SparseMatrix * loadSparseMatrix(char * filename, int pad)
{
char input[128];
strcpy(input,filename);
strcat(input,".bin");
printf("loading binary: %s\n", input);
SparseMatrix * m = bmpaddedread(input,pad);
if(m == NULL)
{
strcpy(input,filename);
strcat(input,".txt");
printf("binary not found, loading text and then saving as binary for next time: %s\n",input);
m = mread(input);
strcphttps://mail.google.com/mail/u/0/?ui=2&shva=1#inboxy(input,filename);
strcat(input,".bin");
bmwrite(input,m);
}
assert(m);
int i;
for(i = 0; i < 10; i++)
{
printf("%d, %e, %d\n",m->cpu_cols[i],m->cpu_vals[i],m->cpu_index[i]);
}
int result;
//make the gpu side data structure
m->gpu_cols = clCreateBuffer(context,CL_MEM_READ_WRITE, sizeof(int)*m->colsLength,NULL,&result);
m->gpu_vals = clCreateBuffer(context,CL_MEM_READ_WRITE, sizeof(TYPE)*m->colsLength,NULL,&result);
m->gpu_index = clCreateBuffer(context,CL_MEM_READ_WRITE, sizeof(int)*m->indexLength,NULL,&result);
m->locality = 0;
return m;
}
int
main(int argc, char **argv)
{
float **data;
char **hdrdata;
float ***workm;
int ival2; /* int value of key2 */
int ival3; /* int value of key3 */
Value val2; /* ... its value */
Value val3; /* ... its value */
int index2;
int index3;
int n1,n2,n3;
int i1,i2,i3;
char *key2=NULL; /* header key word from segy.h */
char *type2=NULL; /* ... its type */
char *key3=NULL; /* header key word from segy.h */
char *type3=NULL; /* ... its type */
int lins;
int line;
int nl;
int dir=2;
int il;
int su;
float cdp;
float t;
float dt;
/* Initialize */
initargs(argc, argv);
requestdoc(1);
MUSTGETPARINT("n2", &n2);
MUSTGETPARINT("n3", &n3);
/* get key2*/
if (!getparstring("key2", &key2)) key2 = "fldr";
type2 = hdtype(key2);
index2 = getindex(key2);
/* get key3*/
if (!getparstring("key3", &key3)) key3 = "tracf";
type3 = hdtype(key3);
index3 = getindex(key3);
if (!getparint("lins", &lins)) lins = 1;
if(lins<1) err(" lins must be larger than 0");
if (!getparint("line", &line)) line = n2;
if (!getparint("su", &su)) su = 1;
/* Get info from first trace */
if (!gettr(&tr)) err ("can't get first trace");
n1 = tr.ns;
if (!getparfloat("dt", &dt)) dt = ((float) tr.dt)/1000000.0;
if (!dt) {
dt = .01;
warn("dt not set, assumed to be .01");
}
data = bmalloc(n1*sizeof(float),n2,n3);
hdrdata = bmalloc(HDRBYTES,n2,n3);
do {
gethval(&tr, index2, &val2);
ival2 = vtoi(type2,val2);
gethval(&tr, index3, &val3);
ival3 = vtoi(type3,val3);
if(ival2>n2 || ival2<0 ) err(" Array in dimension 2 out of bound\n");
if(ival3>n3 || ival3<0 ) err(" Array in dimension 3 out of bound\n");
bmwrite(data,1,ival2-1,ival3-1,1,tr.data);
bmwrite(hdrdata,1,ival2-1,ival3-1,1,&tr);
} while (gettr(&tr));
nl=line-lins+1;
if(dir==2) {
/* n2 direction */
workm = alloc3float(n1,n3,nl);
for(il=lins-1;il<nl;il++) {
for(i3=0;i3<n3;i3++) bmread(data,1,lins+il-1,i3,1,workm[il][i3]);
}
if(su==1) {
fprintf(stdout,"cdp=");
for(il=lins-1;il<nl;il++) {
for(i3=0;i3<n3;i3++) fprintf(stdout,"%d,",(lins+il)*1000+i3+1);
}
for(il=lins-1;il<nl;il++) {
for(i3=0;i3<n3;i3++){
fprintf(stdout,"\ntnmo=");
for(i1=0;i1<n1;i1++) fprintf(stdout,"%.3f,",dt*i1);
