) pad = 0;
if(!getparint("hw", &hw)) hw = 15;
if(!getparint("smooth", &smooth)) smooth = 5;
if(!getparint("above", &above)) above = 0;
if(!getparint("shift", &shift)) shift=12;
if(!getparint("ampest", &est)) ampest=0;
if(!getparint("nb", &nb)) nb=0;
if (!getparfloat("bstart", &bstart)) bstart = 1.0;
if (!getparfloat("bend", &bend)) bend = 1.0;
if (reci && ntap) vwarn("tapering influences the reciprocal result");
/* Reading in wavelet parameters */
if(!getparfloat("fpw", &WP.fp)) WP.fp = -1.0;
if(!getparfloat("fminw", &WP.fmin)) WP.fmin = 10.0;
if(!getparfloat("flefw", &WP.flef)) WP.flef = 20.0;
if(!getparfloat("frigw", &WP.frig)) WP.frig = 50.0;
if(!getparfloat("fmaxw", &WP.fmax)) WP.fmax = 60.0;
else WP.fp = -1;
if(!getparfloat("dbw", &WP.db)) WP.db = -20.0;
if(!getparfloat("t0w", &WP.t0)) WP.t0 = 0.0;
if(!getparint("shiftw", &WP.shift)) WP.shift = 0;
if(!getparint("invw", &WP.inv)) WP.inv = 0;
if(!getparfloat("epsw", &WP.eps)) WP.eps = 1.0;
if(!getparfloat("scalew", &WP.scale)) WP.scale = 1.0;
if(!getparint("scfftw", &WP.scfft)) WP.scfft = 1;
if(!getparint("cmw", &WP.cm)) WP.cm = 10;
if(!getparint("cnw", &WP.cn)) WP.cn = 1;
if(!getparint("wav", &WP.wav)) WP.wav = 0;
if(!getparstring("file_wav", &WP.file_wav)) WP.file_wav=NULL;
if(!getparstring("w", &wavtype)) strcpy(WP.w, "g2");
else strcpy(WP.w, wavtype);
if(!getparfloat("fpws", &WPs.fp)) WPs.fp = -1.0;
if(!getparfloat("fminws", &WPs.fmin)) WPs.fmin = 10.0;
if(!getparfloat("flefws", &WPs.flef)) WPs.flef = 20.0;
if(!getparfloat("frigws", &WPs.frig)) WPs.frig = 50.0;
if(!getparfloat("fmaxws", &WPs.fmax)) WPs.fmax = 60.0;
else WPs.fp = -1;
if(!getparfloat("dbw", &WPs.db)) WPs.db = -20.0;
if(!getparfloat("t0ws", &WPs.t0)) WPs.t0 = 0.0;
if(!getparint("shiftws", &WPs.shift)) WPs.shift = 0;
if(!getparint("invws", &WPs.inv)) WPs.inv = 0;
if(!getparfloat("epsws", &WPs.eps)) WPs.eps = 1.0;
if(!getparfloat("scalews", &WPs.scale)) WPs.scale = 1.0;
if(!getparint("scfftws", &WPs.scfft)) WPs.scfft = 1;
if(!getparint("cmws", &WPs.cm)) WPs.cm = 10;
if(!getparint("cnws", &WPs.cn)) WPs.cn = 1;
if(!getparint("wavs", &WPs.wav)) WPs.wav = 0;
if(!getparstring("file_wavs", &WPs.file_wav)) WPs.file_wav=NULL;
if(!getparstring("ws", &wavtype2)) strcpy(WPs.w, "g2");
else strcpy(WPs.w, wavtype2);
if(!getparint("niter", &niter)) niter = 10;
if(!getparint("niterh", &niterh)) niterh = niter;
/*================ Reading info about shot and initial operator sizes ================*/
ngath = 0; /* setting ngath=0 scans all traces; n2 contains maximum traces/gather */
if (file_ray!=NULL && file_tinv==NULL) {
ret = getFileInfo(file_ray, &n2, &n1, &ngath, &d1, &d2, &f2, &f1, &xmin, &xmax, &scl, &ntraces);
n1 = 1;
ntraces = n2*ngath;
scl = 0.0010;
d1 = -1.0*xmin;
xmin = -1.0*xmax;
xmax = d1;
WP.wav = 1;
WP.xloc = -123456.0;
WP.zloc = -123456.0;
synpos = (int *)calloc(ngath,sizeof(int));
shot.nz = 1;
shot.nx = ngath;
shot.n = shot.nx*shot.nz;
for (l=0; l<shot.nz; l++) {
for (j=0; j<shot.nx; j++) {
synpos[l*shot.nx+j] = j*shot.nz+l;
}
}
}
else if (file_ray==NULL && file_tinv==NULL) {
getParameters(&mod, &rec, &src, &shot, &ray, verbose);
n1 = 1;
n2 = rec.n;
ngath = shot.n;
d1 = mod.dt;
d2 = (rec.x[1]-rec.x[0])*mod.dx;
f1 = 0.0;
f2 = mod.x0+rec.x[0]*mod.dx;
xmin = mod.x0+rec.x[0]*mod.dx;
xmax = mod.x0+rec.x[rec.n-1]*mod.dx;
scl = 0.0010;
ntraces = n2*ngath;
WP.wav = 1;
WP.xloc = -123456.0;
WP.zloc = -123456.0;
synpos = (int *)calloc(ngath,sizeof(int));
for (l=0; l<shot.nz; l++) {
for (j=0; j<shot.nx; j++) {
synpos[l*shot.nx+j] = j*shot.nz+l;
}
}
}
else {
ret = getFileInfo(file_tinv, &n1, &n2, &ngath, &d1, &d2, &f1, &f2, &xmin, &xmax, &scl, &ntraces);
}
Nsyn = ngath;
nxs = n2;
nts = n1;
nt0 = n1;
dxs = d2;
fxs = f2;
ngath = 0; /* setting ngath=0 scans all traces; nx contains maximum traces/gather */
ret = getFileInfo(file_shot, &nt, &nx, &ngath, &d1, &dx, &ft, &fx, &xmin, &xmax, &scl, &ntraces);
nshots = ngath;
assert (nxs >= nshots);
if (!getparfloat("dt", &dt)) dt = d1;
ntfft = optncr(MAX(nt+pad, nts+pad));
nfreq = ntfft/2+1;
nw_low = (int)MIN((fmin*ntfft*dt), nfreq-1);
nw_low = MAX(nw_low, 1);
nw_high = MIN((int)(fmax*ntfft*dt), nfreq-1);
nw = nw_high - nw_low + 1;
scl = 1.0/((float)ntfft);
if (nb > 1) {
db = (bend-bstart)/((float)(nb-1));
}
else if (nb == 1) {
db = 0;
bend = bstart;
}
/*================ Allocating all data arrays ================*/
green = (float *)calloc(Nsyn*nxs*ntfft,sizeof(float));
f2p = (float *)calloc(Nsyn*nxs*ntfft,sizeof(float));
pmin = (float *)calloc(Nsyn*nxs*ntfft,sizeof(float));
f1plus = (float *)calloc(Nsyn*nxs*ntfft,sizeof(float));
f1min = (float *)calloc(Nsyn*nxs*ntfft,sizeof(float));
G_d = (float *)calloc(Nsyn*nxs*ntfft,sizeof(float));
muteW = (int *)calloc(Nsyn*nxs,sizeof(int));
trace = (float *)malloc(ntfft*sizeof(float));
ixpossyn = (int *)malloc(nxs*sizeof(int));
xrcvsyn = (float *)calloc(Nsyn*nxs,sizeof(float));
xsyn = (float *)malloc(Nsyn*sizeof(float));
zsyn = (float *)malloc(Nsyn*sizeof(float));
xnxsyn = (int *)calloc(Nsyn,sizeof(int));
tapersy = (float *)malloc(nxs*sizeof(float));
Refl = (complex *)malloc(nw*nx*nshots*sizeof(complex));
tapersh = (float *)malloc(nx*sizeof(float));
xsrc = (float *)calloc(nshots,sizeof(float));
zsrc = (float *)calloc(nshots,sizeof(float));
xrcv = (float *)calloc(nshots*nx,sizeof(float));
xnx = (int *)calloc(nshots,sizeof(int));
/*================ Read and define mute window based on focusing operator(s) ================*/
/* G_d = p_0^+ = G_d (-t) ~ Tinv */
WPs.nt = ntfft;
WPs.dt = dt;
WP.nt = ntfft;
WP.dt = dt;
if (file_ray!=NULL || file_cp!=NULL) {
makeWindow(WP, file_ray, file_amp, dt, xrcvsyn, xsyn, zsyn, xnxsyn,
Nsyn, nxs, ntfft, mode, muteW, G_d, hw, verbose);
}
else {
mode=-1; /* apply complex conjugate to read in data */
readTinvData(file_tinv, dt, xrcvsyn, xsyn, zsyn, xnxsyn,
Nsyn, nxs, ntfft, mode, muteW, G_d, hw, verbose);
}
/* reading data added zero's to the number of time samples to be the same as ntfft */
nts = ntfft;
/* define tapers to taper edges of acquisition */
if (tap == 1 || tap == 3) {
for (j = 0; j < ntap; j++)
tapersy[j] = (cos(PI*(j-ntap)/ntap)+1)/2.0;
for (j = ntap; j < nxs-ntap; j++)
tapersy[j] = 1.0;
for (j = nxs-ntap; j < nxs; j++)
tapersy[j] =(cos(PI*(j-(nxs-ntap))/ntap)+1)/2.0;
}
else {
for (j = 0; j < nxs; j++) tapersy[j] = 1.0;
}
if (tap == 1 || tap == 3) {
if (verbose) vmess("Taper for operator applied ntap=%d", ntap);
for (l = 0; l < Nsyn; l++) {
for (i = 0; i < nxs; i++) {
for (j = 0; j < nts; j++) {
G_d[l*nxs*nts+i*nts+j] *= tapersy[i];
}
}
}
}
/* check consistency of header values */
dxf = (xrcvsyn[nxs-1] - xrcvsyn[0])/(float)(nxs-1);
if (NINT(dxs*1e3) != NINT(fabs(dxf)*1e3)) {
vmess("dx in hdr.d1 (%.3f) and hdr.gx (%.3f) not equal",d2, dxf);
if (dxf != 0) dxs = fabs(dxf);
vmess("dx in operator => %f", dxs);
}
if (xrcvsyn[0] != 0 || xrcvsyn[1] != 0 ) fxs = xrcvsyn[0];
fxs2 = fxs + (float)(nxs-1)*dxs;
/*================ Reading shot records ================*/
mode=1;
readShotData(file_shot, xrcv, xsrc, zsrc, xnx, Refl, nw, nw_low, ngath, nx, nx, ntfft,
mode, weight, tsq, Q, f0, verbose);
tapersh = (float *)malloc(nx*sizeof(float));
if (tap == 2 || tap == 3) {
for (j = 0; j < ntap; j++)
tapersh[j] = (cos(PI*(j-ntap)/ntap)+1)/2.0;
for (j = ntap; j < nx-ntap; j++)
tapersh[j] = 1.0;
for (j = nx-ntap; j < nx; j++)
tapersh[j] =(cos(PI*(j-(nx-ntap))/ntap)+1)/2.0;
}
else {
for (j = 0; j < nx; j++) tapersh[j] = 1.0;
}
if (tap == 2 || tap == 3) {
if (verbose) vmess("Taper for shots applied ntap=%d", ntap);
for (l = 0; l < nshots; l++) {
for (j = 1; j < nw; j++) {
for (i = 0; i < nx; i++) {
Refl[l*nx*nw+j*nx+i].r *= tapersh[i];
Refl[l*nx*nw+j*nx+i].i *= tapersh[i];
}
}
}
}
free(tapersh);
/* check consistency of header values */
fxf = xsrc[0];
if (nx > 1) dxf = (xrcv[0] - xrcv[nx-1])/(float)(nx-1);
else dxf = d2;
if (NINT(dx*1e3) != NINT(fabs(dxf)*1e3)) {
vmess("dx in hdr.d1 (%.3f) and hdr.gx (%.3f) not equal",dx, dxf);
if (dxf != 0) dx = fabs(dxf);
else verr("gx hdrs not set");
vmess("dx used => %f", dx);
}
dxsrc = (float)xsrc[1] - xsrc[0];
if (dxsrc == 0) {
vwarn("sx hdrs are not filled in!!");
dxsrc = dx;
}
/*================ Check the size of the files ================*/
if (NINT(dxsrc/dx)*dx != NINT(dxsrc)) {
vwarn("source (%.2f) and receiver step (%.2f) don't match",dxsrc,dx);
if (reci == 2) vwarn("step used from operator (%.2f) ",dxs);
