/*
* Class: jfftw_complex_nd_Plan
* Method: transform
* Signature: ([D)[D
*/
JNIEXPORT jdoubleArray JNICALL Java_jfftw_complex_nd_Plan_transform___3D( JNIEnv* env, jobject obj, jdoubleArray in )
{
jdouble *cin, *cout;
jdoubleArray out;
int i;
jclass clazz = (*env)->GetObjectClass( env, obj );
jfieldID id = (*env)->GetFieldID( env, clazz, "plan", "[B" );
jbyteArray arr = (jbyteArray)(*env)->GetObjectField( env, obj, id );
unsigned char* carr = (*env)->GetByteArrayElements( env, arr, 0 );
fftwnd_plan plan = *(fftwnd_plan*)carr;
int length = 1;
for( i = 0; i < plan->rank; ++i ) length *= plan->n[i];
if( length * 2 != (*env)->GetArrayLength( env, in ) )
{
(*env)->ThrowNew( env, (*env)->FindClass( env, "java/lang/IndexOutOfBoundsException" ), "the Plan was created for a different length" );
(*env)->ReleaseByteArrayElements( env, arr, carr, 0 );
return NULL;
}
cin = (*env)->GetDoubleArrayElements( env, in, 0 );
if( plan->rank > 0 && ! plan->plans[0]->flags & FFTW_THREADSAFE )
{
// synchronization
(*env)->MonitorEnter( env, obj );
}
if( plan->is_in_place )
{
out = in;
fftwnd_one( plan, (fftw_complex*)cin, NULL );
}
else
{
out = (*env)->NewDoubleArray( env, length * 2 );
cout = (*env)->GetDoubleArrayElements( env, out, 0 );
fftwnd_one( plan, (fftw_complex*)cin, (fftw_complex*)cout );
(*env)->ReleaseDoubleArrayElements( env, out, cout, 0 );
}
if( plan->rank > 0 && ! plan->plans[0]->flags & FFTW_THREADSAFE )
{
// synchronization
(*env)->MonitorExit( env, obj );
}
(*env)->ReleaseDoubleArrayElements( env, in, cin, 0 );
(*env)->ReleaseByteArrayElements( env, arr, carr, 0 );
return out;
}
//------------------------------------------------------------------------------------
int fdct3d_inverse_center(int N1,int N2,int N3,int b, double L1,double L2,double L3, int s,
CpxCrvletPrtd& C, CpxNumTnsBlkd& W)
{
int mpirank; MPI_Comm_rank(MPI_COMM_WORLD, &mpirank);
vector< vector<int> >& Cowners = C.owners();
if(Cowners[0][0]==mpirank) {
int S1, S2, S3; int F1, F2, F3; double R1, R2, R3; fdct3d_rangecompute(L1, L2, L3, S1, S2, S3, F1, F2, F3, R1, R2, R3);
DblOffVec big1(S1); fdct3d_lowpass(L1, big1);
DblOffVec big2(S2); fdct3d_lowpass(L2, big2);
DblOffVec big3(S3); fdct3d_lowpass(L3, big3);
CpxNumTns T(S1,S2,S3);
CpxNumTns& Cblk = C.block(0,0); //center block
T = Cblk;
fftwnd_plan p = fftw3d_create_plan(S3,S2,S1, FFTW_FORWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
fftwnd_one(p, (fftw_complex*)T.data(), NULL);
fftwnd_destroy_plan(p);
double sqrtprod = sqrt(double(S1*S2*S3));
for(int i=0; i<S1; i++) for(int j=0; j<S2; j++) for(int k=0; k<S3; k++) T(i,j,k) /= sqrtprod;
CpxOffTns A(S1,S2,S3);
fdct3d_fftshift(S1,S2,S3,T,A);
for(int i=-S1/2; i<-S1/2+S1; i++)
for(int j=-S2/2; j<-S2/2+S2; j++)
for(int k=-S3/2; k<-S3/2+S3; k++) {
int bi,bj,bk; int oi,oj,ok; fdct3d_position_aux(N1,N2,N3,b, i,j,k, bi,bj,bk,oi,oj,ok);
CpxNumTns& Wblk = W.block(bi,bj,bk);
Wblk(oi,oj,ok) += A(i,j,k) * (big1(i)*big2(j)*big3(k));
}
//done
}
return 0;
}
//-----------------------------------------------------------------------
int fdct_wrapping_wavelet(CpxOffMat& Xhgh, vector<CpxNumMat>& csc)
{
int N1 = Xhgh.m(); int N2 = Xhgh.n();
int F1 = -Xhgh.s(); int F2 = -Xhgh.t();
CpxNumMat T(N1, N2);
fdct_wrapping_ifftshift(Xhgh, T);
fftwnd_plan p ;
#pragma omp critical
{
p = fftw2d_create_plan(N2, N1, FFTW_BACKWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
}
fftwnd_one(p, (fftw_complex*)T.data(), NULL);
#pragma omp critical
{
fftwnd_destroy_plan(p);
}
double sqrtprod = sqrt(double(N1*N2));
for(int j=0; j<N2; j++)
for(int i=0; i<N1; i++)
T(i,j) /= sqrtprod;
csc[0] = T; //csc[0].resize(N1, N2); //fdct_wrapping_fftshift(T, csc[0]);
return 0;
}
void testnd_out_of_place(int rank, int *n, fftw_direction dir,
fftwnd_plan validated_plan)
{
int istride, ostride;
int N, dim, i;
fftw_complex *in1, *in2, *out1, *out2;
fftwnd_plan p;
int flags = measure_flag | wisdom_flag;
if (coinflip())
flags |= FFTW_THREADSAFE;
N = 1;
for (dim = 0; dim < rank; ++dim)
N *= n[dim];
in1 = (fftw_complex *) fftw_malloc(N * MAX_STRIDE * sizeof(fftw_complex));
out1 = (fftw_complex *) fftw_malloc(N * MAX_STRIDE * sizeof(fftw_complex));
in2 = (fftw_complex *) fftw_malloc(N * sizeof(fftw_complex));
out2 = (fftw_complex *) fftw_malloc(N * sizeof(fftw_complex));
p = fftwnd_create_plan(rank, n, dir, flags);
for (istride = 1; istride <= MAX_STRIDE; ++istride) {
/* generate random inputs */
for (i = 0; i < N; ++i) {
int j;
c_re(in2[i]) = DRAND();
c_im(in2[i]) = DRAND();
for (j = 0; j < istride; ++j) {
c_re(in1[i * istride + j]) = c_re(in2[i]);
c_im(in1[i * istride + j]) = c_im(in2[i]);
}
}
for (ostride = 1; ostride <= MAX_STRIDE; ++ostride) {
int howmany = (istride < ostride) ? istride : ostride;
if (howmany != 1 || istride != 1 || ostride != 1 || coinflip())
fftwnd(p, howmany, in1, istride, 1, out1, ostride, 1);
else
fftwnd_one(p, in1, out1);
fftwnd(validated_plan, 1, in2, 1, 1, out2, 1, 1);
for (i = 0; i < howmany; ++i)
CHECK(compute_error_complex(out1 + i, ostride, out2, 1, N)
< TOLERANCE,
"testnd_out_of_place: wrong answer");
}
}
fftwnd_destroy_plan(p);
fftw_free(out2);
fftw_free(in2);
fftw_free(out1);
fftw_free(in1);
}
int
gmx_fft_3d(gmx_fft_t fft,
enum gmx_fft_direction dir,
void * in_data,
void * out_data)
{
int inplace = (in_data == out_data);
int isforward = (dir == GMX_FFT_FORWARD);
if((fft->ndim != 3) ||
((dir != GMX_FFT_FORWARD) && (dir != GMX_FFT_BACKWARD)))
{
gmx_fatal(FARGS,"FFT plan mismatch - bad plan or direction.");
return EINVAL;
}
fftwnd_one(fft->multi[inplace][isforward],(fftw_complex *)in_data,(fftw_complex *)out_data);
return 0;
}
//---------------------
int fdct_wrapping_invwavelet(vector<CpxNumMat>& csc, CpxOffMat& Xhgh)
{
assert(csc.size()==1);
CpxNumMat& C = csc[0];
int N1 = C.m(); int N2 = C.n();
CpxNumMat T(C); //CpxNumMat T(N1, N2); fdct_wrapping_ifftshift(N1, N2, F1, F2, C, T);
fftwnd_plan p = fftw2d_create_plan(N2, N1, FFTW_FORWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
fftwnd_one(p, (fftw_complex*)T.data(), NULL);
fftwnd_destroy_plan(p);
double sqrtprod = sqrt(double(N1*N2));
for(int j=0; j<N2; j++)
for(int i=0; i<N1; i++)
T(i,j) /= sqrtprod;
Xhgh.resize(N1, N2);
fdct_wrapping_fftshift(T, Xhgh);
return 0;
}
void CMainFrame::Fft_back(Complex *array) {
int m_iDem = m_iDem_ampl;
fftw_complex *in = new fftw_complex[m_iDem * m_iDem];
size_t dxdy = m_iDem * m_iDem;
for(size_t i = 0; i < dxdy; i++) {
in[i].re = array[i].re;
in[i].im = array[i].im;
}
fftwnd_plan p = fftw2d_create_plan(m_iDem, m_iDem, FFTW_BACKWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
fftwnd_one(p, in, NULL);
fftwnd_destroy_plan(p);
for(size_t i = 0; i < dxdy; i++) {
array[i].re = (float) in[i].re;
array[i].im = (float) in[i].im;
}
delete[] in;
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix;
int xx0, xx1, xx2, xx3;
int y0min, y0max, y1min, y1max, y2min, y2max, y3min, y3max;
int y0, y1, y2, y3, iy;
int z0, z1, z2, z3, iz;
int gid, status;
int model_type = -1;
double *disc = (double*)NULL;
double *disc2 = (double*)NULL;
double *work = (double*)NULL;
double q[4], fnorm;
char filename[100], contype[200];
double ratime, retime;
double rmin2, rmax2, rsqr;
complex w, w1;
FILE *ofs;
fftw_complex *in=(fftw_complex*)NULL;
#ifdef MPI
fftwnd_mpi_plan plan_p, plan_m;
#else
fftwnd_plan plan_p, plan_m;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?f:t:")) != -1) {
switch (c) {
case 't':
model_type = atoi(optarg);
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
if(filename_set==0) strcpy(filename, "cvc.input");
fprintf(stdout, "# Reading input from file %s\n", filename);
read_input_parser(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
fprintf(stdout, "\n**************************************************\n");
fprintf(stdout, "* vp_disc_ft\n");
fprintf(stdout, "**************************************************\n\n");
#ifdef MPI
if(g_cart_id==0) fprintf(stdout, "# Warning: MPI-version not yet available; exit\n");
exit(200);
#endif
/*********************************
* initialize MPI parameters
*********************************/
mpi_init(argc, argv);
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifdef MPI
plan_p = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_BACKWARD, FFTW_MEASURE);
plan_m = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_FORWARD, FFTW_MEASURE);
fftwnd_mpi_local_sizes(plan_p, &T, &Tstart, &l_LX_at, &l_LXstart_at, &FFTW_LOC_VOLUME);
#else
plan_p = fftwnd_create_plan(4, dims, FFTW_BACKWARD, FFTW_MEASURE | FFTW_IN_PLACE);
plan_m = fftwnd_create_plan(4, dims, FFTW_FORWARD, FFTW_MEASURE | FFTW_IN_PLACE);
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] fftw parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(2);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
/****************************************
* allocate memory for the contractions
****************************************/
disc = (double*)calloc( 8*VOLUME, sizeof(double));
if( disc == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
disc2 = (double*)calloc( 32*VOLUME, sizeof(double));
if( disc2 == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc2\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
for(ix=0; ix<32*VOLUME; ix++) disc2[ix] = 0.;
work = (double*)calloc(32*VOLUME, sizeof(double));
if( work == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for work\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
/****************************************
* prepare Fourier transformation arrays
****************************************/
in = (fftw_complex*)malloc(FFTW_LOC_VOLUME*sizeof(fftw_complex));
if(in==(fftw_complex*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(4);
}
/***************************************
* set model type function
***************************************/
switch (model_type) {
case 0:
model_type_function = pidisc_model;
fprintf(stdout, "# function pointer set to type pidisc_model\n");
case 1:
model_type_function = pidisc_model1;
fprintf(stdout, "# function pointer set to type pidisc_model1\n");
break;
case 2:
model_type_function = pidisc_model2;
fprintf(stdout, "# function pointer set to type pidisc_model2\n");
break;
case 3:
model_type_function = pidisc_model3;
fprintf(stdout, "# function pointer set to type pidisc_model3\n");
break;
default:
model_type_function = NULL;
fprintf(stdout, "# no model function selected; will add zero\n");
break;
}
/****************************************
* prepare the model for pidisc
* - same for all gauge configurations
****************************************/
rmin2 = g_rmin * g_rmin;
rmax2 = g_rmax * g_rmax;
if(model_type > -1) {
for(mu=0; mu<16; mu++) {
model_type_function(model_mrho, model_dcoeff_re, model_dcoeff_im, work, plan_m, mu);
for(x0=-(T-1); x0<T; x0++) {
y0 = (x0 + T_global) % T_global;
for(x1=-(LX-1); x1<LX; x1++) {
y1 = (x1 + LX) % LX;
for(x2=-(LY-1); x2<LY; x2++) {
y2 = (x2 + LY) % LY;
for(x3=-(LZ-1); x3<LZ; x3++) {
y3 = (x3 + LZ) % LZ;
iy = g_ipt[y0][y1][y2][y3];
rsqr = (double)(x1*x1) + (double)(x2*x2) + (double)(x3*x3);
if(rmin2-rsqr<=_Q2EPS && rsqr-rmax2<=_Q2EPS) continue; /* radius in range for data usage, so continue */
disc2[_GWI(mu,iy,VOLUME) ] += work[2*iy ];
disc2[_GWI(mu,iy,VOLUME)+1] += work[2*iy+1];
}}}}
memcpy((void*)in, (void*)(disc2+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(disc2+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
}
} else {
for(ix=0; ix<32*VOLUME; ix++) disc2[ix] = 0.;
}
/***********************************************
* start loop on gauge id.s
***********************************************/
for(gid=g_gaugeid; gid<=g_gaugeid2; gid+=g_gauge_step) {
if(g_cart_id==0) fprintf(stdout, "# Start working on gauge id %d\n", gid);
/* read the new contractions */
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
sprintf(filename, "%s.%.4d.%.4d", filename_prefix, gid, Nsave);
if(g_cart_id==0) fprintf(stdout, "# Reading contraction data from file %s\n", filename);
if(read_lime_contraction(disc, filename, 4, 0) == 106) {
if(g_cart_id==0) fprintf(stderr, "Error, could not read from file %s, continue\n", filename);
continue;
}
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to read contraction: %e seconds\n", retime-ratime);
/************************************************
* prepare \Pi_\mu\nu (x,y)
************************************************/
# ifdef MPI
ratime = MPI_Wtime();
# else
ratime = (double)clock() / CLOCKS_PER_SEC;
# endif
for(x0=-T+1; x0<T; x0++) {
y0min = x0<0 ? -x0 : 0;
y0max = x0<0 ? T : T-x0;
for(x1=-LX+1; x1<LX; x1++) {
y1min = x1<0 ? -x1 : 0;
y1max = x1<0 ? LX : LX-x1;
for(x2=-LY+1; x2<LY; x2++) {
y2min = x2<0 ? -x2 : 0;
y2max = x2<0 ? LY : LY-x2;
for(x3=-LZ+1; x3<LZ; x3++) {
y3min = x3<0 ? -x3 : 0;
y3max = x3<0 ? LZ : LZ-x3;
xx0 = (x0+T ) % T;
xx1 = (x1+LX) % LX;
xx2 = (x2+LX) % LY;
xx3 = (x3+LX) % LZ;
ix = g_ipt[xx0][xx1][xx2][xx3];
rsqr = (double)(x1*x1) + (double)(x2*x2) + (double)(x3*x3);
if(rmin2-rsqr>_Q2EPS || rsqr-rmax2>_Q2EPS) continue;
for(y0=y0min; y0<y0max; y0++) {
z0 = y0 + x0;
for(y1=y1min; y1<y1max; y1++) {