fprintf(stdout,"\nvnmo=");
for(i1=0;i1<n1;i1++)
fprintf(stdout,"%.3f,",workm[il][i3][i1]);
}
}
} else {
for(il=lins-1;il<nl;il++) {
for(i3=0;i3<n3;i3++) {
cdp=(lins+il)*1000+i3+1;
fwrite(&cdp,sizeof(float),1,stdout);
}
}
for(il=lins-1;il<nl;il++) {
for(i3=0;i3<n3;i3++){
for(i1=0;i1<n1;i1++) {
t=tr.d1*i1;
fwrite(&t,sizeof(float),1,stdout);
}
for(i1=0;i1<n1;i1++)
fwrite(&workm[il][i3][i1],sizeof(float),1,stdout);
}
}
}
free3float(workm);
}
bmfree(data);
bmfree(hdrdata);
return EXIT_SUCCESS;
}
int main( int argc, char *argv[] )
{
int ntr=0; /* number of traces */
int ntrv=0; /* number of traces */
int ns=0;
int nsv=0;
float dt;
float dtv;
cwp_String fs;
cwp_String fv;
FILE *fps;
FILE *fpv;
FILE *headerfp;
float *data; /* data matrix of the migration volume */
float *vel; /* velocity matrix */
float *velfi; /* velocity function interpolated to ns values*/
float *velf; /* velocity function */
float *vdt;
float *ddt;
float *ap; /* array of apperture values in m */
float apr; /* array of apperture values in m */
int *apt=NULL; /* array of apperture time limits in mig. gath*/
float r; /* maximum radius with a given apperture */
float ir2; /* r/d2 */
float ir3; /* r/d3 */
float d2; /* spatial sampling int. in dir 2. */
float d3; /* spatial sampling int. in dir 3. */
float **mgd=NULL; /* migration gather data */
float *migt; /* migrated data trace */
int **mgdnz=NULL; /* migration gather data non zero samples*/
float dm; /* migration gather spatial sample int. */
int im; /* number of traces in migration gather */
int *mtnz; /* migrated trace data non zero smaples */
char **dummyi; /* index array that the trace contains zeros only */
float fac; /* velocity scale factor */
int sphr; /* spherical divergence flag */
int imt; /* mute time sample of trace */
float tmp;
int imoff;
int **igtr=NULL;
int nigtr;
int n2;
int n3;
int verbose;
/* phase shift filter stuff */
float power; /* power of i omega applied to data */
float amp; /* amplitude associated with the power */
float arg; /* argument of power */
float phasefac; /* phase factor */
float phase; /* phase shift = phasefac*PI */
complex exparg; /* cexp(I arg) */
register float *rt; /* real trace */
register complex *ct; /* complex transformed trace */
complex *filt; /* complex power */
float omega; /* circular frequency */
float domega; /* circular frequency spacing (from dt) */
float sign; /* sign in front of i*omega default -1 */
int nfft; /* number of points in nfft */
int nf; /* number of frequencies (incl Nyq) */
float onfft; /* 1 / nfft */
size_t nzeros; /* number of padded zeroes in bytes */
initargs(argc, argv);
requestdoc(1);
MUSTGETPARSTRING("fs",&fs);
MUSTGETPARSTRING("fv",&fv);
MUSTGETPARINT("n2",&n2);
MUSTGETPARINT("n3",&n3);
MUSTGETPARFLOAT("d2",&d2);
MUSTGETPARFLOAT("d3",&d3);
if (!getparfloat("dm", &dm)) dm=(d2+d3)/2.0;
/* open datafile */
fps = efopen(fs,"r");
fpv = efopen(fv,"r");
/* Open tmpfile for headers */
headerfp = etmpfile();
/* get information from the first data trace */
ntr = fgettra(fps,&tr,0);
if(n2*n3!=ntr) err(" Number of traces in file %d not equal to n2*n3 %d \n",