}
di = NINT(dxf/dxs);
if ((NINT(di*dxs) != NINT(dxf)) && verbose)
vwarn("dx in receiver (%.2f) and operator (%.2f) don't match",dx,dxs);
if (nt != nts)
vmess("Time samples in shot (%d) and focusing operator (%d) are not equal",nt, nts);
if (verbose) {
vmess("Number of focusing operators = %d", Nsyn);
vmess("Number of receivers in focusop = %d", nxs);
vmess("number of shots = %d", nshots);
vmess("number of receiver/shot = %d", nx);
vmess("first model position = %.2f", fxs);
vmess("last model position = %.2f", fxs2);
vmess("first source position fxf = %.2f", fxf);
vmess("source distance dxsrc = %.2f", dxsrc);
vmess("last source position = %.2f", fxf+(nshots-1)*dxsrc);
vmess("receiver distance dxf = %.2f", dxf);
vmess("direction of increasing traces = %d", di);
vmess("number of time samples (nt,nts) = %d (%d,%d)", ntfft, nt, nts);
vmess("time sampling = %e ", dt);
if (ampest > 0) vmess("Amplitude correction estimation is switched on");
if (nb > 0) vmess("Scaling estimation in %d step(s) from %.3f to %.3f (db=%.3f)",nb,bstart,bend,db);
if (file_green != NULL) vmess("Green output file = %s ", file_green);
if (file_gmin != NULL) vmess("Gmin output file = %s ", file_gmin);
if (file_gplus != NULL) vmess("Gplus output file = %s ", file_gplus);
if (file_pmin != NULL) vmess("Pmin output file = %s ", file_pmin);
if (file_f2 != NULL) vmess("f2 (=pplus) output file = %s ", file_f2);
if (file_f1min != NULL) vmess("f1min output file = %s ", file_f1min);
if (file_f1plus != NULL)vmess("f1plus output file = %s ", file_f1plus);
if (file_iter != NULL) vmess("Iterations output file = %s ", file_iter);
}
/*================ initializations ================*/
if (ixa || ixb) n2out = ixa + ixb + 1;
else if (reci) n2out = nxs;
else n2out = nshots;
mem = Nsyn*n2out*ntfft*sizeof(float)/1048576.0;
if (verbose) {
vmess("number of output traces = %d", n2out);
vmess("number of output samples = %d", ntfft);
vmess("Size of output data/file = %.1f MB", mem);
}
//memcpy(Ni, G_d, Nsyn*nxs*ntfft*sizeof(float));
if (file_homg!=NULL) {
hG_d = (float *)calloc(nxs*ntfft,sizeof(float));
hmuteW = (int *)calloc(nxs,sizeof(int));
hxrcvsyn = (float *)calloc(nxs,sizeof(float));
hxsyn = (float *)calloc(1,sizeof(float));
hzsyn = (float *)calloc(1,sizeof(float));
hxnxsyn = (int *)calloc(1,sizeof(int));
cshot = (complex *)calloc(nxs*nfreq,sizeof(complex));
if(!getparfloat("xloc", &WPs.xloc)) WPs.xloc = -123456.0;
if(!getparfloat("zloc", &WPs.zloc)) WPs.zloc = -123456.0;
if (WPs.xloc == -123456.0 && WPs.zloc == -123456.0) file_cp = NULL;
if (WPs.xloc == -123456.0) WPs.xloc = 0.0;
if (WPs.zloc == -123456.0) WPs.zloc = 0.0;
xloc = WPs.xloc;
zloc = WPs.zloc;
ngath = 1;
if (file_rays!=NULL || file_cp!=NULL) {
WPs.wav=1;
makeWindow(WPs, file_rays, file_amps, dt, hxrcvsyn, hxsyn, hzsyn, hxnxsyn, ngath, nxs, ntfft, mode, hmuteW, hG_d, hw, verbose);
}
else {
mode=-1; /* apply complex conjugate to read in data */
readTinvData(file_tinvs, dt, hxrcvsyn, hxsyn, hzsyn, hxnxsyn,
ngath, nxs, ntfft, mode, hmuteW, hG_d, hw, verbose);
}
WPs.xloc = -123456.0;
WPs.zloc = -123456.0;
if (tap == 1 || tap == 3) {
if (verbose) vmess("Taper for operator applied ntap=%d", ntap);
for (i = 0; i < nxs; i++) {
for (j = 0; j < nts; j++) {
hG_d[i*nts+j] *= tapersy[i];
}
}
}
ngath = omp_get_max_threads();
synthesisPosistions(nx, nt, nxs, nts, dt, hxsyn, 1, xrcv, xsrc, fxs2, fxs,
dxs, dxsrc, dx, ixa, ixb, reci, nshots, ixpossyn, &npossyn, verbose);
iterations(Refl,nx,nt,nxs,nts,dt,hxsyn,1,xrcv,xsrc,fxs2,fxs,dxs,dxsrc,dx,ixa,ixb,
ntfft,nw,nw_low,nw_high,mode,reci,nshots,ixpossyn,npossyn,pmin,f1min,f1plus,
f2p,hG_d,hmuteW,smooth,shift,above,pad,nt0,&first,niterh,verbose);
/* compute full Green's function G = int R * f2(t) + f2(-t) = Pplus + Pmin */
for (i = 0; i < npossyn; i++) {
j = 0;
/* set green to zero if mute-window exceeds nt/2 */
if (hmuteW[ixpossyn[i]] >= nts/2) {
memset(&green[i*nts],0, sizeof(float)*nt);
continue;
}
green[i*nts+j] = f2p[i*nts+j] + pmin[i*nts+j];
for (j = 1; j < nts; j++) {
green[i*nts+j] = f2p[i*nts+nts-j] + pmin[i*nts+j];
}
}
applyMute(green, hmuteW, smooth, 4, 1, nxs, nts, ixpossyn, npossyn, shift, pad, nt0);
omp_set_num_threads(ngath);
/* Transform the green position to the frequency domain */
/*for (i = 0; i < npossyn; i++) {
rc1fft(&green[i*nts],&cshot[i*nfreq],ntfft,-1);
}*/
//free(hG_d);free(hmuteW);free(hxrcvsyn);
free(hmuteW);free(hxrcvsyn);
free(hxsyn);free(hzsyn);free(hxnxsyn);free(cshot);
}
/* dry-run of synthesis to get all x-positions calcalated by the integration */
synthesisPosistions(nx, nt, nxs, nts, dt, xsyn, Nsyn, xrcv, xsrc, fxs2, fxs,
dxs, dxsrc, dx, ixa, ixb, reci, nshots, ixpossyn, &npossyn, verbose);
if (verbose) {
vmess("synthesisPosistions: nshots=%d npossyn=%d", nshots, npossyn);
}
t1 = wallclock_time();
tread = t1-t0;
iterations(Refl,nx,nt,nxs,nts,dt,xsyn,Nsyn,xrcv,xsrc,fxs2,fxs,dxs,dxsrc,dx,ixa,ixb,
ntfft,nw,nw_low,nw_high,mode,reci,nshots,ixpossyn,npossyn,pmin,f1min,f1plus,
f2p,G_d,muteW,smooth,shift,above,pad,nt0,&first,niter,verbose);
/*if (niter==0) {
for (l = 0; l < Nsyn; l++) {
for (i = 0; i < npossyn; i++) {
j = 0;
ix = ixpossyn[i];
f2p[l*nxs*nts+i*nts+j] = G_d[l*nxs*nts+ix*nts+j];
f1plus[l*nxs*nts+i*nts+j] = G_d[l*nxs*nts+ix*nts+j];
for (j = 1; j < nts; j++) {
f2p[l*nxs*nts+i*nts+j] = G_d[l*nxs*nts+ix*nts+j];
f1plus[l*nxs*nts+i*nts+j] = G_d[l*nxs*nts+ix*nts+j];
}
}
}
}*/
if (niterh==0) {
for (l = 0; l < Nsyn; l++) {
for (i = 0; i < npossyn; i++) {
j = 0;
ix = ixpossyn[i];
green[i*nts+j] = hG_d[ix*nts+j];
for (j = 1; j < nts; j++) {
green[i*nts+j] = hG_d[ix*nts+nts-j];
}
}
}
}
if (file_img!=NULL) {
/*================ set variables for output data ================*/
hdrs_im = (segy *) calloc(shot.nx,sizeof(segy));
if (hdrs_im == NULL) verr("allocation for hdrs_out");
Image = (float *)calloc(Nsyn,sizeof(float));
first=0;
imaging(Image,WPs,Refl,nx,nt,nxs,nts,dt,xsyn,Nsyn,xrcv,xsrc,fxs2,fxs,dxs,dxsrc,dx,ixa,ixb,
ntfft,nw,nw_low,nw_high,mode,reci,nshots,ixpossyn,npossyn,pmin,f1min,f1plus,
f2p,G_d,muteW,smooth,shift,above,pad,nt0,synpos,verbose);
/*============= write output files ================*/
fp_out = fopen(file_img, "w+");
for (i = 0; i < shot.nx; i++) {
hdrs_im[i].fldr = 1;
hdrs_im[i].tracl = 1;
hdrs_im[i].tracf = i+1;
hdrs_im[i].scalco = -1000;
hdrs_im[i].scalel = -1000;
hdrs_im[i].sdepth = 0;
hdrs_im[i].trid = 1;
hdrs_im[i].ns = shot.nz;
hdrs_im[i].trwf = shot.nx;
hdrs_im[i].ntr = hdrs_im[i].fldr*hdrs_im[i].trwf;
hdrs_im[i].f1 = zsyn[0];
hdrs_im[i].f2 = xsyn[0];
hdrs_im[i].dt = dt*(1E6);
hdrs_im[i].d1 = (float)zsyn[shot.nx]-zsyn[0];
hdrs_im[i].d2 = (float)xsyn[1]-xsyn[0];
hdrs_im[i].sx = (int)roundf(xsyn[0] + (i*hdrs_im[i].d2));
hdrs_im[i].gx = (int)roundf(xsyn[0] + (i*hdrs_im[i].d2));
hdrs_im[i].offset = (hdrs_im[i].gx - hdrs_im[i].sx)/1000.0;
}
ret = writeData(fp_out, &Image[0], hdrs_im, shot.nz, shot.nx);
if (ret < 0 ) verr("error on writing output file.");
fclose(fp_out);
}
if (file_homg!=NULL) {
/*================ set variables for output data ================*/
hdrs_homg = (segy *) calloc(shot.nx,sizeof(segy));
if (hdrs_homg == NULL) verr("allocation for hdrs_out");
HomG = (float *)calloc(Nsyn*ntfft,sizeof(float));
homogeneousg(HomG,green,Refl,nx,nt,nxs,nts,dt,xsyn,Nsyn,xrcv,xsrc,fxs2,fxs,dxs,dxsrc,dx,ixa,ixb,
ntfft,nw,nw_low,nw_high,mode,reci,nshots,ixpossyn,npossyn,pmin,f1min,f1plus,
f2p,G_d,muteW,smooth,shift,above,pad,nt0,synpos,verbose);
/*============= write output files ================*/
fp_out = fopen(file_homg, "w+");
for (j = 0; j < ntfft; j++) {
for (i = 0; i < shot.nx; i++) {
hdrs_homg[i].fldr = j+1;
hdrs_homg[i].tracl = j*shot.nx+i+1;
hdrs_homg[i].tracf = i+1;
hdrs_homg[i].scalco = -1000;
hdrs_homg[i].scalel = -1000;
hdrs_homg[i].sdepth = (int)(zloc*1000.0);
hdrs_homg[i].trid = 1;
hdrs_homg[i].ns = shot.nz;
hdrs_homg[i].trwf = shot.nx;
hdrs_homg[i].ntr = hdrs_homg[i].fldr*hdrs_homg[i].trwf;
hdrs_homg[i].f1 = zsyn[0];
hdrs_homg[i].f2 = xsyn[0];
hdrs_homg[i].dt = dt*(1E6);
hdrs_homg[i].d1 = (float)zsyn[shot.nx]-zsyn[0];
hdrs_homg[i].d2 = (float)xsyn[1]-xsyn[0];
hdrs_homg[i].sx = (int)roundf(xsyn[0] + (i*hdrs_homg[i].d2));
hdrs_homg[i].gx = (int)roundf(xsyn[0] + (i*hdrs_homg[i].d2));
hdrs_homg[i].offset = (hdrs_homg[i].gx - hdrs_homg[i].sx)/1000.0;
}
ret = writeData(fp_out, &HomG[j*shot.n], hdrs_homg, shot.nz, shot.nx);
if (ret < 0 ) verr("error on writing output file.");