z1 = y1 + x1;
for(y2=y2min; y2<y2max; y2++) {
z2 = y2 + x2;
for(y3=y3min; y3<y3max; y3++) {
z3 = y3 + x3;
iy = g_ipt[y0][y1][y2][y3];
iz = g_ipt[z0][z1][z2][z3];
i=0;
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
iix = _GWI(i,ix,VOLUME);
_co_eq_co_ti_co(&w, (complex*)(disc+_GWI(mu,iz,VOLUME)), (complex*)(disc+_GWI(nu,iy,VOLUME)));
work[iix ] += w.re;
work[iix+1] += w.im;
i++;
}}
}}}}
}}}}
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to calculate \\Pi_\\mu\\nu in position space: %e seconds\n", retime-ratime);
/***********************************************
* Fourier transform
***********************************************/
for(mu=0; mu<16; mu++) {
memcpy((void*)in, (void*)(work+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(work+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
}
fnorm = 1. / ((double)T_global * (double)(LX*LY*LZ));
if(g_cart_id==0) fprintf(stdout, "# P-fnorm = %16.5e\n", fnorm);
for(x0=0; x0<T; x0++) {
q[0] = (double)(x0+Tstart) / (double)T_global;
for(x1=0; x1<LX; x1++) {
q[1] = (double)x1 / (double)LX;
for(x2=0; x2<LY; x2++) {
q[2] = (double)x2 / (double)LY;
for(x3=0; x3<LZ; x3++) {
q[3] = (double)x3 / (double)LZ;
ix = g_ipt[x0][x1][x2][x3];
i=0;
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
iix = _GWI(i,ix,VOLUME);
w.re = cos(M_PI * (q[mu] - q[nu]));
w.im = sin(M_PI * (q[mu] - q[nu]));
work[iix ] = work[iix ] * fnorm + disc2[iix ];
work[iix+1] = work[iix+1] * fnorm + disc2[iix+1];
_co_eq_co_ti_co(&w1, (complex*)(work+iix), &w);
work[iix ] = w1.re;
work[iix+1] = w1.im;
i++;
}}
}}}}
/***********************************************
* save results
***********************************************/
sprintf(filename, "%s.%.4d.%.4d", filename_prefix2, gid, Nsave);
if(g_cart_id==0) fprintf(stdout, "# Saving results to file %s\n", filename);
sprintf(contype, "cvc-disc-P");
write_lime_contraction(work, filename, 64, 16, contype, gid, Nsave);
/*
sprintf(filename, "%sascii.%.4d.%.4d", filename_prefix2, gid, Nsave);
write_contraction(work, NULL, filename, 16, 2, 0);
*/
if(g_cart_id==0) fprintf(stdout, "# Finished working on gauge id %d\n", gid);
} /* of loop on gid */
/***********************************************
* free the allocated memory, finalize
***********************************************/
free_geometry();
fftw_free(in);
free(disc);
free(disc2);
free(work);
#ifdef MPI
fftwnd_mpi_destroy_plan(plan_p);
fftwnd_mpi_destroy_plan(plan_m);
MPI_Finalize();
#else
fftwnd_destroy_plan(plan_p);
fftwnd_destroy_plan(plan_m);
#endif
return(0);
}
void testnd_in_place(int rank, int *n, fftw_direction dir,
fftwnd_plan validated_plan,
int alternate_api, int specific, int force_buffered)
{
int istride;
int N, dim, i;
fftw_complex *in1, *in2, *out2;
fftwnd_plan p;
int flags = measure_flag | wisdom_flag | FFTW_IN_PLACE;
if (coinflip())
flags |= FFTW_THREADSAFE;
if (force_buffered)
flags |= FFTWND_FORCE_BUFFERED;
N = 1;
for (dim = 0; dim < rank; ++dim)
N *= n[dim];
in1 = (fftw_complex *) fftw_malloc(N * MAX_STRIDE * sizeof(fftw_complex));
in2 = (fftw_complex *) fftw_malloc(N * sizeof(fftw_complex));
out2 = (fftw_complex *) fftw_malloc(N * sizeof(fftw_complex));
if (!specific) {
if (alternate_api && (rank == 2 || rank == 3)) {
if (rank == 2)
p = fftw2d_create_plan(n[0], n[1], dir, flags);
else
p = fftw3d_create_plan(n[0], n[1], n[2], dir, flags);
} else /* standard api */
p = fftwnd_create_plan(rank, n, dir, flags);
} else { /* specific plan creation */
if (alternate_api && (rank == 2 || rank == 3)) {
if (rank == 2)
p = fftw2d_create_plan_specific(n[0], n[1], dir, flags,
in1, 1,
(fftw_complex *) NULL, 1);
else
p = fftw3d_create_plan_specific(n[0], n[1], n[2], dir, flags,
in1, 1,
(fftw_complex *) NULL, 1);
} else /* standard api */
p = fftwnd_create_plan_specific(rank, n, dir, flags,
in1, 1,
(fftw_complex *) NULL, 1);
}
for (istride = 1; istride <= MAX_STRIDE; ++istride) {
/*
* generate random inputs */
for (i = 0; i < N; ++i) {
int j;
c_re(in2[i]) = DRAND();
c_im(in2[i]) = DRAND();
for (j = 0; j < istride; ++j) {
c_re(in1[i * istride + j]) = c_re(in2[i]);
c_im(in1[i * istride + j]) = c_im(in2[i]);
}
}
if (istride != 1 || istride != 1 || coinflip())
fftwnd(p, istride, in1, istride, 1, (fftw_complex *) NULL, 1, 1);
else
fftwnd_one(p, in1, NULL);
fftwnd(validated_plan, 1, in2, 1, 1, out2, 1, 1);
for (i = 0; i < istride; ++i)
CHECK(compute_error_complex(in1 + i, istride, out2, 1, N) < TOLERANCE,
"testnd_in_place: wrong answer");
}
fftwnd_destroy_plan(p);
fftw_free(out2);
fftw_free(in2);
fftw_free(in1);
}
//-------------------------------------------------------------------------------
int ifdct_wrapping(int N1, int N2, int nbscales, int nbangles_coarse, int allcurvelets, vector< vector<CpxNumMat> >& c, CpxNumMat& x)
{
assert(nbscales==c.size() && nbangles_coarse==c[1].size());
//int F1 = N1/2; int F2 = N2/2;
//-------------------------------------------angles to Xhgh
vector<int> nbangles(nbscales);
vector<CpxOffMat> Xhghs; Xhghs.resize(nbscales);
if(allcurvelets==1) {
//----
nbangles[0] = 1;
for(int sc=1; sc<nbscales; sc++) nbangles[sc] = nbangles_coarse * pow2( int(ceil(double(sc-1)/2)) );
double XL1 = 4.0*N1/3.0; double XL2 = 4.0*N2/3.0;
for(int sc=nbscales-1; sc>0; sc--) {
fdct_wrapping_invsepangle(XL1, XL2, nbangles[sc], c[sc], Xhghs[sc]);
XL1 /= 2; XL2 /= 2;
}
fdct_wrapping_invwavelet(c[0], Xhghs[0]);
} else {
//----
nbangles[0] = 1;
for(int sc=1; sc<nbscales-1; sc++) nbangles[sc] = nbangles_coarse * pow2( int(ceil(double(sc-1)/2)) );
nbangles[nbscales-1] = 1;
fdct_wrapping_invwavelet(c[nbscales-1], Xhghs[nbscales-1]);
double XL1 = 2.0*N1/3.0; double XL2 = 2.0*N2/3.0;
for(int sc=nbscales-2; sc>0; sc--) {
fdct_wrapping_invsepangle(XL1, XL2, nbangles[sc], c[sc], Xhghs[sc]);
XL1 /= 2; XL2 /= 2;
}
fdct_wrapping_invwavelet(c[0], Xhghs[0]);
}
//-------------------------------------------xhghs to O
//combine
CpxOffMat X;
if(allcurvelets==1) {
double XL1 = 4.0*N1/3.0; double XL2 = 4.0*N2/3.0; //range
int XS1, XS2; int XF1, XF2; double XR1, XR2; fdct_wrapping_rangecompute(XL1, XL2, XS1, XS2, XF1, XF2, XR1, XR2);
X.resize(XS1, XS2);
} else {
X.resize(N1, N2);
}
double XL1 = 4.0*N1/3.0; double XL2 = 4.0*N2/3.0;
int XS1, XS2; int XF1, XF2; double XR1, XR2; fdct_wrapping_rangecompute(XL1, XL2, XS1, XS2, XF1, XF2, XR1, XR2);
for(int sc=nbscales-1; sc>0; sc--) {
double XL1n = XL1/2; double XL2n = XL2/2;
int XS1n, XS2n; int XF1n, XF2n; double XR1n, XR2n;
fdct_wrapping_rangecompute(XL1n, XL2n, XS1n, XS2n, XF1n, XF2n, XR1n, XR2n);
DblOffMat lowpass(XS1n, XS2n);
fdct_wrapping_lowpasscompute(XL1n, XL2n, lowpass);
DblOffMat hghpass(XS1n, XS2n);
for(int j=-XF2n; j<-XF2n+XS2n; j++)
for(int i=-XF1n; i<-XF1n+XS1n; i++)
hghpass(i,j) = sqrt(1-lowpass(i,j)*lowpass(i,j));
for(int j=-XF2n; j<-XF2n+XS2n; j++)
for(int i=-XF1n; i<-XF1n+XS1n; i++)
Xhghs[sc](i,j) *= hghpass(i,j);
for(int j=-XF2n; j<-XF2n+XS2n; j++)
for(int i=-XF1n; i<-XF1n+XS1n; i++)
Xhghs[sc-1](i,j) *= lowpass(i,j);
CpxOffMat& G = Xhghs[sc];
for(int j=G.t(); j<G.t()+G.n(); j++)
for(int i=G.s(); i<G.s()+G.m(); i++)
X(i,j) += G(i,j);
XL1 = XL1/2; XL2 = XL2/2;
fdct_wrapping_rangecompute(XL1, XL2, XS1, XS2, XF1, XF2, XR1, XR2);
}
for(int j=-XF2; j<-XF2+XS2; j++)
for(int i=-XF1; i<-XF1+XS1; i++)
X(i,j) += Xhghs[0](i,j);
// fold
CpxOffMat O(N1, N2);
if(allcurvelets==1) {
double XL1 = 4.0*N1/3.0; double XL2 = 4.0*N2/3.0;
int XS1, XS2; int XF1, XF2; double XR1, XR2; fdct_wrapping_rangecompute(XL1, XL2, XS1, XS2, XF1, XF2, XR1, XR2);
//times pou;
DblOffMat lowpass(XS1,XS2);
fdct_wrapping_lowpasscompute(XL1, XL2, lowpass);
for(int j=-XF2; j<-XF2+XS2; j++)
for(int i=-XF1; i<-XF1+XS1; i++)
X(i,j) *= lowpass(i,j);
IntOffVec t1(XS1);
for(int i=-XF1; i<-XF1+XS1; i++) if( i<-N1/2) t1(i) = i+int(N1); else if(i>(N1-1)/2) t1(i) = i-int(N1); else t1(i) = i;
IntOffVec t2(XS2);
for(int i=-XF2; i<-XF2+XS2; i++) if( i<-N2/2) t2(i) = i+int(N2); else if(i>(N2-1)/2) t2(i) = i-int(N2); else t2(i) = i;
for(int j=-XF2; j<-XF2+XS2; j++)
for(int i=-XF1; i<-XF1+XS1; i++)
O(t1(i), t2(j)) += X(i,j);
} else {
O = X;
}
//------------------------------------------------------------
CpxNumMat T(N1,N2);
fdct_wrapping_ifftshift(O, T);
fftwnd_plan p = fftw2d_create_plan(N2, N1, FFTW_BACKWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
fftwnd_one(p, (fftw_complex*)T.data(), NULL);
fftwnd_destroy_plan(p);
double sqrtprod = sqrt(double(N1*N2)); //scale
for(int i=0; i<N1; i++) for(int j=0; j<N2; j++) T(i,j) /= sqrtprod;
x = T;
//x.resize(N1, N2);
//fdct_wrapping_fftshift(T, x);
return 0;
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix;
int dxm[4], dxn[4], ixpm, ixpn;
int sid;
double *disc = (double*)NULL;
double *work = (double*)NULL;
double q[4], fnorm;
int verbose = 0;
int do_gt = 0;
char filename[100];
double ratime, retime;
double plaq, _2kappamu, hpe3_coeff, onepmutilde2, mutilde2;
double spinor1[24], spinor2[24], U_[18], U1_[18], U2_[18];
double *gauge_trafo=(double*)NULL;
complex w, w1, w2, *cp1, *cp2, *cp3;
FILE *ofs;
fftw_complex *in=(fftw_complex*)NULL;
#ifdef MPI
fftwnd_mpi_plan plan_p, plan_m;
int *status;
#else
fftwnd_plan plan_p, plan_m;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vgf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'g':
do_gt = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
set_default_input_values();
if(filename_set==0) strcpy(filename, "cvc.input");
/* read the input file */
read_input(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
/* initialize MPI parameters */
mpi_init(argc, argv);
#ifdef MPI
if((status = (int*)calloc(g_nproc, sizeof(int))) == (int*)NULL) {
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(7);
}
#endif
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifdef MPI
plan_p = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_BACKWARD, FFTW_MEASURE);
plan_m = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_FORWARD, FFTW_MEASURE);
fftwnd_mpi_local_sizes(plan_p, &T, &Tstart, &l_LX_at, &l_LXstart_at, &FFTW_LOC_VOLUME);
#else
plan_p = fftwnd_create_plan(4, dims, FFTW_BACKWARD, FFTW_MEASURE | FFTW_IN_PLACE);
plan_m = fftwnd_create_plan(4, dims, FFTW_FORWARD, FFTW_MEASURE | FFTW_IN_PLACE);
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] fftw parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(2);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
/* read the gauge field */
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
xchange_gauge();
/* measure the plaquette */
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value: %25.16e\n", plaq);
if(do_gt==1) {
/***********************************
* initialize gauge transformation
***********************************/
init_gauge_trafo(&gauge_trafo, 1.);
apply_gt_gauge(gauge_trafo);
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value after gauge trafo: %25.16e\n", plaq);
}
/****************************************
* allocate memory for the spinor fields
****************************************/
no_fields = 3;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/****************************************
* allocate memory for the contractions
****************************************/
disc = (double*)calloc( 8*VOLUME, sizeof(double));
if( disc == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
work = (double*)calloc(48*VOLUME, sizeof(double));
if( work == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for work\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
/****************************************
* prepare Fourier transformation arrays
****************************************/
in = (fftw_complex*)malloc(FFTW_LOC_VOLUME*sizeof(fftw_complex));
if(in==(fftw_complex*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(4);
}
/***********************************************
* start loop on source id.s
***********************************************/
for(sid=g_sourceid; sid<=g_sourceid2; sid++) {
/********************************
* read the first propagator
********************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid);
if(read_lime_spinor(g_spinor_field[2], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, Nconf, sid);
if(read_cmi(g_spinor_field[2], filename) != 0) break;
}
xchange_field(g_spinor_field[2]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "time to read prop.: %e seconds\n", retime-ratime);
if(do_gt==1) {
/******************************************
* gauge transform the propagators for sid
******************************************/
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_cm_ti_fv(spinor1, gauge_trafo+18*ix, g_spinor_field[2]+_GSI(ix));
_fv_eq_fv(g_spinor_field[2]+_GSI(ix), spinor1);
}
xchange_field(g_spinor_field[2]);
}
/************************************************
* calculate the source: apply Q_phi_tbc
************************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_phi_tbc(g_spinor_field[0], g_spinor_field[2]);
xchange_field(g_spinor_field[0]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to calculate source: %e seconds\n", retime-ratime);