ntr,n2*n3);
ns=tr.ns;
if (!getparfloat("dt", &dt)) dt = ((float) tr.dt)/1000000.0;
if (!dt) {
dt = .002;
warn("dt not set, assumed to be .002");
}
/* get information from the first velocity trace */
ntrv = fgettra(fpv,&trv,0);
if(ntrv!=ntr) err(" Number of traces in velocity file %d differ from %d \n",
ntrv,ntr);
nsv=trv.ns;
if (!getparfloat("dtv", &dtv)) dtv = ((float) trv.dt)/1000000.0;
if (!dtv) {
dtv = .002;
warn("dtv not set, assumed to be .002 for velocity");
}
if (!getparfloat("fac", &fac)) fac=2.0;
if (!getparint("verbose", &verbose)) verbose=0;
if (!getparint("sphr", &sphr)) sphr=0;
if (!getparfloat("apr", &apr)) apr=75;
apr*=3.141592653/180;
/* allocate arrays */
data = bmalloc(sizeof(float),ns,ntr);
vel = bmalloc(sizeof(float),nsv,ntr);
velf = ealloc1float(nsv);
velfi = ealloc1float(ns);
migt = ealloc1float(ns);
vdt = ealloc1float(nsv);
ddt = ealloc1float(ns);
ap = ealloc1float(ns);
mtnz = ealloc1int(ns);
dummyi = (char **) ealloc2(n2,n3,sizeof(char));
/* Times to do interpolation of velocity from sparse sampling */
/* to fine sampling of the data */
{ register int it;
for(it=0;it<nsv;it++) vdt[it]=it*dtv;
for(it=0;it<ns;it++) ddt[it]=it*dt;
}
/* Read traces into data */
/* Store headers in tmpfile */
ntr=0;
erewind(fps);
erewind(fpv);
{ register int i2,i3;
for(i3=0;i3<n3;i3++)
for(i2=0;i2<n2;i2++) {
fgettr(fps,&tr);
fgettr(fpv,&trv);
if(tr.trid > 2) dummyi[i3][i2]=1;
else dummyi[i3][i2]=0;
efwrite(&tr, 1, HDRBYTES, headerfp);
bmwrite(data,1,0,i3*n2+i2,ns,tr.data);
bmwrite(vel,1,0,i3*n2+i2,nsv,trv.data);
}
erewind(headerfp);
/* set up the phase filter */
power = 1.0;sign = 1.0;phasefac = 0.5;
phase = phasefac * PI;
/* Set up for fft */
nfft = npfaro(ns, LOOKFAC * ns);
if (nfft >= SU_NFLTS || nfft >= PFA_MAX)
err("Padded nt=%d -- too big", nfft);
nf = nfft/2 + 1;
onfft = 1.0 / nfft;
nzeros = (nfft - ns) * FSIZE;
domega = TWOPI * onfft / dt;
/* Allocate fft arrays */
rt = ealloc1float(nfft);
ct = ealloc1complex(nf);
filt = ealloc1complex(nf);
/* Set up args for complex power evaluation */
arg = sign * PIBY2 * power + phase;
exparg = cexp(crmul(I, arg));
{
register int i;
for (i = 0 ; i < nf; ++i) {
omega = i * domega;
/* kludge to handle omega=0 case for power < 0 */
if (power < 0 && i == 0) omega = FLT_MAX;
/* calculate filter */
amp = pow(omega, power) * onfft;
filt[i] = crmul(exparg, amp);
}
}
/* set up constants for migration */
if(verbose) fprintf(stderr," Setting up constants....\n");
r=0;
for(i3=0;i3<n3;i3++)
for(i2=0;i2<n2;i2++) {
if(dummyi[i3][i2] < 1) {
/* get the velocity function */
bmread(vel,1,0,i3*n2+i2,nsv,velf);
/* linear interpolation from nsv to ns values */
intlin(nsv,vdt,velf,velf[0],velf[nsv-1],ns,ddt,velfi);
/* Apply scale factor to velocity */
{ register int it;
for(it=0;it<ns;it++) velfi[it] *=fac;
}
/* compute maximum radius from apperture and velocity */
{ register int it;
for(it=0;it<ns;it++)
ap[it] = ddt[it]*velfi[it]*tan(apr)/2.0;
}
tmp = ap[isamax(ns,ap,1)];
if(tmp>r) r=tmp;
}
}
r=MIN(r,sqrt(SQR((n2-1)*d2)+SQR((n3-1)*d3)));
ir2 = (int)(2*r/d2)+1;
ir3 = (int)(2*r/d3)+1;