}
fclose(fp_out);
}
if (verbose) {
t1 = wallclock_time();
vmess("and CPU-time write data = %.3f", t1-t2);
}
free(tapersy);
exit(0);
}
void synthesis(complex *Refl, complex *Fop, float *Top, float *iRN, int nx, int nt, int nxs, int nts, float dt, float *xsyn, int
Nfoc, float *xrcv, float *xsrc, int *xnx, float fxse, float fxsb, float dxs, float dxsrc, float dx, int ntfft, int
nw, int nw_low, int nw_high, int mode, int reci, int nshots, int *ixpos, int npos, double *tfft, int *isxcount, int
*reci_xsrc, int *reci_xrcv, float *ixmask, int verbose)
{
int nfreq, size, inx;
float scl;
int i, j, l, m, iw, ix, k, ixsrc, il, ik;
float *rtrace, idxs;
complex *sum, *ctrace;
int npe;
static int first=1, *ixrcv;
static double t0, t1, t;
size = nxs*nts;
nfreq = ntfft/2+1;
/* scale factor 1/N for backward FFT,
* scale dt for correlation/convolution along time,
* scale dx (or dxsrc) for integration over receiver (or shot) coordinates */
scl = 1.0*dt/((float)ntfft);
#ifdef _OPENMP
npe = omp_get_max_threads();
/* parallelisation is over number of shot positions (nshots) */
if (npe > nshots) {
vmess("Number of OpenMP threads set to %d (was %d)", nshots, npe);
omp_set_num_threads(nshots);
}
#endif
t0 = wallclock_time();
/* reset output data to zero */
memset(&iRN[0], 0, Nfoc*nxs*nts*sizeof(float));
ctrace = (complex *)calloc(ntfft,sizeof(complex));
/* this first check is done to support an acquisition geometry that has more receiver than source
* postions. In the first iteration the int R(x_r,x_s) Fop(x_r) d x_r results in a grid on x_s.
* so for the next interations onlt x_s traces have to be computed on Fop */
if (!first) {
/* transform muted Ni (Top) to frequency domain, input for next iteration */
for (l = 0; l < Nfoc; l++) {
/* set Fop to zero, so new operator can be defined within ixpos points */
memset(&Fop[l*nxs*nw].r, 0, nxs*nw*2*sizeof(float));
for (i = 0; i < npos; i++) {
rc1fft(&Top[l*size+i*nts],ctrace,ntfft,-1);
ix = ixpos[i];
for (iw=0; iw<nw; iw++) {
Fop[l*nxs*nw+iw*nxs+ix].r = ctrace[nw_low+iw].r;
Fop[l*nxs*nw+iw*nxs+ix].i = mode*ctrace[nw_low+iw].i;
}
}
}
}
else { /* only for first call to synthesis using all nxs traces in G_d */
/* transform G_d to frequency domain, over all nxs traces */
first=0;
for (l = 0; l < Nfoc; l++) {
/* set Fop to zero, so new operator can be defined within all ix points */
memset(&Fop[l*nxs*nw].r, 0, nxs*nw*2*sizeof(float));
for (i = 0; i < nxs; i++) {
rc1fft(&Top[l*size+i*nts],ctrace,ntfft,-1);
for (iw=0; iw<nw; iw++) {
Fop[l*nxs*nw+iw*nxs+i].r = ctrace[nw_low+iw].r;
Fop[l*nxs*nw+iw*nxs+i].i = mode*ctrace[nw_low+iw].i;
}
}
}
idxs = 1.0/dxs;
ixrcv = (int *)malloc(nshots*nx*sizeof(int));
for (k=0; k<nshots; k++) {
for (i = 0; i < nx; i++) {
ixrcv[k*nx+i] = NINT((xrcv[k*nx+i]-fxsb)*idxs);
}
}
}
free(ctrace);
t1 = wallclock_time();
*tfft += t1 - t0;
if (reci == 0 || reci == 1) {
/*================ SYNTHESIS ================*/
#pragma omp parallel default(none) \
shared(iRN, dx, npe, nw, verbose, nshots, xnx) \
shared(Refl, Nfoc, reci, xsrc, xsyn, fxsb, fxse, nxs, dxs) \
shared(nx, dxsrc, nfreq, nw_low, nw_high) \
shared(Fop, size, nts, ntfft, scl, ixrcv) \
private(l, ix, j, m, i, sum, rtrace, k, ixsrc, inx)
{ /* start of parallel region */
sum = (complex *)malloc(nfreq*sizeof(complex));
rtrace = (float *)calloc(ntfft,sizeof(float));
/* Loop over total number of shots */
#pragma omp for schedule(guided,1)
for (k=0; k<nshots; k++) {
if ((xsrc[k] < 0.999*fxsb) || (xsrc[k] > 1.001*fxse)) continue;
ixsrc = NINT((xsrc[k] - fxsb)/dxs);
inx = xnx[k]; /* number of traces per shot */
for (l = 0; l < Nfoc; l++) {
/* compute integral over receiver positions */
/* multiply R with Fop and sum over nx */
memset(&sum[0].r,0,nfreq*2*sizeof(float));
for (j = nw_low, m = 0; j <= nw_high; j++, m++) {
for (i = 0; i < inx; i++) {
ix = ixrcv[k*nx+i];
sum[j].r += Refl[k*nw*nx+m*nx+i].r*Fop[l*nw*nxs+m*nxs+ix].r -
Refl[k*nw*nx+m*nx+i].i*Fop[l*nw*nxs+m*nxs+ix].i;
sum[j].i += Refl[k*nw*nx+m*nx+i].i*Fop[l*nw*nxs+m*nxs+ix].r +
Refl[k*nw*nx+m*nx+i].r*Fop[l*nw*nxs+m*nxs+ix].i;
}
}
/* transfrom result back to time domain */
cr1fft(sum, rtrace, ntfft, 1);
/* place result at source position ixsrc; dx = receiver distance */
for (j = 0; j < nts; j++)
iRN[l*size+ixsrc*nts+j] += rtrace[j]*scl*dx;
} /* end of parallel Nfoc loop */
if (verbose>4) vmess("*** Shot gather %d processed ***", k);
} /* end of nshots (k) loop */
free(sum);
free(rtrace);
} /* end of parallel region */
} /* end of if reci */
/* if reciprocal traces are enabled start a new loop over reciprocal shot positions */
if (reci != 0) {
#pragma omp parallel default(none) \
shared(iRN, dx, nw, verbose) \
shared(Refl, Nfoc, reci, xsrc, xsyn, fxsb, fxse, nxs, dxs) \
shared(nx, dxsrc, nfreq, nw_low, nw_high) \
shared(reci_xrcv, reci_xsrc, ixmask, isxcount) \
shared(Fop, size, nts, ntfft, scl, ixrcv) \
private(l, ix, j, m, i, k, sum, rtrace, ik, il, ixsrc, inx)
{ /* start of parallel region */
sum = (complex *)malloc(nfreq*sizeof(complex));
rtrace = (float *)calloc(ntfft,sizeof(float));
#pragma omp for schedule(guided,1)
for (k=0; k<nxs; k++) {
if (isxcount[k] == 0) continue;
ixsrc = k;
inx = isxcount[ixsrc]; /* number of traces per reciprocal shot */
for (l = 0; l < Nfoc; l++) {
/* compute integral over (reciprocal) source positions */
/* multiply R with Fop and sum over nx */
memset(&sum[0].r,0,nfreq*2*sizeof(float));
for (j = nw_low, m = 0; j <= nw_high; j++, m++) {
for (i = 0; i < inx; i++) {
il = reci_xrcv[ixsrc*nxs+i];
ik = reci_xsrc[ixsrc*nxs+i];
ix = NINT((xsrc[il] - fxsb)/dxs);
sum[j].r += Refl[il*nw*nx+m*nx+ik].r*Fop[l*nw*nxs+m*nxs+ix].r -
Refl[il*nw*nx+m*nx+ik].i*Fop[l*nw*nxs+m*nxs+ix].i;
sum[j].i += Refl[il*nw*nx+m*nx+ik].i*Fop[l*nw*nxs+m*nxs+ix].r +
Refl[il*nw*nx+m*nx+ik].r*Fop[l*nw*nxs+m*nxs+ix].i;
}
}
/* transfrom result back to time domain */
cr1fft(sum, rtrace, ntfft, 1);
/* place result at source position ixsrc; dxsrc = shot distance */
for (j = 0; j < nts; j++)
iRN[l*size+ixsrc*nts+j] = ixmask[ixsrc]*(iRN[l*size+ixsrc*nts+j]+rtrace[j]*scl*dxsrc);
} /* end of Nfoc loop */
} /* end of parallel reciprocal shots (k) loop */
free(sum);
free(rtrace);
} /* end of parallel region */
} /* end of if reci */
t = wallclock_time() - t0;
if (verbose>2) {
vmess("OMP: parallel region = %f seconds (%d threads)", t, npe);
}
return;
}
double wallclock_time_(void)
{
return (double)wallclock_time();
}
int main(int argc, char **argv)
{
modPar mod;
recPar rec;
snaPar sna;
wavPar wav;
srcPar src;
bndPar bnd;
shotPar shot;
float **src_nwav;
float *rox, *roz, *l2m, *lam, *mul;
float *tss, *tes, *tep, *p, *q, *r;
float *vx, *vz, *tzz, *txz, *txx;
float *rec_vx, *rec_vz, *rec_p;
float *rec_txx, *rec_tzz, *rec_txz;
float *rec_pp, *rec_ss;
float *rec_udp, *rec_udvz;
float *beam_vx, *beam_vz, *beam_p;
float *beam_txx, *beam_tzz, *beam_txz;
float *beam_pp, *beam_ss;
float sinkvel;
double t0, t1, t2, t3, tt, tinit;
size_t size, sizem, nsamp, memsize;
int n1, ix, iz, ir, ishot, i;
int ioPx, ioPz;
int it0, it1, its, it, fileno, isam;
int ixsrc, izsrc;
int verbose;
t0= wallclock_time();
initargs(argc,argv);
requestdoc(0);
if (!getparint("verbose",&verbose)) verbose=0;
getParameters(&mod, &rec, &sna, &wav, &src, &shot, &bnd, verbose);
/* allocate arrays for model parameters: the different schemes use different arrays */
n1 = mod.naz;
sizem=mod.nax*mod.naz;
rox = (float *)calloc(sizem,sizeof(float));
roz = (float *)calloc(sizem,sizeof(float));
l2m = (float *)calloc(sizem,sizeof(float));
if (mod.ischeme==2) {
tss = (float *)calloc(sizem,sizeof(float));
tep = (float *)calloc(sizem,sizeof(float));
q = (float *)calloc(sizem,sizeof(float));
}
if (mod.ischeme>2) {
lam = (float *)calloc(sizem,sizeof(float));
mul = (float *)calloc(sizem,sizeof(float));
}
if (mod.ischeme==4) {
tss = (float *)calloc(sizem,sizeof(float));
tes = (float *)calloc(sizem,sizeof(float));
tep = (float *)calloc(sizem,sizeof(float));
r = (float *)calloc(sizem,sizeof(float));
p = (float *)calloc(sizem,sizeof(float));
q = (float *)calloc(sizem,sizeof(float));
}
/* read velocity and density files */
readModel(mod, bnd, rox, roz, l2m, lam, mul, tss, tes, tep);
/* read and/or define source wavelet(s) */
/* Using a random source, which can have a random length
for each source position, a pointer array with variable
length (wav.nsamp[i]) is used.