/************************************************
* HPE: apply BH5
************************************************/
BH5(g_spinor_field[1], g_spinor_field[2]);
for(ix=0; ix<8*VOLUME; ix++) {disc[ix] = 0.;}
/* add new contractions to (existing) disc */
# ifdef MPI
ratime = MPI_Wtime();
# else
ratime = (double)clock() / CLOCKS_PER_SEC;
# endif
for(mu=0; mu<4; mu++) { /* loop on Lorentz index of the current */
iix = _GWI(mu,0,VOLUME);
for(ix=0; ix<VOLUME; ix++) { /* loop on lattice sites */
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
/* first contribution */
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
/* second contribution */
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
iix += 2;
} /* of ix */
} /* of mu */
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "[%2d] time to contract cvc: %e seconds\n", g_cart_id, retime-ratime);
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/* Fourier transform data, copy to work */
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(work+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
}
/********************************
* read the second propagator
********************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid+g_resume);
if(read_lime_spinor(g_spinor_field[2], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, Nconf, sid+g_resume);
if(read_cmi(g_spinor_field[2], filename) != 0) break;
}
xchange_field(g_spinor_field[2]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "time to read prop.: %e seconds\n", retime-ratime);
if(do_gt==1) {
/******************************************
* gauge transform the propagators for sid
******************************************/
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_cm_ti_fv(spinor1, gauge_trafo+18*ix, g_spinor_field[2]+_GSI(ix));
_fv_eq_fv(g_spinor_field[2]+_GSI(ix), spinor1);
}
xchange_field(g_spinor_field[2]);
}
/************************************************
* calculate the source: apply Q_phi_tbc
************************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_phi_tbc(g_spinor_field[0], g_spinor_field[2]);
xchange_field(g_spinor_field[0]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to calculate source: %e seconds\n", retime-ratime);
/************************************************
* HPE: apply BH5
************************************************/
BH5(g_spinor_field[1], g_spinor_field[2]);
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
/* add new contractions to (existing) disc */
# ifdef MPI
ratime = MPI_Wtime();
# else
ratime = (double)clock() / CLOCKS_PER_SEC;
# endif
for(mu=0; mu<4; mu++) { /* loop on Lorentz index of the current */
iix = _GWI(mu,0,VOLUME);
for(ix=0; ix<VOLUME; ix++) { /* loop on lattice sites */
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
/* first contribution */
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
/* second contribution */
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
iix += 2;
} /* of ix */
} /* of mu */
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "[%2d] time to contract cvc: %e seconds\n", g_cart_id, retime-ratime);
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/* Fourier transform data, copy to work */
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
memcpy((void*)(work+_GWI(4+mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
}
fnorm = 1. / ((double)(T_global*LX*LY*LZ));
fprintf(stdout, "fnorm = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(work+_GWI(mu,0,VOLUME));
cp2 = (complex*)(work+_GWI(4+nu,0,VOLUME));
cp3 = (complex*)(work+_GWI(8+4*mu+nu,0,VOLUME));
for(x0=0; x0<T; x0++) {
q[0] = (double)(x0+Tstart) / (double)T_global;
for(x1=0; x1<LX; x1++) {
q[1] = (double)(x1) / (double)LX;
for(x2=0; x2<LY; x2++) {
q[2] = (double)(x2) / (double)LY;
for(x3=0; x3<LZ; x3++) {
q[3] = (double)(x3) / (double)LZ;
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI * (q[mu]-q[nu]) );
w.im = sin( M_PI * (q[mu]-q[nu]) );
_co_eq_co_ti_co(&w1, cp1, cp2);
_co_eq_co_ti_co(cp3, &w1, &w);
_co_ti_eq_re(cp3, fnorm);
cp1++; cp2++; cp3++;
}
}
}
}
}
}
/* save the result in momentum space */
sprintf(filename, "cvc_hpe5_ft.%.4d.%.2d", Nconf, sid);
write_contraction(work+_GWI(8,0,VOLUME), NULL, filename, 16, 0, 0);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to save cvc results: %e seconds\n", retime-ratime);
} /* of loop on sid */
/***********************************************
* free the allocated memory, finalize
***********************************************/
free(g_gauge_field);
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
free_geometry();
fftw_free(in);
free(disc);
free(work);
#ifdef MPI
fftwnd_mpi_destroy_plan(plan_p);
fftwnd_mpi_destroy_plan(plan_m);
free(status);
MPI_Finalize();
#else
fftwnd_destroy_plan(plan_p);
fftwnd_destroy_plan(plan_m);
#endif
return(0);
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix;
int sid, status;
double *disc = (double*)NULL;
double *data = (double*)NULL;
double *bias = (double*)NULL;
double *work = (double*)NULL;
double q[4], fnorm;
int verbose = 0;
char filename[100], contype[200];
double ratime, retime;
double plaq;
double spinor1[24], spinor2[24], U_[18];
complex w, w1, *cp1, *cp2, *cp3, *cp4;
fftw_complex *in=(fftw_complex*)NULL;
#ifdef MPI
fftwnd_mpi_plan plan_p, plan_m;
#else
fftwnd_plan plan_p, plan_m;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
if(filename_set==0) strcpy(filename, "cvc.input");
fprintf(stdout, "# Reading input from file %s\n", filename);
read_input_parser(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa <= 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.\n");
usage();
}
if(hpe_order%2==0 && hpe_order>0) {
if(g_proc_id==0) fprintf(stdout, "HPE order should be odd\n");
usage();
}
fprintf(stdout, "\n**************************************************\n"\
"* vp_disc_hpe_stoch_subtract with HPE of order %d\n"\
"**************************************************\n\n", hpe_order);
/*********************************
* initialize MPI parameters
*********************************/
mpi_init(argc, argv);
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifdef MPI
plan_p = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_BACKWARD, FFTW_MEASURE);
plan_m = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_FORWARD, FFTW_MEASURE);
fftwnd_mpi_local_sizes(plan_p, &T, &Tstart, &l_LX_at, &l_LXstart_at, &FFTW_LOC_VOLUME);
#else
plan_p = fftwnd_create_plan(4, dims, FFTW_BACKWARD, FFTW_MEASURE | FFTW_IN_PLACE);
plan_m = fftwnd_create_plan(4, dims, FFTW_FORWARD, FFTW_MEASURE | FFTW_IN_PLACE);
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] fftw parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(101);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(102);
}
geometry();
/************************************************
* read the gauge field, measure the plaquette
************************************************/
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "# reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
xchange_gauge();
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "# measured plaquette value: %25.16e\n", plaq);
/****************************************
* allocate memory for the spinor fields
****************************************/
no_fields = 3;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/****************************************
* allocate memory for the contractions
****************************************/
disc = (double*)calloc(16*VOLUME, sizeof(double));
if( disc == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(103);
}
data = (double*)calloc(16*VOLUME, sizeof(double));
if( data== (double*)NULL ) {
fprintf(stderr, "could not allocate memory for data\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(104);
}
for(ix=0; ix<16*VOLUME; ix++) data[ix] = 0.;
work = (double*)calloc(32*VOLUME, sizeof(double));
if( work == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for work\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(105);
}
bias = (double*)calloc(32*VOLUME, sizeof(double));
if( bias == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for bias\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(106);
}
for(ix=0; ix<32*VOLUME; ix++) bias[ix] = 0.;
/****************************************
* prepare Fourier transformation arrays
****************************************/
in = (fftw_complex*)malloc(FFTW_LOC_VOLUME*sizeof(fftw_complex));
if(in==(fftw_complex*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(107);
}
/***********************************************
* start loop on source id.s
***********************************************/
for(sid=g_sourceid; sid<=g_sourceid2; sid+=g_sourceid_step) {
for(ix=0; ix<16*VOLUME; ix++) disc[ix] = 0.;
/* read the new propagator */
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid);
if(read_lime_spinor(g_spinor_field[2], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, Nconf, sid);
if(read_cmi(g_spinor_field[2], filename) != 0) break;
}
xchange_field(g_spinor_field[2]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to read prop.: %e seconds\n", retime-ratime);
count++;
/************************************************
* calculate the source: apply Q_phi_tbc
************************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_phi_tbc(g_spinor_field[0], g_spinor_field[2]);
xchange_field(g_spinor_field[0]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to calculate source: %e seconds\n", retime-ratime);
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/************************************************
* HPE: apply BH to order hpe_order+2
************************************************/
if(hpe_order>0) {
BHn(g_spinor_field[1], g_spinor_field[2], hpe_order+2);
} else {
memcpy((void*)g_spinor_field[1], (void*)g_spinor_field[2], 24*VOLUMEPLUSRAND*sizeof(double));
}
/************************************************
* add new contractions to (existing) disc
************************************************/
for(mu=0; mu<4; mu++) {
iix = _GWI(mu,0,VOLUME);
for(ix=0; ix<VOLUME; ix++) {
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[iix ] = -0.5 * w.re;
disc[iix+1] = -0.5 * w.im;
data[iix ] -= 0.5 * w.re;
data[iix+1] -= 0.5 * w.im;
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
data[iix ] -= 0.5 * w.re;
data[iix+1] -= 0.5 * w.im;
iix += 2;
}
}
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to contract cvc: %e seconds\n", retime-ratime);
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
memcpy((void*)(disc+_GWI(4+mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(disc+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
} /* of mu =0 ,..., 3*/
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(disc+_GWI(mu, 0,VOLUME));
cp2 = (complex*)(disc+_GWI(4+nu, 0,VOLUME));
cp3 = (complex*)(bias+_GWI(4*mu+nu,0,VOLUME));
for(ix=0; ix<VOLUME; ix++) {
_co_eq_co_ti_co(&w1, cp1, cp2);
cp3->re += w1.re;
cp3->im += w1.im;
cp1++; cp2++; cp3++;
}
}}
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time for Fourier trafo and adding to bias: %e seconds\n",
retime-ratime);
} /* of loop on sid */
/************************************************
* save results for count == Nsave
************************************************/
if(count==Nsave) {
if(g_cart_id == 0) fprintf(stdout, "# save results for count = %d\n", count);
for(ix=0; ix<16*VOLUME; ix++) disc[ix] = 0.;
if(hpe_order>0) {
sprintf(filename, "vp_disc_hpe%.2d_loops_X.%.4d", hpe_order, Nconf);
if(g_cart_id==0) fprintf(stdout, "# reading loop part from file %s\n", filename);
if( (status = read_lime_contraction(disc, filename, 4, 0)) != 0 ) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(108);
}
}
/* save the result in position space */
fnorm = 1. / ( (double)count * g_prop_normsqr );
if(g_cart_id==0) fprintf(stdout, "# X-fnorm = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(ix=0; ix<VOLUME; ix++) {
work[_GWI(mu,ix,VOLUME) ] = data[_GWI(mu,ix,VOLUME) ] * fnorm + disc[_GWI(mu,ix,VOLUME) ];
work[_GWI(mu,ix,VOLUME)+1] = data[_GWI(mu,ix,VOLUME)+1] * fnorm + disc[_GWI(mu,ix,VOLUME)+1];
}
}
sprintf(filename, "vp_disc_hpe%.2d_subtracted_X.%.4d.%.4d", hpe_order, Nconf, count);
sprintf(contype, "cvc-disc-hpe-loops-%2d-to-%2d-stoch-subtracted-X", hpe_order, hpe_order+2);
write_lime_contraction(work, filename, 64, 4, contype, Nconf, count);
/*
sprintf(filename, "vp_disc_hpe%.2d_subtracted_X.%.4d.%.4d.ascii", hpe_order, Nconf, count);
write_contraction(work, NULL, filename, 4, 2, 0);
*/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(data+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
memcpy((void*)(data+_GWI(4+mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
memcpy((void*)in, (void*)(data+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(data+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
}
fnorm = 1. / ( g_prop_normsqr*g_prop_normsqr * (double)count * (double)(count-1) );
if(g_cart_id==0) fprintf(stdout, "# P-fnorm for purely stochastic part = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(data+_GWI(mu, 0,VOLUME));