im = (int)(r/dm)+1;
/* allocate migration gather */
mgd = ealloc2float(ns,im);
mgdnz = ealloc2int(ns,im);
apt = ealloc1int(im);
/* set up the stencil for selecting traces */
igtr = ealloc2int(ir2*ir3,2);
stncl(r, d2, d3,igtr,&nigtr);
if(verbose) {
fprintf(stderr," Maximum radius %f\n",r);
fprintf(stderr," Maximum offset %f\n",
sqrt(SQR((n2-1)*d2)+SQR((n3-1)*d3)));
}
/* main processing loop */
for(i3=0;i3<n3;i3++)
for(i2=0;i2<n2;i2++) {
memset( (void *) tr.data, (int) '\0',ns*FSIZE);
if(dummyi[i3][i2] < 1) {
memset( (void *) mgd[0], (int) '\0',ns*im*FSIZE);
memset( (void *) mgdnz[0], (int) '\0',ns*im*ISIZE);
/* get the velocity function */
bmread(vel,1,0,i3*n2+i2,nsv,velf);
/* linear interpolation from nsv to ns values */
intlin(nsv,vdt,velf,velf[0],velf[nsv-1],ns,ddt,velfi);
/* Apply scale factor to velocity */
{ register int it;
for(it=0;it<ns;it++) velfi[it] *=fac;
}
/* create the migration gather */
{ register int itr,ist2,ist3;
for(itr=0;itr<nigtr;itr++) {
ist2=i2+igtr[0][itr];
ist3=i3+igtr[1][itr];
if(ist2 >= 0 && ist2 <n2)
if(ist3 >= 0 && ist3 <n3) {
if(dummyi[ist3][ist2] <1) {
imoff = (int) (
sqrt(SQR(igtr[0][itr]*d2)
+SQR(igtr[1][itr]*d3))/dm+0.5);
bmread(data,1,0,ist3*n2+ist2,ns,tr.data);
imoff=MIN(imoff,im-1);
{ register int it;
/* get the mute time for this
offset, apperture and velocity */
xindex(ns,ap,imoff*dm,&imt);
for(it=imt;it<ns;it++)
if(tr.data[it]!=0) {
mgd[imoff][it]+=tr.data[it];
mgdnz[imoff][it]+=1;
}
}
}
}
}
}
/* normalize the gather */
{ register int ix,it;
for(ix=0;ix<im;ix++)
for(it=0;it<ns;it++)
if(mgdnz[ix][it] > 1) mgd[ix][it] /=(float) mgdnz[ix][it];
}
memset( (void *) tr.data, (int) '\0',ns*FSIZE);
memset( (void *) mtnz, (int) '\0',ns*ISIZE);
/* do a knmo */
{ register int ix,it;
for(ix=0;ix<im;ix++) {
/* get the mute time for this
offset, apperture and velocity */
xindex(ns,ap,ix*dm,&imt);
knmo(mgd[ix],migt,ns,velfi,0,ix*dm,dt,imt,sphr);
/* stack the gather */
for(it=0;it<ns;it++) {
if(migt[it]!=0.0) {
tr.data[it] += migt[it];
mtnz[it]++;
}
/* tr.data[it] += mgd[ix][it]; */
}
}
}
{ register int it;
for(it=0;it<ns;it++)
if(mtnz[it]>1) tr.data[it] /=(float)mtnz[it];
}
/*Do the phase filtering before the trace is released*/
/* Load trace into rt (zero-padded) */
memcpy( (void *) rt, (const void *) tr.data, ns*FSIZE);
memset((void *) (rt + ns), (int) '\0', nzeros);
pfarc(1, nfft, rt, ct);
{ register int i;
for (i = 0; i < nf; ++i) ct[i] = cmul(ct[i], filt[i]);
}
pfacr(-1, nfft, ct, rt);
memcpy( (void *) tr.data, (const void *) rt, ns*FSIZE);
} /* end of dummy if */
/* spit out the gather */
efread(&tr, 1, HDRBYTES, headerfp);
puttr(&tr);
if(verbose) fprintf(stderr," %d %d\n",i2,i3);
} /* end of i2 loop */
} /* end of i3 loop */
/* This should be the last thing */
efclose(headerfp);
/* Free memory */
free2int(igtr);
free2float(mgd);
free2int(mgdnz);
free1int(apt);
bmfree(data);
bmfree(vel);
free1float(velfi);
free1float(velf);
free1float(ddt);
free1float(vdt);
free1float(ap);
free1int(mtnz);
free1float(migt);
free1float(rt);
free1complex(ct);
free1complex(filt);
free2((void **) dummyi);
return EXIT_SUCCESS;
}