The total length of all the source lengths together is wav.nst */
if (wav.random) {
src_nwav = (float **)calloc(wav.nx,sizeof(float *));
src_nwav[0] = (float *)calloc(wav.nst,sizeof(float));
assert(src_nwav[0] != NULL);
nsamp = 0;
for (i=0; i<wav.nx; i++) {
src_nwav[i] = (float *)(src_nwav[0] + nsamp);
nsamp += wav.nsamp[i];
}
}
else {
src_nwav = (float **)calloc(wav.nx,sizeof(float *));
src_nwav[0] = (float *)calloc(wav.nt*wav.nx,sizeof(float));
assert(src_nwav[0] != NULL);
for (i=0; i<wav.nx; i++) {
src_nwav[i] = (float *)(src_nwav[0] + wav.nt*i);
}
}
defineSource(wav, src, src_nwav, mod.grid_dir, verbose);
/* allocate arrays for wavefield and receiver arrays */
vx = (float *)calloc(sizem,sizeof(float));
vz = (float *)calloc(sizem,sizeof(float));
tzz = (float *)calloc(sizem,sizeof(float)); /* =P field for acoustic */
if (mod.ischeme>2) {
txz = (float *)calloc(sizem,sizeof(float));
txx = (float *)calloc(sizem,sizeof(float));
}
if (rec.type.vz) rec_vz = (float *)calloc(size,sizeof(float));
size = rec.n*rec.nt;
if (rec.type.vz) rec_vz = (float *)calloc(size,sizeof(float));
if (rec.type.vx) rec_vx = (float *)calloc(size,sizeof(float));
if (rec.type.p) rec_p = (float *)calloc(size,sizeof(float));
if (rec.type.txx) rec_txx = (float *)calloc(size,sizeof(float));
if (rec.type.tzz) rec_tzz = (float *)calloc(size,sizeof(float));
if (rec.type.txz) rec_txz = (float *)calloc(size,sizeof(float));
if (rec.type.pp) rec_pp = (float *)calloc(size,sizeof(float));
if (rec.type.ss) rec_ss = (float *)calloc(size,sizeof(float));
if (rec.type.ud) { /* get velcity and density at first receiver location */
ir = mod.ioZz + rec.z[0]+(rec.x[0]+mod.ioZx)*n1;
rec.rho = mod.dt/(mod.dx*roz[ir]);
rec.cp = sqrt(l2m[ir]*(roz[ir]))*mod.dx/mod.dt;
rec_udvz = (float *)calloc(mod.nax*rec.nt,sizeof(float));
rec_udp = (float *)calloc(mod.nax*rec.nt,sizeof(float));
}
if(sna.beam) {
size = sna.nz*sna.nx;
if (sna.type.vz) beam_vz = (float *)calloc(size,sizeof(float));
if (sna.type.vx) beam_vx = (float *)calloc(size,sizeof(float));
if (sna.type.p) beam_p = (float *)calloc(size,sizeof(float));
if (sna.type.txx) beam_txx = (float *)calloc(size,sizeof(float));
if (sna.type.tzz) beam_tzz = (float *)calloc(size,sizeof(float));
if (sna.type.txz) beam_txz = (float *)calloc(size,sizeof(float));
if (sna.type.pp) beam_pp = (float *)calloc(size,sizeof(float));
if (sna.type.ss) beam_ss = (float *)calloc(size,sizeof(float));
}
t1= wallclock_time();
if (verbose) {
tinit = t1-t0;
vmess("*******************************************");
vmess("************* runtime info ****************");
vmess("*******************************************");
vmess("CPU time for intializing arrays and model = %f", tinit);
}
/* Sinking source and receiver arrays:
If P-velocity==0 the source and receiver
postions are placed deeper until the P-velocity changes.
The free-surface position is stored in bnd.surface[ix].
Setting the option rec.sinkvel only sinks the receiver position
(not the source) and uses the velocity
of the first receiver to sink through to the next layer. */
ioPx=mod.ioPx;
ioPz=mod.ioPz;
if (bnd.lef==4 || bnd.lef==2) ioPx += bnd.ntap;
if (bnd.top==4 || bnd.top==2) ioPz += bnd.ntap;
if (rec.sinkvel) sinkvel=l2m[(rec.x[0]+ioPx)*n1+rec.z[0]+ioPz];
else sinkvel = 0.0;
/* sink receivers to value different than sinkvel */
for (ir=0; ir<rec.n; ir++) {
iz = rec.z[ir];
ix = rec.x[ir];
while(l2m[(ix+ioPx)*n1+iz+ioPz] == sinkvel) iz++;
rec.z[ir]=iz+rec.sinkdepth;
rec.zr[ir]=rec.zr[ir]+(rec.z[ir]-iz)*mod.dz;
// rec.zr[ir]=rec.z[ir]*mod.dz;
if (verbose>3) vmess("receiver position %d at grid[ix=%d, iz=%d] = (x=%f z=%f)", ir, ix+ioPx, rec.z[ir]+ioPz, rec.xr[ir]+mod.x0, rec.zr[ir]+mod.z0);
}
/* sink sources to value different than zero */
for (ishot=0; ishot<shot.n; ishot++) {
iz = shot.z[ishot];
ix = shot.x[ishot];
while(l2m[(ix+ioPx)*n1+iz+ioPz] == 0.0) iz++;
shot.z[ishot]=iz+src.sinkdepth;
}
/* scan for free surface boundary in case it has a topography */
for (ix=0; ix<mod.nx; ix++) {
iz = ioPz;
while(l2m[(ix+ioPx)*n1+iz] == 0.0) iz++;
bnd.surface[ix+ioPx] = iz;
if ((verbose>3) && (iz != ioPz)) vmess("Topgraphy surface x=%.2f z=%.2f", mod.x0+mod.dx*ix, mod.z0+mod.dz*(iz-ioPz));
}
for (ix=0; ix<ioPx; ix++) {
bnd.surface[ix] = bnd.surface[ioPx];
}
for (ix=ioPx+mod.nx; ix<mod.iePx; ix++) {
bnd.surface[ix] = bnd.surface[mod.iePx-1];
}
if (verbose>3) writeSrcRecPos(&mod, &rec, &src, &shot);
/* Outer loop over number of shots */
for (ishot=0; ishot<shot.n; ishot++) {
izsrc = shot.z[ishot];
ixsrc = shot.x[ishot];
fileno= 0;
memset(vx,0,sizem*sizeof(float));
memset(vz,0,sizem*sizeof(float));
memset(tzz,0,sizem*sizeof(float));
if (mod.ischeme==2) {
memset(q,0,sizem*sizeof(float));
}
if (mod.ischeme>2) {
memset(txz,0,sizem*sizeof(float));
memset(txx,0,sizem*sizeof(float));
}
if (mod.ischeme==4) {
memset(r,0,sizem*sizeof(float));
memset(p,0,sizem*sizeof(float));
memset(q,0,sizem*sizeof(float));
}
if (verbose) {
if (!src.random) {
vmess("Modeling source %d at gridpoints ix=%d iz=%d", ishot, shot.x[ishot], shot.z[ishot]);
vmess(" which are actual positions x=%.2f z=%.2f", mod.x0+mod.dx*shot.x[ishot], mod.z0+mod.dz*shot.z[ishot]);
}
vmess("Receivers at gridpoint x-range ix=%d - %d", rec.x[0], rec.x[rec.n-1]);
vmess(" which are actual positions x=%.2f - %.2f", mod.x0+rec.xr[0], mod.x0+rec.xr[rec.n-1]);
vmess("Receivers at gridpoint z-range iz=%d - %d", rec.z[0], rec.z[rec.n-1]);
vmess(" which are actual positions z=%.2f - %.2f", mod.z0+rec.zr[0], mod.z0+rec.zr[rec.n-1]);
}
if (mod.grid_dir) { /* reverse time modeling */
it0=-mod.nt+1;
it1=0;
its=-1;
it0=0;
it1=mod.nt;
its=1;
}
else {
it0=0;
it1=mod.nt;
its=1;
}
/* Main loop over the number of time steps */
for (it=it0; it<it1; it++) {
#pragma omp parallel default (shared) \
shared (rox, roz, l2m, lam, mul, txx, txz, tzz, vx, vz) \
shared (tss, tep, tes, r, q, p) \
shared (tinit, it0, it1, its) \
shared(beam_vx, beam_vz, beam_txx, beam_tzz, beam_txz, beam_p, beam_pp, beam_ss) \
shared(rec_vx, rec_vz, rec_txx, rec_tzz, rec_txz, rec_p, rec_pp, rec_ss) \
private (tt, t2, t3) \
shared (shot, bnd, mod, src, wav, rec, ixsrc, izsrc, it, src_nwav, verbose)
{
switch ( mod.ischeme ) {
case 1 : /* Acoustic FD kernel */
if (mod.iorder==2) {
acoustic2(mod, src, wav, bnd, it, ixsrc, izsrc, src_nwav,
vx, vz, tzz, rox, roz, l2m, verbose);
}
else if (mod.iorder==4) {
if (mod.sh) {
acousticSH4(mod, src, wav, bnd, it, ixsrc, izsrc, src_nwav,
vx, vz, tzz, rox, roz, l2m, verbose);
}
else {
acoustic4(mod, src, wav, bnd, it, ixsrc, izsrc, src_nwav,
vx, vz, tzz, rox, roz, l2m, verbose);
}
}
else if (mod.iorder==6) {
acoustic6(mod, src, wav, bnd, it, ixsrc, izsrc, src_nwav,
vx, vz, tzz, rox, roz, l2m, verbose);
}
break;
case 2 : /* Visco-Acoustic FD kernel */
viscoacoustic4(mod, src, wav, bnd, it, ixsrc, izsrc, src_nwav,
vx, vz, tzz, rox, roz, l2m, tss, tep, q, verbose);
break;
case 3 : /* Elastic FD kernel */
if (mod.iorder==4) {
elastic4(mod, src, wav, bnd, it, ixsrc, izsrc, src_nwav,
vx, vz, tzz, txx, txz, rox, roz, l2m, lam, mul, verbose);
}
else if (mod.iorder==6) {
elastic6(mod, src, wav, bnd, it, ixsrc, izsrc, src_nwav,
vx, vz, tzz, txx, txz, rox, roz, l2m, lam, mul, verbose);
}
break;
case 4 : /* Visco-Elastic FD kernel */
viscoelastic4(mod, src, wav, bnd, it, ixsrc, izsrc, src_nwav,
vx, vz, tzz, txx, txz, rox, roz, l2m, lam, mul,
tss, tep, tes, r, q, p, verbose);
break;
}
/* write samples to file if rec.nt samples are calculated */
#pragma omp master