cp2 = (complex*)(data+_GWI(4+nu, 0,VOLUME));
cp3 = (complex*)(work+_GWI(4*mu+nu,0,VOLUME));
cp4 = (complex*)(bias+_GWI(4*mu+nu,0,VOLUME));
for(ix=0; ix<VOLUME; ix++) {
_co_eq_co_ti_co(&w1, cp1, cp2);
cp3->re = ( w1.re - cp4->re ) * fnorm;
cp3->im = ( w1.im - cp4->im ) * fnorm;
cp1++; cp2++; cp3++; cp4++;
}
}}
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
memcpy((void*)(disc+_GWI(4+mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(disc+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
}
fnorm = 1. / ( g_prop_normsqr * (double)count );
if(g_cart_id==0) fprintf(stdout, "# P-fnorm for mixed stochastic-loop part = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(data + _GWI(mu, 0,VOLUME));
cp2 = (complex*)(disc + _GWI(4+nu, 0,VOLUME));
cp3 = (complex*)(work + _GWI(4*mu+nu,0,VOLUME));
for(ix=0; ix<VOLUME; ix++) {
_co_eq_co_ti_co(&w1, cp1, cp2);
cp3->re += w1.re * fnorm;
cp3->im += w1.im * fnorm;
cp1++; cp2++; cp3++;
}
cp1 = (complex*)(disc + _GWI(mu, 0,VOLUME));
cp2 = (complex*)(data + _GWI(4+nu, 0,VOLUME));
cp3 = (complex*)(work + _GWI(4*mu+nu,0,VOLUME));
for(ix=0; ix<VOLUME; ix++) {
_co_eq_co_ti_co(&w1, cp1, cp2);
cp3->re += w1.re * fnorm;
cp3->im += w1.im * fnorm;
cp1++; cp2++; cp3++;
}
}}
fnorm = 1. / ( (double)T_global * (double)(LX*LY*LZ) );
if(g_cart_id==0) fprintf(stdout, "# P-fnorm for final estimator (1/T/V) = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(disc + _GWI(mu, 0,VOLUME));
cp2 = (complex*)(disc + _GWI(4+nu, 0,VOLUME));
cp3 = (complex*)(work + _GWI(4*mu+nu,0,VOLUME));
for(x0=0; x0<T; x0++) {
q[0] = (double)(x0+Tstart) / (double)T_global;
for(x1=0; x1<LX; x1++) {
q[1] = (double)x1 / (double)LX;
for(x2=0; x2<LY; x2++) {
q[2] = (double)x2 / (double)LY;
for(x3=0; x3<LZ; x3++) {
q[3] = (double)x3 / (double)LZ;
ix = g_ipt[x0][x1][x2][x3];
w.re = cos(M_PI * ( q[mu] - q[nu] ) );
w.im = sin(M_PI * ( q[mu] - q[nu] ) );
_co_eq_co_ti_co(&w1, cp1, cp2);
cp3->re += w1.re;
cp3->im += w1.im;
_co_eq_co_ti_co(&w1, cp3, &w);
cp3->re = w1.re * fnorm;
cp3->im = w1.im * fnorm;
cp1++; cp2++; cp3++;
}}}}
}}
sprintf(filename, "vp_disc_hpe%.2d_subtracted_P.%.4d.%.4d", hpe_order, Nconf, count);
sprintf(contype, "cvc-disc-hpe-loops-%2d-to-%2d-stoch-subtracted-P", hpe_order, hpe_order+2);
write_lime_contraction(work, filename, 64, 16, contype, Nconf, count);
/*
sprintf(filename, "vp_disc_hpe%.2d_subtracted_P.%.4d.%.4d.ascii", hpe_order, Nconf, count);
write_contraction(work, NULL, filename, 16, 2, 0);
*/
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to save cvc results: %e seconds\n", retime-ratime);
} /* of if count == Nsave */
/***********************************************
* free the allocated memory, finalize
***********************************************/
free(g_gauge_field);
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
free_geometry();
fftw_free(in);
free(disc);
free(bias);
free(data);
free(work);
#ifdef MPI
fftwnd_mpi_destroy_plan(plan_p);
fftwnd_mpi_destroy_plan(plan_m);
MPI_Finalize();
#else
fftwnd_destroy_plan(plan_p);
fftwnd_destroy_plan(plan_m);
#endif
return(0);
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix;
int sx0, sx1, sx2, sx3;
int sid;
double *disc = (double*)NULL;
double *disc2 = (double*)NULL;
double *work = (double*)NULL;
double q[4], fnorm;
double cvc_lnuy[8];
double *gauge_trafo=(double*)NULL;
double unit_trace[2], D_trace[2];
int verbose = 0;
int do_gt = 0;
char filename[100];
double ratime, retime;
double plaq;
double spinor1[24], spinor2[24], U_[18];
complex w, w1, *cp1, *cp2, *cp3;
FILE *ofs;
fftw_complex *in=(fftw_complex*)NULL;
#ifdef MPI
fftwnd_mpi_plan plan_p;
int *status;
#else
fftwnd_plan plan_p;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vgf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'g':
do_gt = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
set_default_input_values();
if(filename_set==0) strcpy(filename, "cvc.input");
/* read the input file */
read_input(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
/* initialize MPI parameters */
mpi_init(argc, argv);
#ifdef MPI
if((status = (int*)calloc(g_nproc, sizeof(int))) == (int*)NULL) {
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(7);
}
#endif
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifdef MPI
plan_p = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_BACKWARD, FFTW_MEASURE);
fftwnd_mpi_local_sizes(plan_p, &T, &Tstart, &l_LX_at, &l_LXstart_at, &FFTW_LOC_VOLUME);
#else
plan_p = fftwnd_create_plan(4, dims, FFTW_BACKWARD, FFTW_MEASURE | FFTW_IN_PLACE);
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] fftw parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(2);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
/* read the gauge field */
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
xchange_gauge();
/* measure the plaquette */
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value: %25.16e\n", plaq);
/* get the source location coordinates */
sx0 = g_source_location / (LX*LY*LZ );
sx1 = ( g_source_location % (LX*LY*LZ) ) / (LY*LZ);
sx2 = ( g_source_location % (LY*LZ) ) / LZ;
sx3 = ( g_source_location % LZ );
/* read the data for lnuy */
sprintf(filename, "cvc_lnuy_X.%.4d", Nconf);
ofs = fopen(filename, "r");
fprintf(stdout, "reading cvc lnuy from file %s\n", filename);
for(mu=0; mu<4; mu++) {
fscanf(ofs, "%lf%lf", cvc_lnuy+2*mu, cvc_lnuy+2*mu+1);
}
fclose(ofs);
/* allocate memory for the spinor fields */
no_fields = 2;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/****************************************
* allocate memory for the contractions
****************************************/
disc = (double*)calloc(8*VOLUME, sizeof(double));
if( disc == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
disc2 = (double*)calloc(8*VOLUME, sizeof(double));
if( disc2 == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc2\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc2[ix] = 0.;
work = (double*)calloc(48*VOLUME, sizeof(double));
if( work == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for work\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(3);
}
/****************************************
* prepare Fourier transformation arrays
****************************************/
in = (fftw_complex*)malloc(FFTW_LOC_VOLUME*sizeof(fftw_complex));
if(in==(fftw_complex*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(4);
}
if(g_resume==1) { /* read current disc from file */
sprintf(filename, ".outcvc_current.%.4d", Nconf);
c = read_contraction(disc, &count, filename, 8);
#ifdef MPI
MPI_Gather(&c, 1, MPI_INT, status, 1, MPI_INT, 0, g_cart_grid);
if(g_cart_id==0) {
/* check the entries in status */
for(i=0; i<g_nproc; i++)
if(status[i]!=0) { status[0] = 1; break; }
}
MPI_Bcast(status, 1, MPI_INT, 0, g_cart_grid);
if(status[0]==1) {
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
count = 0;
}
#else
if(c != 0) {
fprintf(stdout, "could not read current disc; start new\n");
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
count = 0;
}
#endif
if(g_cart_id==0) fprintf(stdout, "starting with count = %d\n", count);
} /* of g_resume == 1 */
if(do_gt==1) {
/***********************************
* initialize gauge transformation
***********************************/
init_gauge_trafo(&gauge_trafo,1.0);
fprintf(stdout, "applying gauge trafo to gauge field\n");
apply_gt_gauge(gauge_trafo);
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "plaquette plaq = %25.16e\n", plaq);
}
unit_trace[0] = 0.;
unit_trace[1] = 0.;
D_trace[0] = 0.;
D_trace[1] = 0.;
/****************************************
* start loop on source id.s
****************************************/
for(sid=g_sourceid; sid<=g_sourceid2; sid++) {
/****************************************
* read the new propagator
****************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/****************************************
* check: write source before D-appl.
****************************************/
/*
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d", filename_prefix, Nconf, sid);
read_lime_spinor(g_spinor_field[0], filename, 0);
}
for(ix=0; ix<12*VOLUME; ix++) {
fprintf(stdout, "source: %6d%25.16e%25.16e\n", ix, g_spinor_field[0][2*ix], g_spinor_field[0][2*ix+1]);
}
*/
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid);
/* sprintf(filename, "%s.%.4d.%.2d", filename_prefix, Nconf, sid); */
if(read_lime_spinor(g_spinor_field[1], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, Nconf, sid);
if(read_cmi(g_spinor_field[1], filename) != 0) break;
}
xchange_field(g_spinor_field[1]);
if(do_gt==1) {
fprintf(stdout, "applying gt on propagators\n");
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_cm_ti_fv(spinor1, gauge_trafo+18*ix, g_spinor_field[1]+_GSI(ix));
_fv_eq_fv(g_spinor_field[1]+_GSI(ix), spinor1);
}
}
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "time to read prop.: %e seconds\n", retime-ratime);
count++;
/****************************************
* calculate the source: apply Q_phi_tbc
****************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_phi_tbc(g_spinor_field[0], g_spinor_field[1]);
xchange_field(g_spinor_field[0]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to calculate source: %e seconds\n", retime-ratime);
/****************************************
* check: write source after D-appl.
****************************************/
/*
for(ix=0; ix<12*VOLUME; ix++) {
fprintf(stdout, "D_source: %6d%25.16e%25.16e\n", ix, g_spinor_field[0][2*ix], g_spinor_field[0][2*ix+1]);
}
*/
/****************************************
* add new contractions to (existing) disc
****************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
for(mu=0; mu<4; mu++) { /* loop on Lorentz index of the current */
iix = _GWI(mu,0,VOLUME);
for(ix=0; ix<VOLUME; ix++) { /* loop on lattice sites */
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
/* first contribution */
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
/* second contribution */
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc2[iix ] -= 0.5 * w.re;
disc2[iix+1] -= 0.5 * w.im;
iix += 2;
} /* of ix */
} /* of mu */
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "[%2d] contractions for CVC in %e seconds\n", g_cart_id, retime-ratime);
/***************************************************
* check: convergence of trace of unit matrix
***************************************************/
_co_eq_fv_dag_ti_fv(&w, g_spinor_field[0]+_GSI(g_source_location), g_spinor_field[0]+_GSI(g_source_location));
unit_trace[0] += w.re;
unit_trace[1] += w.im;
fprintf(stdout, "unit_trace: %4d%25.16e%25.16e\n", count, w.re, w.im);
_co_eq_fv_dag_ti_fv(&w, g_spinor_field[0]+_GSI(g_source_location), g_spinor_field[0]+_GSI(g_iup[g_source_location][0]));
fprintf(stdout, "shift_trace: %4d%25.16e%25.16e\n", count, w.re, w.im);
/***************************************************
* check: convergence of trace D_u(source_location, source_location)
***************************************************/
Q_phi_tbc(g_spinor_field[1], g_spinor_field[0]);
_co_eq_fv_dag_ti_fv(&w, g_spinor_field[0]+_GSI(g_source_location), g_spinor_field[1]+_GSI(g_source_location));
D_trace[0] += w.re;
D_trace[1] += w.im;
/* fprintf(stdout, "D_trace: %4d%25.16e%25.16e\n", count, D_trace[0]/(double)count, D_trace[1]/(double)count); */
fprintf(stdout, "D_trace: %4d%25.16e%25.16e\n", count, w.re, w.im);
/***************************************************
* save results for count = multiple of Nsave
***************************************************/
if(count%Nsave == 0) {
if(g_cart_id == 0) fprintf(stdout, "save results for count = %d\n", count);
/* save the result in position space */
/* divide by number of propagators */
for(ix=0; ix<8*VOLUME; ix++) work[ix] = disc[ix] / (double)count;
sprintf(filename, "outcvc_Xm.%.4d.%.4d", Nconf, count);
write_contraction(work, NULL, filename, 4, 2, 0);
for(ix=0; ix<8*VOLUME; ix++) work[ix] = disc2[ix] / (double)count;
sprintf(filename, "outcvc_Xp.%.4d.%.4d", Nconf, count);
write_contraction(work, NULL, filename, 4, 2, 0);
for(ix=0; ix<8*VOLUME; ix++) work[ix] = (disc[ix] + disc2[ix]) / (double)count;
sprintf(filename, "outcvc_X.%.4d.%.4d", Nconf, count);
write_contraction(work, NULL, filename, 4, 2, 0);
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/****************************************
* Fourier transform data, copy to work
****************************************/
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(work+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(work+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
} /* of mu =0 ,..., 3*/
/* fnorm = 1. / ((double)count); */
fprintf(stdout, "fnorm = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(work+_GWI(mu,0,VOLUME));
cp2 = (complex*)(cvc_lnuy+2*nu);
cp3 = (complex*)(work+_GWI(4+4*mu+nu,0,VOLUME));
for(x0=0; x0<T; x0++) {
q[0] = (double)(x0+Tstart) / (double)T_global;
for(x1=0; x1<LX; x1++) {
q[1] = (double)(x1) / (double)LX;
for(x2=0; x2<LY; x2++) {
q[2] = (double)(x2) / (double)LY;
for(x3=0; x3<LZ; x3++) {
q[3] = (double)(x3) / (double)LZ;
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI * ( q[mu] - q[nu] - 2.*(sx0*q[0]+sx1*q[1]+sx2*q[2]+sx3*q[3])) );
w.im = sin( M_PI * ( q[mu] - q[nu] - 2.*(sx0*q[0]+sx1*q[1]+sx2*q[2]+sx3*q[3])) );
/* fprintf(stdout, "mu=%3d, nu=%3d, t=%3d, x=%3d, y=%3d, z=%3d, phase= %21.12e + %21.12ei\n", \
mu, nu, x0, x1, x2, x3, w.re, w.im); */
_co_eq_co_ti_co(&w1, cp1, cp2);
_co_eq_co_ti_co(cp3, &w1, &w);
/* _co_ti_eq_re(cp3, fnorm); */
cp1++; cp3++;
}
}
}
}
}
}
/* save the result in momentum space */
sprintf(filename, "outcvc_P.%.4d.%.4d", Nconf, count);
write_contraction(work+_GWI(4,0,VOLUME), NULL, filename, 16, 2, 0);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to cvc save results: %e seconds\n", retime-ratime);
} /* of count % Nsave == 0 */
} /* of loop on sid */
if(g_resume==1) {
/* write current disc to file */
sprintf(filename, ".outcvc_current.%.4d", Nconf);
write_contraction(disc, &count, filename, 4, 0, 0);
}
/**************************************
* free the allocated memory, finalize
**************************************/
free(g_gauge_field);
if(do_gt==1) free(gauge_trafo);
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
free_geometry();
fftw_free(in);
free(disc); free(disc2);
free(work);
#ifdef MPI
fftwnd_mpi_destroy_plan(plan_p);
free(status);
MPI_Finalize();
#else
fftwnd_destroy_plan(plan_p);
#endif
return(0);
}
static void TRAN_FFT_Electrode_Grid(MPI_Comm comm1, int isign)
/* #define grid_e_ref(i,j,k) ( (i)*Ngrid2*(l3[1]-l3[0]+1)+ (j)*(l3[1]-l3[0]+1) + (k)-l3[0] )
*#define fft2d_ref(i,j) ( (i)*Ngrid2+(j) )
*/
{
int side;
int i,j,k;
int l1[2];
#ifdef fftw2
fftwnd_plan p;
#else
fftw_plan p;
#endif
fftw_complex *in,*out;
double factor;
int myid;
MPI_Comm_rank(comm1,&myid);
if (print_stdout){
printf("TRAN_FFT_Electrode_Grid in\n");
}
/* allocation
* ElectrodedVHart_Grid_c is a global variable
* do not free these
*/
l1[0]= 0;
l1[1]= TRAN_grid_bound[0];
ElectrodedVHart_Grid_c[0]=(dcomplex*)malloc(sizeof(dcomplex)*Ngrid3*Ngrid2*(l1[1]-l1[0]+1) );
/* ElectrodedVHart_Grid_c = Vh_electrode(kx,ky, z=[0:TRAN_grid_bound[0]]), TRAN_grid_bound: integer */
l1[0]= TRAN_grid_bound[1];
l1[1]= Ngrid1-1;
ElectrodedVHart_Grid_c[1]=(dcomplex*)malloc(sizeof(dcomplex)*Ngrid3*Ngrid2*(l1[1]-l1[0]+1) );
/* ElectrodedVHart_Grid_c = Vh_electrode(kx,ky, z=[TRAN_grid_bound[1]:Ngrid3-1]), TRAN_grid_bound: integer */
/* allocation for fft ,
* free these at last
*/
#ifdef fftw2
in = (fftw_complex*)malloc(sizeof(fftw_complex)*Ngrid3*Ngrid2);
out = (fftw_complex*)malloc(sizeof(fftw_complex)*Ngrid3*Ngrid2);
#else
in = fftw_malloc(sizeof(fftw_complex)*Ngrid3*Ngrid2);
out = fftw_malloc(sizeof(fftw_complex)*Ngrid3*Ngrid2);
#endif
#ifdef fftw2
p=fftw2d_create_plan(Ngrid2,Ngrid3,isign,FFTW_ESTIMATE);
#else
p=fftw_plan_dft_2d(Ngrid2,Ngrid3,in,out,isign,FFTW_ESTIMATE);
#endif
/* left side */
side=0;
l1[0]= 0;
l1[1]= TRAN_grid_bound[0];
factor = 1.0/( (double)(Ngrid2*Ngrid3) ) ;
#define grid_e_ref(i,j,k) ( ( (i)-l1[0])*Ngrid2*Ngrid3+(j)*Ngrid3+(k) )
#define fft2d_ref(j,k) ( (j)*Ngrid3+ (k) )
for (i=l1[0];i<=l1[1];i++) {
for (j=0;j<Ngrid2;j++){
for (k=0;k<Ngrid3;k++) {
#ifdef fftw2
c_re(in[fft2d_ref(j,k)]) = ElectrodedVHart_Grid[side][grid_e_ref(i,j,k)];
c_im(in[fft2d_ref(j,k)]) = 0.0;
#else
in[fft2d_ref(j,k)][0]= ElectrodedVHart_Grid[side][grid_e_ref(i,j,k)];
in[fft2d_ref(j,k)][1]= 0.0;
#endif
}
}
#ifdef fftw2
fftwnd_one(p, in, out);
#else
fftw_execute(p);
#endif
for (j=0;j<Ngrid2;j++) {
for (k=0;k<Ngrid3;k++) {
#ifdef fftw2
ElectrodedVHart_Grid_c[side][grid_e_ref(i,j,k)].r = c_re(out[fft2d_ref(j,k)])*factor;
ElectrodedVHart_Grid_c[side][grid_e_ref(i,j,k)].i = c_im(out[fft2d_ref(j,k)])*factor;
#else
ElectrodedVHart_Grid_c[side][grid_e_ref(i,j,k)].r = out[fft2d_ref(j,k)][0]*factor;
ElectrodedVHart_Grid_c[side][grid_e_ref(i,j,k)].i = out[fft2d_ref(j,k)][1]*factor;
#endif
}
}
} /* k */
#ifdef DEBUG
/*debug*/
{ char name[100];int i; double R[4]; for(i=1;i<=3;i++) R[i]=0.0;
sprintf(name,"ElectrodedVHart_Grid_c_lr.%d",myid);
TRAN_Print_Grid_c(name,"ElectrodedVHart_Grid_c_li",
Grid_Origin, gtv, l1[1]-l1[0]+1,Ngrid2, 0, Ngrid3-1, R, ElectrodedVHart_Grid_c[side]);
}
#endif
/* right side */
side=1;
l1[0]= TRAN_grid_bound[1];
l1[1]= Ngrid1-1;
factor = 1.0/( (double) Ngrid2*Ngrid3 );
for (i=l1[0];i<=l1[1];i++) {
for (j=0;j<Ngrid2;j++) {
for (k=0;k<Ngrid3;k++) {
#ifdef fftw2
c_re(in[fft2d_ref(j,k)]) = ElectrodedVHart_Grid[side][grid_e_ref(i,j,k)];
c_im(in[fft2d_ref(j,k)]) = 0.0;
#else
in[fft2d_ref(j,k)][0]= ElectrodedVHart_Grid[side][grid_e_ref(i,j,k)];
in[fft2d_ref(j,k)][1]= 0.0;
#endif
}
}
#ifdef fftw2
fftwnd_one(p, in, out);
#else
fftw_execute(p);
#endif
for (j=0;j<Ngrid2;j++) {
for (k=0;k<Ngrid3;k++) {
#ifdef fftw2
ElectrodedVHart_Grid_c[side][grid_e_ref(i,j,k)].r = c_re(out[fft2d_ref(j,k)])*factor;
ElectrodedVHart_Grid_c[side][grid_e_ref(i,j,k)].i = c_im(out[fft2d_ref(j,k)])*factor;
#else
ElectrodedVHart_Grid_c[side][grid_e_ref(i,j,k)].r = out[fft2d_ref(j,k)][0]*factor;
ElectrodedVHart_Grid_c[side][grid_e_ref(i,j,k)].i = out[fft2d_ref(j,k)][1]*factor;
#endif
}
}
} /* k */
#ifdef DEBUG
/*debug*/
{ char name[100];int i; double R[4]; for(i=1;i<=3;i++) R[i]=0.0;
sprintf(name,"ElectrodedVHart_Grid_c_rr.%d",myid);
TRAN_Print_Grid_c(name,"ElectrodedVHart_Grid_c_ri",
Grid_Origin, gtv,l1[1]-l1[0]+1,Ngrid2, 0, Ngrid3-1, R, ElectrodedVHart_Grid_c[side]);
}
#endif
#ifdef fftw2
fftwnd_destroy_plan(p);
#else
fftw_destroy_plan(p);
#endif
#ifdef fftw2
free(out);
free(in);
#else
fftw_free(out);
fftw_free(in);
#endif
if (print_stdout){
printf("TRAN_FFT_Electrode_Grid out\n");
}
}
void F77_FUNC_(fftwnd_f77_one,FFTWND_F77_ONE)
(fftwnd_plan *p, fftw_complex *in, fftw_complex *out)
{
fftwnd_one(*p,in,out);
}
//-----------------------------------------------------------------------
int fdct_wrapping_sepangle(double XL1, double XL2, int nbangle, CpxOffMat& Xhgh, vector<CpxNumMat>& csc)
{
//WEDGE ORDERING: from -45 degree, counter-clockwise
typedef pair<int,int> intpair;
map<intpair, fftwnd_plan> planmap;
int nbquadrants = 4;
int nd = nbangle / 4;
int wcnt = 0;
//backup
CpxOffMat Xhghb(Xhgh);
double XL1b = XL1; double XL2b = XL2;
int qvec[] = {2,1,0,3};
for(int qi=0; qi<nbquadrants; qi++) {
int q = qvec[qi];
//ROTATE data to its right position
fdct_wrapping_rotate_forward(q, XL1b, XL2b, XL1, XL2); XL1 = abs(XL1); XL2 = abs(XL2);
fdct_wrapping_rotate_forward(q, Xhghb, Xhgh);
//figure out XS, XF, XR
double XW1 = XL1/nd; double XW2 = XL2/nd;
int XS1, XS2; int XF1, XF2; double XR1, XR2; fdct_wrapping_rangecompute(XL1, XL2, XS1, XS2, XF1, XF2, XR1, XR2);
for(int w=nd-1; w>=0; w--) {
double xs = XR1/4 - (XW1/2)/4;
double xe = XR1;
double ys = -XR2 + (w-0.5)*XW2;
double ye = -XR2 + (w+1.5)*XW2; //x range
int xn = int(ceil(xe-xs)); int yn = int(ceil(ye-ys));
//MAKE THEM ODD
if(xn%2==0) xn++; if(yn%2==0) yn++;
int xf = int(ceil(xs)); //int yf = int(ceil(ys));
//theta
double thts, thtm, thte; //y direction
if(w==0) {
thts = atan2(-1.0, 1.0-1.0/nd);
thtm = atan2(-1.0+1.0/nd, 1.0);
thte = atan2(-1.0+3.0/nd, 1.0);
} else if(w==nd-1) {
thts = atan2(-1.0+(2.0*w-1.0)/nd, 1.0);
thtm = atan2(-1.0+(2.0*w+1.0)/nd, 1.0);
thte = atan2(1.0, 1.0-1.0/nd);
} else {
thts = atan2(-1.0+(2.0*w-1.0)/nd, 1.0);
thtm = atan2(-1.0+(2.0*w+1.0)/nd, 1.0);
thte = atan2(-1.0+(2.0*w+3.0)/nd, 1.0);
}
//wrapping
int xh = xn/2; int yh = yn/2; //half length
double R21 = XR2/XR1; //ratio
CpxOffMat wpdata(xn,yn);
for(int xcur=xf; xcur<xe; xcur++) { //for each layer
int yfm = (int)ceil( max(-XR2, R21*xcur*tan(thts)) );
int yto = (int)floor( min(XR2, R21*xcur*tan(thte)) );
for(int ycur=yfm; ycur<=yto; ycur++) {
int tmpx = xcur%xn; if(tmpx<-xh) tmpx+=xn; if(tmpx>=-xh+xn) tmpx-=xn;
int tmpy = ycur%yn; if(tmpy<-yh) tmpy+=yn; if(tmpy>=-yh+yn) tmpy-=yn;
wpdata(tmpx,tmpy) = Xhgh(xcur,ycur);
//partition of unity
double thtcur = atan2(ycur/XR2, xcur/XR1);
double wtht;
if(thtcur<thtm) {
double l,r; fdct_wrapping_window((thtcur-thts)/(thtm-thts), l, r);
wtht = l;
} else {
double l,r; fdct_wrapping_window((thtcur-thtm)/(thte-thtm), l, r);
wtht = r;
}
double pou = wtht;
wpdata(tmpx,tmpy) *= pou;
}
}
//IFFT
{
//rotate backward
CpxOffMat rpdata;
fdct_wrapping_rotate_backward(q, wpdata, rpdata);
//ifftshift
int xn = rpdata.m(); int yn = rpdata.n(); //reset xn, yn
CpxNumMat tpdata(xn,yn);
fdct_wrapping_ifftshift(rpdata, tpdata);
//ifft
fftwnd_plan p = NULL;
map<intpair,fftwnd_plan>::iterator mit=planmap.find( intpair(xn,yn) );
if(mit!=planmap.end()) {
p = (*mit).second;
} else {
p = fftw2d_create_plan(yn, xn, FFTW_BACKWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
planmap[ intpair(xn, yn) ] = p;
}
fftwnd_one(p, (fftw_complex*)tpdata.data(), NULL);
double sqrtprod = sqrt(double(xn*yn));
for(int j=0; j<yn; j++) for(int i=0; i<xn; i++) tpdata(i,j) /= sqrtprod;
//store
csc[wcnt] = tpdata;
}
//fdct_wrapping_fftshift(xn,yn,xh,yh,tpdata,wpdata);
//ROTATION
//fdct_wrapping_rotate_backward(q, wpdata, csc[wcnt]);
wcnt++;
} //end of w loop
} //end of q loop
//PUT THE RIGHT DATA BACK
Xhgh = Xhghb;
XL1 = XL1b; XL2 = XL2b;
for(map<intpair, fftwnd_plan>::iterator mit=planmap.begin(); mit!=planmap.end(); mit++) {
fftwnd_plan p = (*mit).second;
fftwnd_destroy_plan(p);
}
return 0;
}
//-------------------------------------------------------------------
int fdct_wrapping(int N1, int N2, int nbscales, int nbangles_coarse, int allcurvelets, CpxNumMat& x, vector< vector<CpxNumMat> >& c)
{
//---------------------------------------------
assert(N1==x.m() && N2==x.n());
int F1 = N1/2; int F2 = N2/2;
// ifft original data
CpxNumMat T(x);
fftwnd_plan p = fftw2d_create_plan(N2, N1, FFTW_FORWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
fftwnd_one(p, (fftw_complex*)T.data(), NULL);
fftwnd_destroy_plan(p);
double sqrtprod = sqrt(double(N1*N2));
for(int j=0; j<N2; j++) for(int i=0; i<N1; i++) T(i,j) /= sqrtprod;
CpxOffMat O(N1, N2);
fdct_wrapping_fftshift(T, O);
//-----------------------------------------------------------------------------
vector<CpxOffMat> Xhghs;
Xhghs.resize(nbscales);
CpxOffMat X;
//unfold or not
if(allcurvelets==1) {
//--------------------------
double XL1 = 4.0*N1/3.0; double XL2 = 4.0*N2/3.0; //range
int XS1, XS2; int XF1, XF2; double XR1, XR2; fdct_wrapping_rangecompute(XL1, XL2, XS1, XS2, XF1, XF2, XR1, XR2);
IntOffVec t1(XS1);
for(int i=-XF1; i<-XF1+XS1; i++) if( i<-N1/2) t1(i) = i+int(N1); else if(i>(N1-1)/2) t1(i) = i-int(N1); else t1(i) = i;
IntOffVec t2(XS2);
for(int i=-XF2; i<-XF2+XS2; i++) if( i<-N2/2) t2(i) = i+int(N2); else if(i>(N2-1)/2) t2(i) = i-int(N2); else t2(i) = i;
X.resize(XS1, XS2);
for(int j=-XF2; j<-XF2+XS2; j++)
for(int i=-XF1; i<-XF1+XS1; i++)
X(i,j) = O(t1(i), t2(j));
DblOffMat lowpass(XS1,XS2);
fdct_wrapping_lowpasscompute(XL1, XL2, lowpass); //compute the low pass filter
for(int j=-XF2; j<-XF2+XS2; j++)
for(int i=-XF1; i<-XF1+XS1; i++)
X(i,j) *= lowpass(i,j);
} else {
//--------------------------
X = O;
}
//separate
double XL1 = 4.0*N1/3.0; double XL2 = 4.0*N2/3.0; //range
for(int sc=nbscales-1; sc>0; sc--) {
double XL1n = XL1/2; double XL2n = XL2/2;
int XS1n, XS2n; int XF1n, XF2n; double XR1n, XR2n;
fdct_wrapping_rangecompute(XL1n, XL2n, XS1n, XS2n, XF1n, XF2n, XR1n, XR2n);
//computer filter
DblOffMat lowpass(XS1n, XS2n);
fdct_wrapping_lowpasscompute(XL1n, XL2n, lowpass);
DblOffMat hghpass(XS1n, XS2n);
for(int j=-XF2n; j<-XF2n+XS2n; j++)
for(int i=-XF1n; i<-XF1n+XS1n; i++)
hghpass(i,j) = sqrt(1-lowpass(i,j)*lowpass(i,j));
//separate
CpxOffMat Xhgh(X);
for(int j=-XF2n; j<-XF2n+XS2n; j++)
for(int i=-XF1n; i<-XF1n+XS1n; i++)
Xhgh(i,j) *= hghpass(i,j);
CpxOffMat Xlow(XS1n, XS2n);
for(int j=-XF2n; j<-XF2n+XS2n; j++)
for(int i=-XF1n; i<-XF1n+XS1n; i++)
Xlow(i,j) = X(i,j) * lowpass(i,j);
//store and prepare for next level
Xhghs[sc] = Xhgh;
X = Xlow;
XL1 = XL1/2; XL2 = XL2/2;
}
Xhghs[0] = X;
//-----------------------------------------------------------------------------
vector<int> nbangles(nbscales);
if(allcurvelets==1) {
//nbangles
nbangles[0] = 1;
for(int sc=1; sc<nbscales; sc++) nbangles[sc] = nbangles_coarse * pow2( int(ceil(double(sc-1)/2)) );
//c
c.resize(nbscales);
for(int sc=0; sc<nbscales; sc++) c[sc].resize( nbangles[sc] );
double XL1 = 4.0*N1/3.0; double XL2 = 4.0*N2/3.0; //range
for(int sc=nbscales-1; sc>0; sc--) {
fdct_wrapping_sepangle(XL1, XL2, nbangles[sc], Xhghs[sc], c[sc]);
XL1 /= 2; XL2 /= 2;
}
fdct_wrapping_wavelet(Xhghs[0], c[0]);
} else {
//nbangles
nbangles[0] = 1;
for(int sc=1; sc<nbscales-1; sc++) nbangles[sc] = nbangles_coarse * pow2( int(ceil(double(sc-1)/2)) );
nbangles[nbscales-1] = 1;
//c
c.resize(nbscales);
for(int sc=0; sc<nbscales; sc++) c[sc].resize( nbangles[sc] );
fdct_wrapping_wavelet(Xhghs[nbscales-1], c[nbscales-1]);
double XL1 = 2.0*N1/3.0; double XL2 = 2.0*N2/3.0; //range
for(int sc=nbscales-2; sc>0; sc--) {
fdct_wrapping_sepangle(XL1, XL2, nbangles[sc], Xhghs[sc], c[sc]);
XL1 /= 2; XL2 /= 2;
}
fdct_wrapping_wavelet(Xhghs[0], c[0]);
}
return 0;
}
/* Call fftw for a 1 band complex image.