{
if ( (((it-rec.delay) % rec.skipdt)==0) && (it >= rec.delay) ) {
int writeToFile, itwritten;
writeToFile = ! ( (((it-rec.delay)/rec.skipdt)+1)%rec.nt );
itwritten = fileno*(rec.nt)*rec.skipdt;
isam = (it-rec.delay-itwritten)/rec.skipdt;
/* store time at receiver positions */
getRecTimes(mod, rec, bnd, it, isam, vx, vz, tzz, txx, txz,
rec_vx, rec_vz, rec_txx, rec_tzz, rec_txz,
rec_p, rec_pp, rec_ss, rec_udp, rec_udvz, verbose);
/* at the end of modeling a shot, write receiver array to output file(s) */
if (writeToFile && (it+rec.skipdt <= it1-1) ) {
fileno = ( ((it-rec.delay)/rec.skipdt)+1)/rec.nt;
writeRec(rec, mod, bnd, wav, ixsrc, izsrc, isam+1, ishot, fileno,
rec_vx, rec_vz, rec_txx, rec_tzz, rec_txz,
rec_p, rec_pp, rec_ss, rec_udp, rec_udvz, verbose);
}
}
/* write snapshots to output file(s) */
if (sna.nsnap) {
writeSnapTimes(mod, sna, ixsrc, izsrc, it, vx, vz, tzz, txx, txz, verbose);
}
/* calculate beams */
if(sna.beam) {
getBeamTimes(mod, sna, vx, vz, tzz, txx, txz,
beam_vx, beam_vz, beam_txx, beam_tzz, beam_txz,
beam_p, beam_pp, beam_ss, verbose);
}
}
/* taper the edges of the model */
// taperEdges(mod, bnd, vx, vz, verbose);
#pragma omp master
{
if (verbose) {
if (it==it0+100*its) t2=wallclock_time();
if (it==(it0+500*its)) {
t3=wallclock_time();
tt=(t3-t2)*(((it1-it0)*its)/400.0);
vmess("Estimated compute time = %.2f s. per shot.",tt);
vmess("Estimated total compute time = %.2f s.",tinit+shot.n*tt);
}
}
}
} /* end of OpenMP parallel section */
} /* end of loop over time steps it */
/* write output files: receivers and or beams */
if (fileno) fileno++;
if (rec.scale==1) { /* scale receiver with distance src-rcv */
float xsrc, zsrc, Rrec, rdx, rdz;
int irec;
xsrc=mod.x0+mod.dx*ixsrc;
zsrc=mod.z0+mod.dz*izsrc;
for (irec=0; irec<rec.n; irec++) {
rdx=mod.x0+rec.xr[irec]-xsrc;
rdz=mod.z0+rec.zr[irec]-zsrc;
Rrec = sqrt(rdx*rdx+rdz*rdz);
fprintf(stderr,"Rec %d is scaled with distance %f R=%.2f,%.2f S=%.2f,%.2f\n", irec, Rrec,rdx,rdz,xsrc,zsrc);
for (it=0; it<rec.nt; it++) {
rec_p[irec*rec.nt+it] *= sqrt(Rrec);
}
}
}
writeRec(rec, mod, bnd, wav, ixsrc, izsrc, isam+1, ishot, fileno,
rec_vx, rec_vz, rec_txx, rec_tzz, rec_txz,
rec_p, rec_pp, rec_ss, rec_udp, rec_udvz, verbose);
writeBeams(mod, sna, ixsrc, izsrc, ishot, fileno,
beam_vx, beam_vz, beam_txx, beam_tzz, beam_txz,
beam_p, beam_pp, beam_ss, verbose);
} /* end of loop over number of shots */
t1= wallclock_time();
if (verbose) {
vmess("Total compute time FD modelling = %.2f s.", t1-t0);
}
/* free arrays */
free(rox);
free(roz);
free(l2m);
free(src_nwav[0]);
free(src_nwav);
free(vx);
free(vz);
free(tzz);
if (rec.type.vz) free(rec_vz);
if (rec.type.vx) free(rec_vx);
if (rec.type.p) free(rec_p);
if (rec.type.txx) free(rec_txx);
if (rec.type.tzz) free(rec_tzz);
if (rec.type.txz) free(rec_txz);
if (rec.type.pp) free(rec_pp);
if (rec.type.ss) free(rec_ss);
if (rec.type.ud) {
free(rec_udvz);
free(rec_udp);
}
if(sna.beam) {
if (sna.type.vz) free(beam_vz);
if (sna.type.vx) free(beam_vx);
if (sna.type.p) free(beam_p);
if (sna.type.txx) free(beam_txx);
if (sna.type.tzz) free(beam_tzz);
if (sna.type.txz) free(beam_txz);
if (sna.type.pp) free(beam_pp);
if (sna.type.ss) free(beam_ss);
}
if (mod.ischeme==2) {
free(tss);
free(tep);
free(q);
}
if (mod.ischeme>2) {
free(lam);
free(mul);
free(txz);
free(txx);
}
if (mod.ischeme==4) {
free(tss);
free(tes);
free(tep);
free(r);
free(p);
free(q);
}
return 0;
}
int main(int argc, char *argv[])
{
FILE *shot_file, *vel_file, *out_file, *beam_file;
size_t nread, bytes, size, trace_sz, size_out;
int verbose, method, ntraces, verb_root;
int nxv, nyv, nzv, binary_file, dstep, id, id1, id2;
int d, nt, ndepth, i, j, conjg, conjgs, mode, out_su;
int ntap, tap_opt, order, McC, oplx, oply, fine, MB;
int stackmigr, imc, area, ixmin, ixmax, iymin, iymax, ns;
int nfft, nfreq, nw_high, nw_low, nw, sx, sy, ix, iy;
int npages_w, sxy, iw, one_shot, traces_shot, sign, is;
int traces_read_in, nxy, fd, nx, ny, num_threads, nel, oper_opt;
int fldr_w, power_of_2, beam_su, nterms, filter_inc, beam;
Area shot_area;
float alpha, weight, scl, sclw;
float fmin, fmax, dt;
float *tot_beam, *beams, scale;
float *velocity, weights, tshift, zrcv;
float xvmin, yvmin, zvmin, dxv, dyv, dzv, vmin, vmax;
float dw, df, om, dtw, tdw, t_w, tr, ti;
double t0, t1, t2, t3, t_migr=0, t_io=0, t_table=0, t_init=0;
double t_comm=0;
complex *rec_all, *rec_field, *rec;
char *file_vel, *file_in, *file_out, *file_beam, *file_table;
char *tmp_dir, sys_call[256];
segy *hdr;
int npes, pe, root_pe=0, nlw, maxlw, *freq_index, fdist, ipe;
#ifdef MPI
int *nlwcounts, *recvcounts, *displacements;
int nlw_tag;
complex *gath_rec_field;
MPI_Status status;
MPI_Request request;
MPI_Init( &argc, &argv );
MPI_Comm_size( MPI_COMM_WORLD, &npes );
MPI_Comm_rank( MPI_COMM_WORLD, &pe );
#else
npes = 1;
pe = 0;
#endif
t0 = wallclock_time();
/* Read in parameters */
initargs(argc,argv);
requestdoc(0);
if(!getparstring("file_in", &file_in)) file_in=NULL;
if(!getparstring("file_vel", &file_vel)) file_vel=NULL;
if(!getparstring("file_out", &file_out)) file_out=NULL;
if(!getparstring("file_beam", &file_beam)) file_beam=" ";
if(!getparstring("file_table", &file_table)) file_table=NULL;
if(!getparstring("tmp_dir", &tmp_dir)) tmp_dir="/tmp";
if(!getparfloat("fmin", &fmin)) fmin = 0.0;
if(!getparfloat("fmax", &fmax)) fmax = 45.0;
if(!getparint("mode", &mode)) mode = 1;
if(!getparint("ntap", &ntap)) ntap = 0;
if(!getparint("tap_opt", &tap_opt)) tap_opt = 1;
if(!getparint("method", &method)) method = 1;
if(!getparint("conjg", &conjg)) conjg = 0;
if(!getparint("conjgs", &conjgs)) conjgs = 0;
if(!getparint("area", &area)) area = 0;
if(!getparint("oplx", &oplx)) oplx = 25;
if(!getparint("oply", &oply)) oply = oplx;
if(!getparint("order", &order)) order = 13;
if(!getparint("McC", &McC)) McC = 1;
if(!getparint("oper_opt", &oper_opt)) oper_opt = 1;
if(!getparint("nterms", &nterms)) nterms = 1;
if(!getparint("filter_inc", &filter_inc)) filter_inc = 1;
if(!getparfloat("alpha", &alpha)) alpha = 65.0;
if(!getparfloat("weight", &weight)) weight = 5e-5;
if(!getparfloat("weights", &weights)) weights = 1e-2;
if(!getparint("fine", &fine)) fine = 2;
if(!getparint("beam", &beam)) beam = 0;
if(!getparint("verbose", &verbose)) verbose = 0;
if(!ISODD(oplx)) oplx += 1;
if(!ISODD(oply)) oply += 1;
if(conjg) conjg = -1; else conjg = 1;
if(conjgs) conjgs = -1; else conjgs = 1;
if(mode >= 0) mode = 1;
if(mode < 0) mode = -1;
assert(McC <= 2 && McC >= 1);
assert(method <= 4 && method >= 1);
assert(file_vel != NULL);
assert(file_out != NULL);
out_su = (strstr(file_out, ".su")!=NULL);
beam_su = (strstr(file_beam, ".su")!=NULL);
if (verbose && pe==root_pe) verb_root=verbose;
else verb_root = 0;
t1 = wallclock_time(); t_init += t1-t0;
/* Clean up 'old' velocity files */
sprintf(sys_call,"rm -rf %s/velocity*.bin\n",tmp_dir);
system(sys_call);
#ifdef MPI
MPI_Barrier(MPI_COMM_WORLD);
#endif
/* Open velocity file and determine the size of the file */
openVelocityFile(file_vel, &vel_file, &shot_area, verb_root);
#ifdef MPI
MPI_Barrier(MPI_COMM_WORLD);
#endif
xvmin = shot_area.xmin;
yvmin = shot_area.ymin;
zvmin = shot_area.zmin;
nxv = shot_area.nx;
nyv = shot_area.ny;
nzv = shot_area.nz;
dxv = shot_area.dx;
dyv = shot_area.dy;
dzv = shot_area.dz;
nxy = shot_area.sxy;
if(!getparfloat("zrcv", &zrcv)) zrcv = zvmin+(nzv-1)*dzv;
ndepth = NINT((zrcv-zvmin)/dzv);
if(!getparint("dstep", &dstep)) dstep = MIN(5, ndepth);