*/
static int
cfwfft1( IMAGE *dummy, IMAGE *in, IMAGE *out )
{
fftwnd_plan plan;
double *buf, *q, *p;
int x, y;
IMAGE *cmplx = im_open_local( dummy, "fwfft1:1", "t" );
/* Make dp complex image.
*/
if( !cmplx || im_pincheck( in ) || im_outcheck( out ) )
return( -1 );
if( in->Coding != IM_CODING_NONE || in->Bands != 1 ) {
im_error( "im_fwfft", _( "one band uncoded only" ) );
return( -1 );
}
if( im_clip2dcm( in, cmplx ) )
return( -1 );
/* Make the plan for the transform.
*/
if( !(plan = fftw2d_create_plan( in->Ysize, in->Xsize,
FFTW_FORWARD,
FFTW_MEASURE | FFTW_USE_WISDOM | FFTW_IN_PLACE )) ) {
im_error( "im_fwfft", _( "unable to create transform plan" ) );
return( -1 );
}
fftwnd_one( plan, (fftw_complex *) cmplx->data, NULL );
fftwnd_destroy_plan( plan );
/* WIO to out.
*/
if( im_cp_desc( out, in ) )
return( -1 );
out->Bbits = IM_BBITS_DPCOMPLEX;
out->BandFmt = IM_BANDFMT_DPCOMPLEX;
if( im_setupout( out ) )
return( -1 );
if( !(buf = (double *) IM_ARRAY( dummy,
IM_IMAGE_SIZEOF_LINE( out ), PEL )) )
return( -1 );
/* Copy to out, normalise.
*/
for( p = (double *) cmplx->data, y = 0; y < out->Ysize; y++ ) {
int size = out->Xsize * out->Ysize;
q = buf;
for( x = 0; x < out->Xsize; x++ ) {
q[0] = p[0] / size;
q[1] = p[1] / size;
p += 2;
q += 2;
}
if( im_writeline( y, out, (PEL *) buf ) )
return( -1 );
}
return( 0 );
}
//------------------------------------------------------------------------------------
int fdct3d_inverse_angles(int N1,int N2,int N3,int b, double L1,double L2,double L3, int s,int nd,
CpxCrvletPrtd& C, CpxNumTnsBlkd& W)
{
int mpirank; MPI_Comm_rank(MPI_COMM_WORLD, &mpirank);
int mpisize; MPI_Comm_size(MPI_COMM_WORLD, &mpisize);
vector< vector<int> >& Cowners = C.owners();
vector<int>& crvowners = Cowners[s]; //LEXING: the owner information for wedges in scale s
int nf = 6;
int wcnt = 0;
int S1, S2, S3; int F1, F2, F3; double R1, R2, R3; fdct3d_rangecompute(L1, L2, L3, S1, S2, S3, F1, F2, F3, R1, R2, R3);
DblOffVec big1(S1); fdct3d_lowpass(L1, big1);
DblOffVec big2(S2); fdct3d_lowpass(L2, big2);
DblOffVec big3(S3); fdct3d_lowpass(L3, big3);
double Lh1 = L1/2; double Lh2 = L2/2; double Lh3 = L3/2;
int Sh1, Sh2, Sh3; int Fh1, Fh2, Fh3; double Rh1, Rh2, Rh3; fdct3d_rangecompute(Lh1, Lh2, Lh3, Sh1, Sh2, Sh3, Fh1, Fh2, Fh3, Rh1, Rh2, Rh3);
DblOffVec sma1(S1); fdct3d_lowpass(Lh1, sma1);
DblOffVec sma2(S2); fdct3d_lowpass(Lh2, sma2);
DblOffVec sma3(S3); fdct3d_lowpass(Lh3, sma3);
double W1 = L1/nd; double W2 = L2/nd; double W3 = L3/nd;
typedef pair<int,int> intpair; typedef pair<int, intpair> inttriple;
map<inttriple, fftwnd_plan> planmap;
//face 0: x,y,z
for(int h=0; h<nd; h++) { //(y first z second)
for(int g=0; g<nd; g++) {
if(crvowners[wcnt]==mpirank) {
double xs = R1/4-(W1/2)/4; double xe = R1;
double ys = -R2 + (2*g-1)*W2/2; double ye = -R2 + (2*g+3)*W2/2;
double zs = -R3 + (2*h-1)*W3/2; double ze = -R3 + (2*h+3)*W3/2;
int xn = int(ceil(xe-xs)); int yn = int(ceil(ye-ys)); int zn = int(ceil(ze-zs));
double thts, thtm, thte; //y to x
if(g==0) {
thts = atan2(-1.0, 1.0-1.0/nd); thtm = atan2(-1.0+1.0/nd, 1.0); thte = atan2(-1.0+3.0/nd, 1.0);
} else if(g==nd-1) {
thts = atan2(-1.0+(2.0*g-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*g+1.0)/nd, 1.0); thte = atan2(1.0, 1.0-1.0/nd);
} else {
thts = atan2(-1.0+(2.0*g-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*g+1.0)/nd, 1.0); thte = atan2(-1.0+(2.0*g+3.0)/nd, 1.0);
}
double phis, phim, phie; //z to x
if(h==0) {
phis = atan2(-1.0, 1.0-1.0/nd); phim = atan2(-1.0+1.0/nd, 1.0); phie = atan2(-1.0+3.0/nd, 1.0);
} else if(h==nd-1) {
phis = atan2(-1.0+(2.0*h-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*h+1.0)/nd, 1.0); phie = atan2(1.0, 1.0-1.0/nd);
} else {
phis = atan2(-1.0+(2.0*h-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*h+1.0)/nd, 1.0); phie = atan2(-1.0+(2.0*h+3.0)/nd, 1.0);
}
int xh = xn/2; int yh = yn/2; int zh = zn/2; //half
double R21 = R2/R1; double R31 = R3/R1;
CpxNumTns tpdata(xn,yn,zn);
CpxNumTns& Cblk = C.block(s,wcnt);
tpdata = Cblk;
//fft
fftwnd_plan p = NULL;
map<inttriple, fftwnd_plan>::iterator mit = planmap.find( inttriple(xn, intpair(yn,zn)) );
if(mit!=planmap.end()) { p = (*mit).second;
} else {
p = fftw3d_create_plan(zn, yn, xn, FFTW_FORWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
planmap[ inttriple(xn, intpair(yn,zn)) ] = p;
}
fftwnd_one(p, (fftw_complex*)tpdata.data(), NULL); //cerr<<"wedge s"<<endl;
double sqrtprod = sqrt(double(xn*yn*zn));
for(int i=0; i<xn; i++) for(int j=0; j<yn; j++) for(int k=0; k<zn; k++) tpdata(i,j,k) /= sqrtprod;
CpxOffTns wpdata(xn,yn,zn);
fdct3d_fftshift(xn,yn,zn,tpdata,wpdata);
for(int xcur=(int)ceil(xs); xcur<xe; xcur++) {
int yfm = (int)ceil( max(-R2, R21*xcur*tan(thts)) );
int yto = (int)floor( min(R2, R21*xcur*tan(thte)) );
int zfm = (int)ceil( max(-R3, R31*xcur*tan(phis)) );
int zto = (int)floor( min(R3, R31*xcur*tan(phie)) );
for(int ycur=yfm; ycur<=yto; ycur++)
for(int zcur=zfm; zcur<=zto; zcur++) {
int tmpx = xcur%xn; if(tmpx<-xh) tmpx+=xn; if(tmpx>=-xh+xn) tmpx-=xn;
int tmpy = ycur%yn; if(tmpy<-yh) tmpy+=yn; if(tmpy>=-yh+yn) tmpy-=yn;
int tmpz = zcur%zn; if(tmpz<-zh) tmpz+=zn; if(tmpz>=-zh+zn) tmpz-=zn;
double thtcur = atan2(ycur/R2, xcur/R1);
double phicur = atan2(zcur/R3, xcur/R1);
double glbpou; fdct3d_globalpou(thtcur, phicur, M_PI/4-atan2(1.0-1.0/nd, 1.0), glbpou);
double wtht;
if(thtcur<thtm) {
if(g==0) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thts)/(thtm-thts), l, r); wtht = l; }
} else {
if(g==nd-1) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thtm)/(thte-thtm), l, r); wtht = r; }
}
double wphi;
if(phicur<phim) {
if(h==0) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phis)/(phim-phis), l, r); wphi = l; }
} else {
if(h==nd-1) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phim)/(phie-phim), l, r); wphi = r; }
}
double pou = glbpou * wtht * wphi;
wpdata(tmpx, tmpy, tmpz) *= pou;
double ss = sma1(xcur)*sma2(ycur)*sma3(zcur); double bb = big1(xcur)*big2(ycur)*big3(zcur);
int bi,bj,bk; int oi,oj,ok; fdct3d_position_aux(N1,N2,N3,b, xcur,ycur,zcur, bi,bj,bk,oi,oj,ok);
CpxNumTns& Wblk = W.block(bi,bj,bk);
Wblk(oi,oj,ok) += wpdata(tmpx,tmpy,tmpz) * bb * sqrt(1.0-ss*ss);
}
} //xcur
} //if
wcnt++;
}
} //end of face
//face 1. y z x
for(int f=0; f<nd; f++) {
for(int h=0; h<nd; h++) {
if(crvowners[wcnt]==mpirank) {
double ys = R2/4-(W2/2)/4; double ye = R2;
double zs = -R3 + (2*h-1)*W3/2; double ze = -R3 + (2*h+3)*W3/2;
double xs = -R1 + (2*f-1)*W1/2; double xe = -R1 + (2*f+3)*W1/2;
int xn = int(ceil(xe-xs)); int yn = int(ceil(ye-ys)); int zn = int(ceil(ze-zs));
double thts, thtm, thte; //z to y
if(h==0) {
thts = atan2(-1.0, 1.0-1.0/nd); thtm = atan2(-1.0+1.0/nd, 1.0); thte = atan2(-1.0+3.0/nd, 1.0);
} else if(h==nd-1) {
thts = atan2(-1.0+(2.0*h-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*h+1.0)/nd, 1.0); thte = atan2(1.0, 1.0-1.0/nd);
} else {
thts = atan2(-1.0+(2.0*h-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*h+1.0)/nd, 1.0); thte = atan2(-1.0+(2.0*h+3.0)/nd, 1.0);
}
double phis, phim, phie; //z to x
if(f==0) {
phis = atan2(-1.0, 1.0-1.0/nd); phim = atan2(-1.0+1.0/nd, 1.0); phie = atan2(-1.0+3.0/nd, 1.0);
} else if(f==nd-1) {
phis = atan2(-1.0+(2.0*f-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*f+1.0)/nd, 1.0); phie = atan2(1.0, 1.0-1.0/nd);
} else {
phis = atan2(-1.0+(2.0*f-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*f+1.0)/nd, 1.0); phie = atan2(-1.0+(2.0*f+3.0)/nd, 1.0);
}
int xh = xn/2; int yh = yn/2; int zh = zn/2;
double R32 = R3/R2; double R12 = R1/R2;
CpxNumTns tpdata(xn,yn,zn);
CpxNumTns& Cblk = C.block(s,wcnt);
tpdata = Cblk;
//fft
fftwnd_plan p = NULL;
map<inttriple, fftwnd_plan>::iterator mit = planmap.find( inttriple(xn, intpair(yn,zn)) );
if(mit!=planmap.end()) { p = (*mit).second;
} else {
p = fftw3d_create_plan(zn, yn, xn, FFTW_FORWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
planmap[ inttriple(xn, intpair(yn,zn)) ] = p;
}
fftwnd_one(p, (fftw_complex*)tpdata.data(), NULL); //cerr<<"wedge s"<<endl;
double sqrtprod = sqrt(double(xn*yn*zn));
for(int i=0; i<xn; i++) for(int j=0; j<yn; j++) for(int k=0; k<zn; k++) tpdata(i,j,k) /= sqrtprod;
CpxOffTns wpdata(xn,yn,zn);
fdct3d_fftshift(xn,yn,zn,tpdata,wpdata);
for(int ycur=(int)ceil(ys); ycur<ye; ycur++) {
int zfm = (int)ceil( max(-R3, R32*ycur*tan(thts)) );
int zto = (int)floor( min(R3, R32*ycur*tan(thte)) );
int xfm = (int)ceil( max(-R1, R12*ycur*tan(phis)) );
int xto = (int)floor( min(R1, R12*ycur*tan(phie)) );
for(int zcur=zfm; zcur<=zto; zcur++)
for(int xcur=xfm; xcur<=xto; xcur++) {
int tmpx = xcur%xn; if(tmpx<-xh) tmpx+=xn; if(tmpx>=-xh+xn) tmpx-=xn;
int tmpy = ycur%yn; if(tmpy<-yh) tmpy+=yn; if(tmpy>=-yh+yn) tmpy-=yn;
int tmpz = zcur%zn; if(tmpz<-zh) tmpz+=zn; if(tmpz>=-zh+zn) tmpz-=zn;
double thtcur = atan2(zcur/R3, ycur/R2);
double phicur = atan2(xcur/R1, ycur/R2);
double glbpou; fdct3d_globalpou(thtcur, phicur, M_PI/4-atan2(1.0-1.0/nd, 1.0), glbpou); //CHECK
double wtht;
if(thtcur<thtm) {
if(h==0) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thts)/(thtm-thts), l, r); wtht = l; }
} else {
if(h==nd-1) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thtm)/(thte-thtm), l, r); wtht = r; }
}
double wphi;
if(phicur<phim) {
if(f==0) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phis)/(phim-phis), l, r); wphi = l; }
} else {
if(f==nd-1) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phim)/(phie-phim), l, r); wphi = r; }
}
double pou = glbpou * wtht * wphi;
wpdata(tmpx, tmpy, tmpz) *= pou;
double ss = sma1(xcur)*sma2(ycur)*sma3(zcur); double bb = big1(xcur)*big2(ycur)*big3(zcur);
int bi,bj,bk; int oi,oj,ok; fdct3d_position_aux(N1,N2,N3,b, xcur,ycur,zcur, bi,bj,bk,oi,oj,ok);
CpxNumTns& Wblk = W.block(bi,bj,bk);
Wblk(oi,oj,ok) += wpdata(tmpx,tmpy,tmpz) * bb * sqrt(1.0-ss*ss);
}
} //ycur
}//if
wcnt++;
}
}//end of face
//face 2. z x y
for(int g=0; g<nd; g++) {
for(int f=0; f<nd; f++) {
if(crvowners[wcnt]==mpirank) {
double zs = R3/4-(W3/2)/4; double ze = R3;
double xs = -R1 + (2*f-1)*W1/2; double xe = -R1 + (2*f+3)*W1/2;
double ys = -R2 + (2*g-1)*W2/2; double ye = -R2 + (2*g+3)*W2/2;
int xn = int(ceil(xe-xs)); int yn = int(ceil(ye-ys)); int zn = int(ceil(ze-zs));
double thts, thtm, thte; //y to x
if(f==0) {
thts = atan2(-1.0, 1.0-1.0/nd); thtm = atan2(-1.0+1.0/nd, 1.0); thte = atan2(-1.0+3.0/nd, 1.0);
} else if(f==nd-1) {
thts = atan2(-1.0+(2.0*f-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*f+1.0)/nd, 1.0); thte = atan2(1.0, 1.0-1.0/nd);
} else {
thts = atan2(-1.0+(2.0*f-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*f+1.0)/nd, 1.0); thte = atan2(-1.0+(2.0*f+3.0)/nd, 1.0);
}
double phis, phim, phie; //z to x
if(g==0) {
phis = atan2(-1.0, 1.0-1.0/nd); phim = atan2(-1.0+1.0/nd, 1.0); phie = atan2(-1.0+3.0/nd, 1.0);
} else if(g==nd-1) {
phis = atan2(-1.0+(2.0*g-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*g+1.0)/nd, 1.0); phie = atan2(1.0, 1.0-1.0/nd);
} else {
phis = atan2(-1.0+(2.0*g-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*g+1.0)/nd, 1.0); phie = atan2(-1.0+(2.0*g+3.0)/nd, 1.0);
}
int xh = xn/2; int yh = yn/2; int zh = zn/2;
double R13 = double(F1)/double(F3); double R23 = double(F2)/double(F3);
CpxNumTns tpdata(xn,yn,zn);
CpxNumTns& Cblk = C.block(s,wcnt);
tpdata = Cblk;
//fft
fftwnd_plan p = NULL;
map<inttriple, fftwnd_plan>::iterator mit = planmap.find( inttriple(xn, intpair(yn,zn)) );
if(mit!=planmap.end()) { p = (*mit).second;
} else {
p = fftw3d_create_plan(zn, yn, xn, FFTW_FORWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
planmap[ inttriple(xn, intpair(yn,zn)) ] = p;
}
fftwnd_one(p, (fftw_complex*)tpdata.data(), NULL); //cerr<<"wedge s"<<endl;
double sqrtprod = sqrt(double(xn*yn*zn));
for(int i=0; i<xn; i++) for(int j=0; j<yn; j++) for(int k=0; k<zn; k++) tpdata(i,j,k) /= sqrtprod;
CpxOffTns wpdata(xn,yn,zn);
fdct3d_fftshift(xn,yn,zn,tpdata,wpdata);
for(int zcur=(int)ceil(zs); zcur<ze; zcur++) {
int xfm = (int)ceil( max(-R1, R13*zcur*tan(thts)) );
int xto = (int)floor( min(R1, R13*zcur*tan(thte)) );
int yfm = (int)ceil( max(-R2, R23*zcur*tan(phis)) );
int yto = (int)floor( min(R2, R23*zcur*tan(phie)) );
for(int xcur=xfm; xcur<=xto; xcur++)
for(int ycur=yfm; ycur<=yto; ycur++) {
int tmpx = xcur%xn; if(tmpx<-xh) tmpx+=xn; if(tmpx>=-xh+xn) tmpx-=xn;
int tmpy = ycur%yn; if(tmpy<-yh) tmpy+=yn; if(tmpy>=-yh+yn) tmpy-=yn;
int tmpz = zcur%zn; if(tmpz<-zh) tmpz+=zn; if(tmpz>=-zh+zn) tmpz-=zn;
double thtcur = atan2(xcur/R1, zcur/R3);
double phicur = atan2(ycur/R2, zcur/R3);
double glbpou; fdct3d_globalpou(thtcur, phicur, M_PI/4-atan2(1.0-1.0/nd, 1.0), glbpou);
double wtht;
if(thtcur<thtm) {
if(f==0) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thts)/(thtm-thts), l, r); wtht = l; }
} else {
if(f==nd-1) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thtm)/(thte-thtm), l, r); wtht = r; }
}
double wphi;
if(phicur<phim) {
if(g==0) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phis)/(phim-phis), l, r); wphi = l; }
} else {
if(g==nd-1) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phim)/(phie-phim), l, r); wphi = r; }
}
double pou = glbpou * wtht * wphi;
wpdata(tmpx, tmpy, tmpz) *= pou;
double ss = sma1(xcur)*sma2(ycur)*sma3(zcur); double bb = big1(xcur)*big2(ycur)*big3(zcur);
int bi,bj,bk; int oi,oj,ok; fdct3d_position_aux(N1,N2,N3,b, xcur,ycur,zcur, bi,bj,bk,oi,oj,ok);
CpxNumTns& Wblk = W.block(bi,bj,bk);
Wblk(oi,oj,ok) += wpdata(tmpx,tmpy,tmpz) * bb * sqrt(1.0-ss*ss);
}
}//zcur
}//if
wcnt++;
}
}//end of face
//face 3: -x,-y,-z
for(int h=nd-1; h>=0; h--) {
for(int g=nd-1; g>=0; g--) {
if(crvowners[wcnt]==mpirank) {
double xs = -R1; double xe = -R1/4+(W1/2)/4;
double ys = -R2 + (2*g-1)*W2/2; double ye = -R2 + (2*g+3)*W2/2;
double zs = -R3 + (2*h-1)*W3/2; double ze = -R3 + (2*h+3)*W3/2;
int xn = int(ceil(xe-xs)); int yn = int(ceil(ye-ys)); int zn = int(ceil(ze-zs));
double thts, thtm, thte; //y to x
if(g==0) {
thts = atan2(-1.0, 1.0-1.0/nd); thtm = atan2(-1.0+1.0/nd, 1.0); thte = atan2(-1.0+3.0/nd, 1.0);
} else if(g==nd-1) {
thts = atan2(-1.0+(2.0*g-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*g+1.0)/nd, 1.0); thte = atan2(1.0, 1.0-1.0/nd);
} else {
thts = atan2(-1.0+(2.0*g-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*g+1.0)/nd, 1.0); thte = atan2(-1.0+(2.0*g+3.0)/nd, 1.0);
}
double phis, phim, phie; //z to x
if(h==0) {
phis = atan2(-1.0, 1.0-1.0/nd); phim = atan2(-1.0+1.0/nd, 1.0); phie = atan2(-1.0+3.0/nd, 1.0);
} else if(h==nd-1) {
phis = atan2(-1.0+(2.0*h-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*h+1.0)/nd, 1.0); phie = atan2(1.0, 1.0-1.0/nd);
} else {
phis = atan2(-1.0+(2.0*h-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*h+1.0)/nd, 1.0); phie = atan2(-1.0+(2.0*h+3.0)/nd, 1.0);
}
int xh = xn/2; int yh = yn/2; int zh = zn/2;
double R21 = R2/R1; double R31 = R3/R1;
CpxNumTns tpdata(xn,yn,zn);
CpxNumTns& Cblk = C.block(s,wcnt);
tpdata = Cblk;
//fft
fftwnd_plan p = NULL;
map<inttriple, fftwnd_plan>::iterator mit = planmap.find( inttriple(xn, intpair(yn,zn)) );
if(mit!=planmap.end()) { p = (*mit).second;
} else {
p = fftw3d_create_plan(zn, yn, xn, FFTW_FORWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
planmap[ inttriple(xn, intpair(yn,zn)) ] = p;
}
fftwnd_one(p, (fftw_complex*)tpdata.data(), NULL); //cerr<<"wedge s"<<endl;
double sqrtprod = sqrt(double(xn*yn*zn));
for(int i=0; i<xn; i++) for(int j=0; j<yn; j++) for(int k=0; k<zn; k++) tpdata(i,j,k) /= sqrtprod;
CpxOffTns wpdata(xn,yn,zn);
fdct3d_fftshift(xn,yn,zn,tpdata,wpdata);
for(int xcur=(int)ceil(xs); xcur<xe; xcur++) {
int yfm = (int)ceil( max(-R2, R21*(-xcur)*tan(thts)) );
int yto = (int)floor( min(R2, R21*(-xcur)*tan(thte)) );
int zfm = (int)ceil( max(-R3, R31*(-xcur)*tan(phis)) );
int zto = (int)floor( min(R3, R31*(-xcur)*tan(phie)) );
for(int ycur=yfm; ycur<=yto; ycur++)
for(int zcur=zfm; zcur<=zto; zcur++) {
int tmpx = xcur%xn; if(tmpx<-xh) tmpx+=xn; if(tmpx>=-xh+xn) tmpx-=xn;
int tmpy = ycur%yn; if(tmpy<-yh) tmpy+=yn; if(tmpy>=-yh+yn) tmpy-=yn;
int tmpz = zcur%zn; if(tmpz<-zh) tmpz+=zn; if(tmpz>=-zh+zn) tmpz-=zn;
double thtcur = atan2(ycur/R2, (-xcur)/R1);
double phicur = atan2(zcur/R3, (-xcur)/R1);
double glbpou; fdct3d_globalpou(thtcur, phicur, M_PI/4-atan2(1.0-1.0/nd, 1.0), glbpou);
double wtht;
if(thtcur<thtm) {
if(g==0) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thts)/(thtm-thts), l, r); wtht = l; }
} else {
if(g==nd-1) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thtm)/(thte-thtm), l, r); wtht = r; }
}
double wphi;
if(phicur<phim) {
if(h==0) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phis)/(phim-phis), l, r); wphi = l; }
} else {
if(h==nd-1) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phim)/(phie-phim), l, r); wphi = r; }
}
double pou = glbpou * wtht * wphi;
wpdata(tmpx, tmpy, tmpz) *= pou;
double ss = sma1(xcur)*sma2(ycur)*sma3(zcur); double bb = big1(xcur)*big2(ycur)*big3(zcur);
int bi,bj,bk; int oi,oj,ok; fdct3d_position_aux(N1,N2,N3,b, xcur,ycur,zcur, bi,bj,bk,oi,oj,ok);
CpxNumTns& Wblk = W.block(bi,bj,bk);
Wblk(oi,oj,ok) += wpdata(tmpx,tmpy,tmpz) * bb * sqrt(1.0-ss*ss);
}
} //xcur
} //if
wcnt++;
}
} //end of face
//face 4: -y,-z,-x
for(int f=nd-1; f>=0; f--) {
for(int h=nd-1; h>=0; h--) {
if(crvowners[wcnt]==mpirank) {
double ys = -R2; double ye = -R2/4+(W2/2)/4;
double zs = -R3 + (2*h-1)*W3/2; double ze = -R3 + (2*h+3)*W3/2;
double xs = -R1 + (2*f-1)*W1/2; double xe = -R1 + (2*f+3)*W1/2;
int xn = int(ceil(xe-xs)); int yn = int(ceil(ye-ys)); int zn = int(ceil(ze-zs));
double thts, thtm, thte; //z to y
if(h==0) {
thts = atan2(-1.0, 1.0-1.0/nd); thtm = atan2(-1.0+1.0/nd, 1.0); thte = atan2(-1.0+3.0/nd, 1.0);
} else if(h==nd-1) {
thts = atan2(-1.0+(2.0*h-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*h+1.0)/nd, 1.0); thte = atan2(1.0, 1.0-1.0/nd);
} else {
thts = atan2(-1.0+(2.0*h-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*h+1.0)/nd, 1.0); thte = atan2(-1.0+(2.0*h+3.0)/nd, 1.0);
}
double phis, phim, phie; //z to x
if(f==0) {
phis = atan2(-1.0, 1.0-1.0/nd); phim = atan2(-1.0+1.0/nd, 1.0); phie = atan2(-1.0+3.0/nd, 1.0);
} else if(f==nd-1) {
phis = atan2(-1.0+(2.0*f-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*f+1.0)/nd, 1.0); phie = atan2(1.0, 1.0-1.0/nd);
} else {
phis = atan2(-1.0+(2.0*f-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*f+1.0)/nd, 1.0); phie = atan2(-1.0+(2.0*f+3.0)/nd, 1.0);
}
int xh = xn/2; int yh = yn/2; int zh = zn/2;
double R32 = double(F3)/double(F2); double R12 = double(F1)/double(F2);
CpxNumTns tpdata(xn,yn,zn);
CpxNumTns& Cblk = C.block(s,wcnt);
tpdata = Cblk;
//fft
fftwnd_plan p = NULL;
map<inttriple, fftwnd_plan>::iterator mit = planmap.find( inttriple(xn, intpair(yn,zn)) );
if(mit!=planmap.end()) { p = (*mit).second;
} else {
p = fftw3d_create_plan(zn, yn, xn, FFTW_FORWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
planmap[ inttriple(xn, intpair(yn,zn)) ] = p;
}
fftwnd_one(p, (fftw_complex*)tpdata.data(), NULL); //cerr<<"wedge s"<<endl;
double sqrtprod = sqrt(double(xn*yn*zn));
for(int i=0; i<xn; i++) for(int j=0; j<yn; j++) for(int k=0; k<zn; k++) tpdata(i,j,k) /= sqrtprod;
CpxOffTns wpdata(xn,yn,zn);
fdct3d_fftshift(xn,yn,zn,tpdata,wpdata);
for(int ycur=(int)ceil(ys); ycur<ye; ycur++) {
int zfm = (int)ceil( max(-R3, R32*(-ycur)*tan(thts)) );
int zto = (int)floor( min(R3, R32*(-ycur)*tan(thte)) );
int xfm = (int)ceil( max(-R1, R12*(-ycur)*tan(phis)) );
int xto = (int)floor( min(R1, R12*(-ycur)*tan(phie)) );
for(int zcur=zfm; zcur<=zto; zcur++)
for(int xcur=xfm; xcur<=xto; xcur++) {