if(!getparfloat("vmin", &vmin)) vmin = 1500;
if(!getparfloat("vmax", &vmax)) vmax = 4800;
/* Open file_in file and read first header */
hdr = (segy *)calloc(1,sizeof(segy));
if (file_in == NULL) shot_file = stdin;
else shot_file = fopen( file_in, "r" );
assert( shot_file );
nread = fread( hdr, 1, TRCBYTES, shot_file );
assert (nread == TRCBYTES);
fseek ( shot_file, 0, SEEK_END );
bytes = ftell(shot_file);
nt = hdr[0].ns;
dt = 1e-6*hdr[0].dt;
trace_sz = sizeof(float)*nt+TRCBYTES;
ntraces = (int) (bytes/trace_sz);
nfft = optncr(nt);
nfreq = nfft/2 + 1;
df = 1.0/(nfft*dt);
dw = 2.*M_PI*df;
nw_high = MIN( (int)(fmax/df), nfreq );
nw_low = MAX( (int)((fmin)/df), 1 );
nw = nw_high - nw_low + 1;
sx = hdr[0].sx;
sy = hdr[0].sy;
if (hdr[0].scalco < 0) scl = 1.0/fabs(hdr[0].scalco);
else if (hdr[0].scalco == 0) scl = 1.0;
else scl = hdr[0].scalco;
t2 = wallclock_time(); t_io += t2-t1;
/*======= compute frequency distribution for multiple CPU's ========*/
fdist = 0;
maxlw = ceil((float)nw/(float)npes);
freq_index = (int *)malloc(maxlw*sizeof(int));
nlw = frequency_distribution(nw_low, nw, npes, pe, maxlw,
freq_index, fdist );
#ifdef MPI
if( verbose ) {
/* print out all the frequencies for each process */
MPI_Barrier(MPI_COMM_WORLD);
for( ipe=0; ipe<npes; ipe++ ) {
if( pe == ipe ) {
fprintf(stderr, "pe=%d:\tf[%d] = df*{", pe, nlw );
if( nlw > 0 )
for( iw=0; iw<nlw; iw++ )
fprintf(stderr, " %d", freq_index[iw] );
fprintf( stderr, " }\n" );
fflush(stderr);
}
MPI_Barrier(MPI_COMM_WORLD);
}
}
if (npes == 1) nlw = nw;
nlwcounts = (int *)malloc(npes*sizeof(int));
recvcounts = (int *)malloc(npes*sizeof(int));
displacements = (int *)malloc(npes*sizeof(int));
nlw_tag = 1;
if (pe == root_pe) {
displacements[0] = 0;
nlwcounts[0] = nlw;
for( ipe=1; ipe<npes; ipe++ ) {
MPI_Recv(&nlwcounts[ipe], 1, MPI_INT, ipe, nlw_tag,
MPI_COMM_WORLD, &status);
displacements[ipe] = displacements[ipe-1]+nlwcounts[ipe-1];
}
}
else {
MPI_Send(&nlw, 1, MPI_INT, root_pe, nlw_tag, MPI_COMM_WORLD);
}
#else
nlw = nw;
#endif
fmin = MAX(0,-df + df*freq_index[0]);
fmax = df + df*freq_index[nlw-1];
assert( fmax < 1.0/(2.0*dt) ); /* Nyguist in time */
assert( (2.0*fmax)/vmin < 1.0/dxv ); /* Nyguist in space */
assert( (2.0*fmax)/vmin < 1.0/dyv ); /* Nyguist in space */
size = (size_t)nlw*nxy*sizeof(complex);
if (verb_root) {
fprintf(stderr," minimum velocity = %.2f\n", vmin);
fprintf(stderr," maximum velocity = %.2f\n", vmax);
MB = 1024*1024;
fprintf(stderr,"\n DATA INFORMATION\n");
fprintf(stderr," nw = %d\n", nw);
fprintf(stderr," fmin = %.3f fmax = %.3f\n", nw_low*df, nw_high*df);
fprintf(stderr," dt = %.4f nt = %d nfft = %d\n", dt, nt, nfft);
fprintf(stderr," size of rec_field = %ld Mbytes\n", size/MB);
size_out = (size_t)(nt)*(size_t)(nxy)*sizeof(float);
if (out_su) size_out += (TRCBYTES*nxy);
fprintf(stderr," size of output file = %ld Mbytes\n", size_out/MB);
}
/* Open beam file if beam == 1 */
if (beam && pe == root_pe) {
beam_file = fopen( file_beam, "w+" );
assert( beam_file );
}
t1 = wallclock_time(); t_init += t1-t2;
/* Calculate operator tables */
if (method == 1) {
tablecalc_2D(oplx, oply, nxv, dxv, nyv, dyv, dzv, alpha, fmin, fmax,
vmin, vmax, df, weight, fine, oper_opt, file_table, verb_root);
}
else if (method == 2) {
tablecalc_1D(order, nxv, dxv, dzv, alpha, fmin, fmax, vmin, vmax, df,
fine, oper_opt, verb_root);
}
t2 = wallclock_time(); t_table = t2-t1;
if (verb_root)
fprintf(stderr," time to calculate tables : %.3f s.\n", t_table);
/* ============ INITIALIZE AND CHECK PARAMETERS =============== */
/* allocate rec field to zero and distribute along machine */
rec_field = (complex *)calloc(nlw*nxy, sizeof(complex));
assert(rec_field != NULL);
velocity = (float *)malloc(dstep*nxy*sizeof(float));
assert(velocity != NULL);
if (beam) {
beams = (float *)calloc(dstep*nxy, sizeof(float));
assert(beams != NULL);
#ifdef MPI
tot_beam = (float *)calloc(dstep*nxy, sizeof(float));
assert(tot_beam != NULL);
#endif
}
#ifdef MPI
if (pe == root_pe) {
gath_rec_field = (complex *)calloc(nxy*nw,sizeof(complex));
assert(gath_rec_field != NULL);
}
#endif
one_shot = 1;
traces_shot = 0;
traces_read_in = 0;
ixmin = nxv-1; ixmax = 0;
iymin = nxv-1; iymax = 0;
t1 = wallclock_time(); t_init += t1-t2;
fseek(shot_file, 0, SEEK_SET);
read_FFT_DataFile(shot_file, rec_field, shot_area, nfft, nlw,
freq_index[0], &traces_read_in, &traces_shot,
&ixmin, &ixmax, &iymin, &iymax, &sx, &sy, conjgs, verb_root);
t2 = wallclock_time(); t_io += t2-t1;
if (verb_root)
fprintf(stderr," time to initialize migration : %.3f s.\n", t1-t0);
/* Loop over input traces */
is = 0;
while (one_shot) {
t1 = wallclock_time();
if (verb_root) {
fprintf(stderr,"\n EXTRAPOLATION INFORMATION\n");
fprintf(stderr," source position (x,y) : %.2f, %.2f\n",
sx*scl, sy*scl);
fprintf(stderr," number of traces in shot : %d\n", traces_shot);
fprintf(stderr," traces done = %d to do %d\n",
traces_read_in-traces_shot, ntraces-traces_read_in+traces_shot);
fprintf(stderr," shot region is x:%d-%d y:%d-%d\n",
ixmin, ixmax, iymin, iymax);
}
/* determine aperture to be extrapolated */
if (area>0) {
ixmin = MAX(0,ixmin-area);
ixmax = MIN(nxv-1,ixmax+area);
iymin = MAX(0,iymin-area);
iymax = MIN(nyv-1,iymax+area);
}
else {
ixmin = 0;
iymin = 0;
ixmax = nxv-1;
iymax = nyv-1;
}
nx = ixmax-ixmin+1;
ny = iymax-iymin+1;
shot_area.ixmin = ixmin;
shot_area.ixmax = ixmax;
shot_area.iymin = iymin;
shot_area.iymax = iymax;
shot_area.dx = dxv;
shot_area.dy = dyv;
shot_area.dz = dzv;
shot_area.nx = nxv;
shot_area.ny = nyv;
shot_area.sxy = nxy;
if (verb_root) {
fprintf(stderr," work area is x:%d-%d (%d) y:%d-%d (%d)\n",
ixmin, ixmax, nx, iymin, iymax, ny);
}
t2 = wallclock_time(); t_init += t2-t1;
/* write beam for depth=0 */
if (beam) {
scale = 1.0/(float)(nw);
for (iw=0; iw<nlw; iw++) {
rec = (complex*) (rec_field + iw*nxy);
for (ix = 0; ix < nxy; ix++) {
beams[ix] += sqrt(rec[ix].r*rec[ix].r+rec[ix].i*rec[ix].i)*scale;
}
}
t1 = wallclock_time(); t_init += t1-t2;
#ifdef MPI
MPI_Reduce(beams, tot_beam, nxy, MPI_FLOAT, MPI_SUM, 0, MPI_COMM_WORLD);
#else
tot_beam = beams;
#endif
t2 = wallclock_time(); t_comm += t2-t1;
/* write image to output file */
if (pe == root_pe) {
write_ImageFile(beam_file, tot_beam, shot_area, is, 0, beam_su, verbose);
}
t1 = wallclock_time(); t_io += t1-t2;
}
/* Start of depth loop */
for (d=0; d<ndepth; d+=dstep) {
t1 = wallclock_time();
id1 = d;
id2 = MIN(id1+dstep, ndepth);
nel = (id2-id1)*nxy;
/* Read dstep depth slices */
for (id=id1,i=0; id<id2; id++,i++) {
readVelocitySlice(vel_file, &velocity[i*nxy], id, nyv, nxv);
}
t2 = wallclock_time(); t_io += t2-t1;
if (verb_root > 1) {
fprintf(stderr," extrapolating from depth level ");
fprintf(stderr,"%d (%.2f) to %d (%.2f) \n",
id1, zvmin+dzv*id1, id2, zvmin+dzv*id2);
}
if (beam) memset(&beams[0], 0, nxy*dstep*sizeof(float));
for (iw=0; iw<nlw; iw++) {
om = freq_index[iw]*dw;
rec = (complex*) (rec_field + iw*nxy);
for (id=id1,i=0; id<id2; id++,i++) {
/* Extrapolation */
xwExtr3d(rec, &velocity[i*nxy], vmin, oplx, oply,
order, McC, om, nterms, filter_inc, ntap, tap_opt,
&shot_area, mode, method);
if (beam) {
for (ix = 0; ix < nxy; ix++) {
beams[i*nxy+ix] += sqrt(rec[ix].r*rec[ix].r+rec[ix].i*rec[ix].i)*scale;
}
}
} /* end of depth loop */
} /* end of frequency loop */
t1 = wallclock_time(); t_migr += t1-t2;
if (beam) {
#ifdef MPI
MPI_Reduce(beams, tot_beam, dstep*nxy, MPI_FLOAT, MPI_SUM, 0, MPI_COMM_WORLD);
#else
tot_beam = beams;
#endif
t3 = wallclock_time(); t_comm += t3-t1;
/* write beam to output file */
if (pe == root_pe) {
for (id=id1,i=0; id<id2; id++,i++) {
write_ImageFile(beam_file, &tot_beam[i*nxy], shot_area,
is, id, beam_su, verbose);
}
}
t2 = wallclock_time(); t_io += t2-t3;
}
} /* end of outer (dstep) depth loop */
t2 = wallclock_time();
/* communicate data from all PE's to root_pe */
#ifdef MPI
fflush(stderr);
if (pe == root_pe) {
displacements[0] = 0;
recvcounts[0] = nlw*nxy*2;
for( ipe=1; ipe<npes; ipe++ ) {
recvcounts[ipe] = nlwcounts[ipe]*nxy*2;
displacements[ipe] = displacements[ipe-1]+recvcounts[ipe-1];
}
}
MPI_Gatherv(rec_field, nxy*nlw*2, MPI_FLOAT, gath_rec_field, recvcounts,
displacements, MPI_FLOAT, root_pe, MPI_COMM_WORLD);
if (pe == root_pe) {
rec_all = gath_rec_field;
}
#else
rec_all = rec_field;
#endif
t1 = wallclock_time(); t_comm += t1-t2;
if (pe == root_pe) {
/* Write modelling result to output file */
if (verb_root)
fprintf(stderr," End of depth loop, writing data.\n");
if (is == 0)
out_file = fopen( file_out, "w+" );
else
out_file = fopen( file_out, "a" );
assert( out_file );
write_FFT_DataFile(out_file, rec_all, shot_area, (is+1),
nt, nfft, nw, nw_low, dt, out_su, conjg, verb_root);
fclose(out_file);
}
/* Read next shot record */
if (traces_read_in == ntraces) {
one_shot = 0;
}
else {
read_FFT_DataFile(shot_file, rec_field, shot_area, nfft,
nlw, freq_index[0], &traces_read_in, &traces_shot,
&ixmin, &ixmax, &iymin, &iymax, &sx, &sy, conjgs, verb_root);
}
is++;
t2 = wallclock_time(); t_io += t2-t1;
if (verb_root) {
fprintf(stderr," subtime for extrapolation : %.3f s.\n", t_migr);
fprintf(stderr," subtime for io : %.3f s.\n", t_io);
fprintf(stderr," subtime for communication : %.3f s.\n\n", t_comm);
}
} /* end of while loop over input traces */
free(velocity);
free(hdr);
free(rec_field);
fclose(vel_file);
if (beam) {
if (pe == root_pe) fclose(beam_file);
free(beams);
#ifdef MPI
free(tot_beam);
#endif
}
/* Write total image result to output file */
fclose(shot_file);
t2 = wallclock_time(); t_io += t2-t1;
/* clean temporary velocity files */
sprintf(sys_call,"rm -rf %s/velocity%d.bin\n",tmp_dir, getpid());
system(sys_call);
/* print the timing results */
if (verb_root) {
fprintf(stderr,"Time for total job : %.3f s.\n", t2-t0);
fprintf(stderr," time for extrapolation : %.3f s.\n", t_migr);
fprintf(stderr," time for tables : %.3f s.\n", t_table);
fprintf(stderr," time for io : %.3f s.\n", t_io);
fprintf(stderr," time for communication : %.3f s.\n", t_comm);
fprintf(stderr," time for initialization : %.3f s.\n", t_init);
}
#ifdef MPI
MPI_Barrier(MPI_COMM_WORLD);
if (pe == root_pe) free(gath_rec_field);
free(nlwcounts);
free(recvcounts);
free(displacements);
MPI_Finalize();
#endif
return;
}
void kwZoMigr(float *data, int nx, int nt, float dt, float *velmod, int nxm, int nzm, int ixa, int ixb, float fmin, float fmax, float *xrcv, int izrcv, float ox, float dxm, float dz, int ntap, int conjg, int ndepth, float *image, int verbose, float *exrcv, int ndepthex)
{
int iomin, iomax, iom, ix, d, i, j;
int nfreq, optn, nkx, sign, endkx;
int ixrcv, ixmin, ixmax, ixo, ixn, ikx;
float k, k2, kz2, kx, kx2;
float dom, om, c, dkx, df, sr;
float *taper, scl, scl2, *pdata;
float *trace;
complex *ctrace;
float t0, t1;
complex *cdata, tmp, ez;
complex wa, *ctmp, da, *locdat;
complex *cexrcv=(complex *) exrcv;
/* define some constants */
optn = optncr(nt);
nfreq = optn/2 + 1;
df = 1.0/(optn*dt);
dom = 2.0*M_PI*df;
iomin = (int)MIN((fmin*dt*optn), (nfreq-1));
iomin = MAX(iomin, 1);
iomax = MIN((int)(fmax*dt*optn), (nfreq-1));
/* transformation of shot record to frequency domain */
trace = (float *)calloc(optn,sizeof(float));
ctrace = (complex *)malloc(optn*sizeof(complex));
cdata = (complex *)calloc(nxm*nfreq, sizeof(complex));
if (conjg) scl = -1.0;
else scl = 1.0;
sign = -1;
for (ix = 0; ix < nx; ix++) {
memcpy(trace,&data[ix*nt],nt*sizeof(float));
if (optn > nt)
memset( &trace[nt], 0, sizeof(float)*(optn-nt) );
rc1fft(trace,ctrace,optn,sign);
ixrcv = NINT((xrcv[ix]-ox)/dxm);
if (ixrcv < 0 || ixrcv > nxm-1) {
fprintf(stderr,"kwZoMigr: ixrcv %f (%d) outside model\n", xrcv[ix], ixrcv);
continue;
}
for (iom=0; iom<nfreq; iom++) {
/* positioning of shot record into velocity model */
cdata[iom*nxm+ixrcv].r = ctrace[iom].r;
cdata[iom*nxm+ixrcv].i = ctrace[iom].i*scl;
}
}
/* determine aperture to be calculated */
ixo = nxm;
ixn = 0;
for (ix = 0; ix < nx; ix++) {
ixrcv = NINT((xrcv[ix]-ox)/dxm);
if (ixrcv < ixo) ixo = ixrcv;
if (ixrcv > ixn) ixn = ixrcv;
}
ixmin = MAX(0, ixo-ixb-1);
ixmax = MIN(ixn+ixa+1, nxm-1);
nx = (ixmax-ixmin)+1;
if (verbose>=2) {
vmess("kwZoMigr: calculation aperture: %.2f (%d) <--> %.2f (%d) (%d positions)", ixmin*dxm+ox, ixmin, ixmax*dxm+ox, ixmax, nx);
}
/* define some constants */
scl = 2.0/nfreq;
scl = 1.0/(dt*dxm*dxm);
nkx = optncc(2*ntap+nxm);
ntap = (nkx-nxm)/2;
scl2 = 1.0/nkx;
dkx = 2.0*M_PI/(nkx*dxm);
taper = (float *)malloc(ntap*sizeof(float));
for (ix = 0; ix < ntap; ix++) {
taper[ix] = exp(-1.0*(pow((0.4*(ntap-ix)/ntap), 2)));
}
/* calculate image at depth = 0 */
if(izrcv==0 ) {
for (ix = ixmin; ix <= ixmax; ix++) {
for (iom = iomin; iom <= iomax; iom++) {
image[ix*nzm+0] += scl*cdata[iom*nxm+ix].r;
}
}
}
t0 = wallclock_time();
locdat = (complex *)malloc(nkx*sizeof(complex));
/* start extrapolation for all frequencies, depths and x-positions */
for (iom = iomin; iom <= iomax; iom++) {
memset(locdat,0,nkx*sizeof(complex));
for (ix = ixmin; ix <= ixmax; ix++) {
locdat[ntap+ix] = cdata[iom*nxm+ix];
}
om = iom*dom;
d = izrcv;
/* start extrapolation of receiver arrays */
for (; d < ndepth; d++) {
/* transform to wavenumber domain */
cc1fft(locdat, nkx, 1);
/* Extrapolation of data */
c = 0.0;
for (ix = ixmin; ix <= ixmax; ix++) c += velmod[d*nxm+ix];
k = nx*om/c;
k2 = k*k;
/* kx = 0 */
ez.r = cos(k*dz);
ez.i = -sin(k*dz);
tmp.r = ez.r*locdat[0].r;
tmp.r += ez.i*locdat[0].i;
tmp.i = ez.r*locdat[0].i;
tmp.i -= ez.i*locdat[0].r;
locdat[0] = tmp;
/* kx != 0 */
endkx = MIN((int)(k/dkx),nkx/2);
for (ikx = 1; ikx <= endkx; ikx++) {
kx = ikx*dkx;
kx2 = kx*kx;
kz2 = k2 - kx2;
ez.r = cos(sqrt(kz2)*dz);
ez.i = -sin(sqrt(kz2)*dz);
tmp.r = ez.r*locdat[ikx].r;
tmp.r += ez.i*locdat[ikx].i;
tmp.i = ez.r*locdat[ikx].i;
tmp.i -= ez.i*locdat[ikx].r;
locdat[ikx] = tmp;
tmp.r = ez.r*locdat[nkx-ikx].r;
tmp.r += ez.i*locdat[nkx-ikx].i;
tmp.i = ez.r*locdat[nkx-ikx].i;
tmp.i -= ez.i*locdat[nkx-ikx].r;
locdat[nkx-ikx] = tmp;
}
/* transform data back to space domain */
cc1fft(locdat, nkx, -1);
for (j = 0; j < nkx; j++) {
locdat[j].r *= scl2;
locdat[j].i *= scl2;
}
/* imaging condition */
for (ix = ixmin; ix <= ixmax; ix++) {
image[ix*nzm+d+1]+= scl*locdat[ntap+ix].r;
}
/* save extrapolated field at requested depth */
if (d==ndepthex-1) {
if ( cexrcv != NULL ) {
for (ix = 0; ix < nxm; ix++) {
cexrcv[iom*nxm+ix] = locdat[ntap+ix];
}
}
}
/* taper extrapolated data at edges */
for (j = 0; j < ntap; j++) {
locdat[j].r *= taper[j];
locdat[j].i *= taper[j];
locdat[nkx-j-1].r *= taper[j];
locdat[nkx-j-1].i *= taper[j];
}
} /* end of depth loop */
} /* end of iom loop */
if(exrcv) wx2xt(cexrcv, exrcv, optn, nxm, nxm, optn);
free(locdat);
free(cdata);
free(taper);
t1 = wallclock_time();
vmess("kwZoMigr took: %f seconds", t1-t0);
return;
}
int main (int argc, char **argv)
{
FILE *fp_in1, *fp_in2, *fp_out, *fp_chk, *fp_psline1, *fp_psline2;
int verbose, shift, k, nx1, nt1, nx2, nt2;
int ntmax, nxmax, ret, i, j, jmax, imax, above, check;
int size, ntraces, ngath, *maxval, hw, smooth;
int tstart, tend, scale, *xrcv;
float dt, d2, f1, f2, t0, t1, f1b, f2b, d1, d1b, d2b;
float w1, w2, dxrcv;
float *tmpdata, *tmpdata2, *costaper;
char *file_mute, *file_shot, *file_out;
float scl, sclsxgx, sclshot, xmin, xmax, tmax, lmax;
segy *hdrs_in1, *hdrs_in2;
t0 = wallclock_time();
initargs(argc, argv);
requestdoc(1);
if(!getparstring("file_mute", &file_mute)) file_mute=NULL;
if(!getparstring("file_shot", &file_shot)) file_shot=NULL;
if(!getparstring("file_out", &file_out)) file_out=NULL;
if(!getparint("ntmax", &ntmax)) ntmax = 1024;
if(!getparint("nxmax", &nxmax)) nxmax = 512;
if(!getparint("above", &above)) above = 0;
if(!getparint("check", &check)) check = 0;
if(!getparint("scale", &scale)) scale = 0;
if(!getparint("hw", &hw)) hw = 15;
if(!getparint("smooth", &smooth)) smooth = 0;
if(!getparfloat("w1", &w1)) w1=1.0;
if(!getparfloat("w2", &w2)) w2=1.0;
if(!getparint("shift", &shift)) shift=0;
if(!getparint("verbose", &verbose)) verbose=0;
/* Reading input data for file_mute */
if (file_mute != NULL) {
ngath = 1;
getFileInfo(file_mute, &nt1, &nx1, &ngath, &d1, &d2, &f1, &f2, &xmin, &xmax, &sclsxgx, &ntraces);
if (!getparint("ntmax", &ntmax)) ntmax = nt1;
if (!getparint("nxmax", &nxmax)) nxmax = nx1;
if (verbose>=2 && (ntmax!=nt1 || nxmax!=nx1))
vmess("dimensions overruled: %d x %d",ntmax,nxmax);
if(!getparfloat("dt", &dt)) dt=d1;
fp_in1 = fopen(file_mute, "r");
if (fp_in1 == NULL) verr("error on opening input file_mute=%s", file_mute);
size = ntmax * nxmax;
tmpdata = (float *)malloc(size*sizeof(float));
hdrs_in1 = (segy *) calloc(nxmax,sizeof(segy));
nx1 = readData(fp_in1, tmpdata, hdrs_in1, nt1);
if (nx1 == 0) {
fclose(fp_in1);
if (verbose) vmess("end of file_mute data reached");
}
if (verbose) {
disp_fileinfo(file_mute, nt1, nx1, f1, f2, dt, d2, hdrs_in1);
}
}
/* Reading input data for file_shot */
ngath = 1;
getFileInfo(file_shot, &nt2, &nx2, &ngath, &d1b, &d2b, &f1b, &f2b, &xmin, &xmax, &sclshot, &ntraces);
if (!getparint("ntmax", &ntmax)) ntmax = nt2;
if (!getparint("nxmax", &nxmax)) nxmax = nx2;
size = ntmax * nxmax;
tmpdata2 = (float *)malloc(size*sizeof(float));
hdrs_in2 = (segy *) calloc(nxmax,sizeof(segy));
if (file_shot != NULL) fp_in2 = fopen(file_shot, "r");
else fp_in2=stdin;
if (fp_in2 == NULL) verr("error on opening input file_shot=%s", file_shot);
nx2 = readData(fp_in2, tmpdata2, hdrs_in2, nt2);
if (nx2 == 0) {
fclose(fp_in2);
if (verbose) vmess("end of file_shot data reached");
}
nt2 = hdrs_in2[0].ns;
f1b = hdrs_in2[0].f1;
f2b = hdrs_in2[0].f2;
d1b = (float)hdrs_in2[0].dt*1e-6;
if (verbose) {
disp_fileinfo(file_shot, nt2, nx2, f1b, f2b, d1b, d2b, hdrs_in2);
}
/* file_shot will be used as well to define the mute window */
if (file_mute == NULL) {
nx1=nx2;
nt1=nt2;
dt=d1b;
f1=f1b;
f2=f2b;
tmpdata = tmpdata2;
hdrs_in1 = hdrs_in2;
}
if (verbose) vmess("sampling file_mute=%d, file_shot=%d", nt1, nt2);
/*================ initializations ================*/
maxval = (int *)calloc(nx1,sizeof(int));
xrcv = (int *)calloc(nx1,sizeof(int));
if (file_out==NULL) fp_out = stdout;
else {
fp_out = fopen(file_out, "w+");
if (fp_out==NULL) verr("error on ceating output file");
}
if (check!=0){
fp_chk = fopen("check.su", "w+");
if (fp_chk==NULL) verr("error on ceating output file");
fp_psline1 = fopen("pslinepos.asci", "w+");
if (fp_psline1==NULL) verr("error on ceating output file");
fp_psline2 = fopen("pslineneg.asci", "w+");
if (fp_psline2==NULL) verr("error on ceating output file");
}
if (smooth) {
costaper = (float *)malloc(smooth*sizeof(float));
scl = M_PI/((float)smooth);
for (i=0; i<smooth; i++) {
costaper[i] = 0.5*(1.0+cos((i+1)*scl));
/* fprintf(stderr,"costaper[%d]=%f\n",i,costaper[i]);*/
}
}
/*================ loop over all shot records ================*/
k=1;
while (nx1 > 0) {
if (verbose) vmess("processing input gather %d", k);
/*================ loop over all shot records ================*/
/* find consistent (one event) maximum related to maximum value */
/* find global maximum
xmax=0.0;
for (i = 0; i < nx1; i++) {
tmax=0.0;
jmax = 0;
for (j = 0; j < nt1; j++) {
lmax = fabs(tmpdata[i*nt1+j]);
if (lmax > tmax) {
jmax = j;
tmax = lmax;
if (lmax > xmax) {
imax = i;
xmax=lmax;
}
}
}
maxval[i] = jmax;
}
*/
/* alternative find maximum at source position */
dxrcv = (hdrs_in1[nx1-1].gx - hdrs_in1[0].gx)*sclsxgx/(float)(nx1-1);
imax = NINT(((hdrs_in1[0].sx-hdrs_in1[0].gx)*sclsxgx)/dxrcv);
tmax=0.0;
jmax = 0;
for (j = 0; j < nt1; j++) {
lmax = fabs(tmpdata[imax*nt1+j]);
if (lmax > tmax) {
jmax = j;
tmax = lmax;
if (lmax > xmax) {
xmax=lmax;
}
}
}
maxval[imax] = jmax;
if (verbose >= 3) vmess("Mute max at src-trace %d is sample %d", imax, maxval[imax]);
/* search forward */
for (i = imax+1; i < nx1; i++) {
tstart = MAX(0, (maxval[i-1]-hw));
tend = MIN(nt1-1, (maxval[i-1]+hw));
jmax=tstart;
tmax=0.0;
for(j = tstart; j <= tend; j++) {
lmax = fabs(tmpdata[i*nt1+j]);
if (lmax > tmax) {
jmax = j;
tmax = lmax;
}
}
maxval[i] = jmax;
}
/* search backward */
for (i = imax-1; i >=0; i--) {
tstart = MAX(0, (maxval[i+1]-hw));
tend = MIN(nt1-1, (maxval[i+1]+hw));
jmax=tstart;
tmax=0.0;
for(j = tstart; j <= tend; j++) {
lmax = fabs(tmpdata[i*nt1+j]);
if (lmax > tmax) {
jmax = j;
tmax = lmax;
}
}
maxval[i] = jmax;
}
/* scale with maximum ampltiude */
if (scale==1) {
for (i = 0; i < nx2; i++) {
lmax = fabs(tmpdata2[i*nt2+maxval[i]]);
xrcv[i] = i;
for (j = 0; j < nt2; j++) {
tmpdata2[i*nt2+j] = tmpdata2[i*nt2+j]/lmax;
}
}
}
/*================ apply mute window ================*/
applyMute(tmpdata2, maxval, smooth, above, 1, nx2, nt2, xrcv, nx2, shift);
/*================ write result to output file ================*/
ret = writeData(fp_out, tmpdata2, hdrs_in2, nt2, nx2);
if (ret < 0 ) verr("error on writing output file.");
/* put mute window in file to check correctness of mute */
if (check !=0) {
for (i = 0; i < nx1; i++) {
jmax = maxval[i]-shift;
tmpdata[i*nt1+jmax] = 2*xmax;
}
if (above==0){
for (i = 0; i < nx1; i++) {
jmax = nt2-maxval[i]+shift;
tmpdata[i*nt1+jmax] = 2*xmax;
}
}
ret = writeData(fp_chk, tmpdata, hdrs_in1, nt1, nx1);
if (ret < 0 ) verr("error on writing check file.");
for (i=0; i<nx1; i++) {
jmax = maxval[i]-shift;
ret = fprintf(fp_psline1, "%.5f %.5f \n",jmax*dt,hdrs_in1[i].gx*sclshot);
jmax =-maxval[i]+shift;
ret = fprintf(fp_psline2, "%.5f %.5f \n",jmax*dt,hdrs_in1[i].gx*sclshot);
}
}
/*================ Read next record for muting ================*/
if (file_mute != NULL) {
nx1 = readData(fp_in1, tmpdata, hdrs_in1, nt1);
if (nx1 == 0) {
fclose(fp_in1);
if (verbose) vmess("end of file_mute data reached");
fclose(fp_in2);
if (fp_out!=stdout) fclose(fp_out);
if (check!=0) fclose(fp_chk);
if (check!=0) {
fclose(fp_psline1);
fclose(fp_psline2);
}
break;
}
nt1 = (int)hdrs_in1[0].ns;
if (nt1 > ntmax) verr("n_samples (%d) greater than ntmax", nt1);
if (nx1 > nxmax) verr("n_traces (%d) greater than nxmax", nx1);
}
/*================ Read next shot record(s) ================*/
nx2 = readData(fp_in2, tmpdata2, hdrs_in2, nt2);
if (nx2 == 0) {
if (verbose) vmess("end of file_shot data reached");
fclose(fp_in2);
break;
}
nt2 = (int)hdrs_in2[0].ns;
if (nt2 > ntmax) verr("n_samples (%d) greater than ntmax", nt2);
if (nx2 > nxmax) verr("n_traces (%d) greater than nxmax", nx2);
if (file_mute == NULL) {
nx1=nx2;
nt1=nt2;
hdrs_in1 = hdrs_in2;
tmpdata = tmpdata2;
}
k++;
}
t1 = wallclock_time();
if (verbose) vmess("Total CPU-time = %f",t1-t0);
return 0;
}