int tmpx = xcur%xn; if(tmpx<-xh) tmpx+=xn; if(tmpx>=-xh+xn) tmpx-=xn;
int tmpy = ycur%yn; if(tmpy<-yh) tmpy+=yn; if(tmpy>=-yh+yn) tmpy-=yn;
int tmpz = zcur%zn; if(tmpz<-zh) tmpz+=zn; if(tmpz>=-zh+zn) tmpz-=zn;
double thtcur = atan2(zcur/R3, (-ycur)/R2);
double phicur = atan2(xcur/R1, (-ycur)/R2);
double glbpou; fdct3d_globalpou(thtcur, phicur, M_PI/4-atan2(1.0-1.0/nd, 1.0), glbpou); //CHECK
double wtht;
if(thtcur<thtm) {
if(h==0) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thts)/(thtm-thts), l, r); wtht = l; }
} else {
if(h==nd-1) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thtm)/(thte-thtm), l, r); wtht = r; }
}
double wphi;
if(phicur<phim) {
if(f==0) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phis)/(phim-phis), l, r); wphi = l; }
} else {
if(f==nd-1) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phim)/(phie-phim), l, r); wphi = r; }
}
double pou = glbpou * wtht * wphi;
wpdata(tmpx, tmpy, tmpz) *= pou;
double ss = sma1(xcur)*sma2(ycur)*sma3(zcur); double bb = big1(xcur)*big2(ycur)*big3(zcur);
int bi,bj,bk; int oi,oj,ok; fdct3d_position_aux(N1,N2,N3,b, xcur,ycur,zcur, bi,bj,bk,oi,oj,ok);
CpxNumTns& Wblk = W.block(bi,bj,bk);
Wblk(oi,oj,ok) += wpdata(tmpx,tmpy,tmpz) * bb * sqrt(1.0-ss*ss);
}
} //ycur
}//if
wcnt++;
}
}//end of face
//face 5.-z,-x,-y
for(int g=nd-1; g>=0; g--) {
for(int f=nd-1; f>=0; f--) {
if(crvowners[wcnt]==mpirank) {
double zs = -R3; double ze = -R3/4+(W3/2)/4;
double xs = -R1 + (2*f-1)*W1/2; double xe = -R1 + (2*f+3)*W1/2;
double ys = -R2 + (2*g-1)*W2/2; double ye = -R2 + (2*g+3)*W2/2;
int xn = int(ceil(xe-xs)); int yn = int(ceil(ye-ys)); int zn = int(ceil(ze-zs));
double thts, thtm, thte; //y to x
if(f==0) {
thts = atan2(-1.0, 1.0-1.0/nd); thtm = atan2(-1.0+1.0/nd, 1.0); thte = atan2(-1.0+3.0/nd, 1.0);
} else if(f==nd-1) {
thts = atan2(-1.0+(2.0*f-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*f+1.0)/nd, 1.0); thte = atan2(1.0, 1.0-1.0/nd);
} else {
thts = atan2(-1.0+(2.0*f-1.0)/nd, 1.0); thtm = atan2(-1.0+(2.0*f+1.0)/nd, 1.0); thte = atan2(-1.0+(2.0*f+3.0)/nd, 1.0);
}
double phis, phim, phie; //z to x
if(g==0) {
phis = atan2(-1.0, 1.0-1.0/nd); phim = atan2(-1.0+1.0/nd, 1.0); phie = atan2(-1.0+3.0/nd, 1.0);
} else if(g==nd-1) {
phis = atan2(-1.0+(2.0*g-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*g+1.0)/nd, 1.0); phie = atan2(1.0, 1.0-1.0/nd);
} else {
phis = atan2(-1.0+(2.0*g-1.0)/nd, 1.0); phim = atan2(-1.0+(2.0*g+1.0)/nd, 1.0); phie = atan2(-1.0+(2.0*g+3.0)/nd, 1.0);
}
int xh = xn/2; int yh = yn/2; int zh = zn/2;
double R13 = double(F1)/double(F3); double R23 = double(F2)/double(F3);
CpxNumTns tpdata(xn,yn,zn);
CpxNumTns& Cblk = C.block(s,wcnt);
tpdata = Cblk;
//fft
fftwnd_plan p = NULL;
map<inttriple, fftwnd_plan>::iterator mit = planmap.find( inttriple(xn, intpair(yn,zn)) );
if(mit!=planmap.end()) { p = (*mit).second;
} else {
p = fftw3d_create_plan(zn, yn, xn, FFTW_FORWARD, FFTW_ESTIMATE | FFTW_IN_PLACE);
planmap[ inttriple(xn, intpair(yn,zn)) ] = p;
}
fftwnd_one(p, (fftw_complex*)tpdata.data(), NULL); //cerr<<"wedge s"<<endl;
double sqrtprod = sqrt(double(xn*yn*zn));
for(int i=0; i<xn; i++) for(int j=0; j<yn; j++) for(int k=0; k<zn; k++) tpdata(i,j,k) /= sqrtprod;
CpxOffTns wpdata(xn,yn,zn);
fdct3d_fftshift(xn,yn,zn,tpdata,wpdata);
for(int zcur=(int)ceil(zs); zcur<ze; zcur++) {
int xfm = (int)ceil( max(-R1, R13*(-zcur)*tan(thts)) );
int xto = (int)floor( min(R1, R13*(-zcur)*tan(thte)) );
int yfm = (int)ceil( max(-R2, R23*(-zcur)*tan(phis)) );
int yto = (int)floor( min(R2, R23*(-zcur)*tan(phie)) );
for(int xcur=xfm; xcur<=xto; xcur++)
for(int ycur=yfm; ycur<=yto; ycur++) {
int tmpx = xcur%xn; if(tmpx<-xh) tmpx+=xn; if(tmpx>=-xh+xn) tmpx-=xn;
int tmpy = ycur%yn; if(tmpy<-yh) tmpy+=yn; if(tmpy>=-yh+yn) tmpy-=yn;
int tmpz = zcur%zn; if(tmpz<-zh) tmpz+=zn; if(tmpz>=-zh+zn) tmpz-=zn;
double thtcur = atan2(xcur/R1, (-zcur)/R3);
double phicur = atan2(ycur/R2, (-zcur)/R3);
double glbpou; fdct3d_globalpou(thtcur, phicur, M_PI/4-atan2(1.0-1.0/nd, 1.0), glbpou);
double wtht;
if(thtcur<thtm) {
if(f==0) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thts)/(thtm-thts), l, r); wtht = l; }
} else {
if(f==nd-1) wtht = 1;
else { double l,r; fdct3d_window( (thtcur-thtm)/(thte-thtm), l, r); wtht = r; }
}
double wphi;
if(phicur<phim) {
if(g==0) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phis)/(phim-phis), l, r); wphi = l; }
} else {
if(g==nd-1) wphi = 1;
else { double l,r; fdct3d_window( (phicur-phim)/(phie-phim), l, r); wphi = r; }
}
double pou = glbpou * wtht * wphi;
wpdata(tmpx, tmpy, tmpz) *= pou;
double ss = sma1(xcur)*sma2(ycur)*sma3(zcur); double bb = big1(xcur)*big2(ycur)*big3(zcur);
int bi,bj,bk; int oi,oj,ok; fdct3d_position_aux(N1,N2,N3,b, xcur,ycur,zcur, bi,bj,bk,oi,oj,ok);
CpxNumTns& Wblk = W.block(bi,bj,bk);
Wblk(oi,oj,ok) += wpdata(tmpx,tmpy,tmpz) * bb * sqrt(1.0-ss*ss);
}
}//zcur
}//if
wcnt++;
}
}//end of face
iA(wcnt==nd*nd*nf);
//remove plans
for(map<inttriple, fftwnd_plan>::iterator mit=planmap.begin(); mit!=planmap.end(); mit++) {
fftwnd_plan p = (*mit).second;
fftwnd_destroy_plan(p);
}
return 0;
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix;
int dxm[4], dxn[4], ixpm, ixpn;
int sid;
double *disc = (double*)NULL;
double *disc2 = (double*)NULL;
double *work = (double*)NULL;
double q[4], fnorm;
int verbose = 0;
int do_gt = 0;
char filename[100], contype[200];
double ratime, retime;
double plaq, _2kappamu, hpe3_coeff, onepmutilde2, mutilde2;
double spinor1[24], spinor2[24], U_[18], U1_[18], U2_[18];
double *gauge_trafo=(double*)NULL;
complex w, w1, w2, *cp1, *cp2, *cp3;
FILE *ofs;
fftw_complex *in=(fftw_complex*)NULL;
#ifdef MPI
fftwnd_mpi_plan plan_p, plan_m;
#else
fftwnd_plan plan_p, plan_m;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vgf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'g':
do_gt = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
if(filename_set==0) strcpy(filename, "cvc.input");
fprintf(stdout, "# Reading input from file %s\n", filename);
read_input_parser(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
/* initialize MPI parameters */
mpi_init(argc, argv);
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifdef MPI
plan_p = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_BACKWARD, FFTW_MEASURE);
plan_m = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_FORWARD, FFTW_MEASURE);
fftwnd_mpi_local_sizes(plan_p, &T, &Tstart, &l_LX_at, &l_LXstart_at, &FFTW_LOC_VOLUME);
#else
plan_p = fftwnd_create_plan(4, dims, FFTW_BACKWARD, FFTW_MEASURE | FFTW_IN_PLACE);
plan_m = fftwnd_create_plan(4, dims, FFTW_FORWARD, FFTW_MEASURE | FFTW_IN_PLACE);
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] fftw parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(2);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
/* read the gauge field */
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
#ifdef MPI
xchange_gauge();
#endif
/* measure the plaquette */
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value: %25.16e\n", plaq);
if(do_gt==1) {
/***********************************
* initialize gauge transformation
***********************************/
init_gauge_trafo(&gauge_trafo, 1.);
apply_gt_gauge(gauge_trafo);
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value after gauge trafo: %25.16e\n", plaq);
}
/****************************************
* allocate memory for the spinor fields
****************************************/
no_fields = 3;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/****************************************
* allocate memory for the contractions
****************************************/
disc = (double*)calloc( 8*VOLUME, sizeof(double));
if( disc == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
disc2 = (double*)calloc( 8*VOLUME, sizeof(double));
if( disc2 == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc2\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc2[ix] = 0.;
work = (double*)calloc(48*VOLUME, sizeof(double));
if( work == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for work\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
/****************************************
* prepare Fourier transformation arrays
****************************************/
in = (fftw_complex*)malloc(FFTW_LOC_VOLUME*sizeof(fftw_complex));
if(in==(fftw_complex*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(4);
}
/************************************************
* HPE: calculate coeff. of 3rd order term
************************************************/
_2kappamu = 2. * g_kappa * g_mu;
onepmutilde2 = 1. + _2kappamu * _2kappamu;
mutilde2 = _2kappamu * _2kappamu;
hpe3_coeff = 16. * g_kappa*g_kappa*g_kappa*g_kappa * (1. + 6. * mutilde2 + mutilde2*mutilde2) / onepmutilde2 / onepmutilde2 / onepmutilde2 / onepmutilde2;
/*
hpe3_coeff = 8. * g_kappa*g_kappa*g_kappa * \
(1. + 6.*_2kappamu*_2kappamu + _2kappamu*_2kappamu*_2kappamu*_2kappamu) / (1. + _2kappamu*_2kappamu) / (1. + _2kappamu*_2kappamu) / (1. + _2kappamu*_2kappamu) / (1. + _2kappamu*_2kappamu);
*/
fprintf(stdout, "hpe3_coeff = %25.16e\n", hpe3_coeff);
/************************************************
* HPE: calculate the plaquette terms
************************************************/
for(ix=0; ix<VOLUME; ix++) {
for(mu=0; mu<4; mu++) {
for(i=1; i<4; i++) {
nu = (mu+i)%4;
_cm_eq_cm_ti_cm(U1_, g_gauge_field+_GGI(ix,mu), g_gauge_field+_GGI(g_iup[ix][mu],nu) );
_cm_eq_cm_ti_cm(U2_, g_gauge_field+_GGI(ix,nu), g_gauge_field+_GGI(g_iup[ix][nu],mu) );
_cm_eq_cm_ti_cm_dag(U_, U1_, U2_);
_co_eq_tr_cm(&w1, U_);
iix = g_idn[ix][nu];
_cm_eq_cm_ti_cm(U1_, g_gauge_field+_GGI(iix,mu), g_gauge_field+_GGI(g_iup[iix][mu],nu) );
_cm_eq_cm_ti_cm(U2_, g_gauge_field+_GGI(iix,nu), g_gauge_field+_GGI(g_iup[iix][nu],mu) );
_cm_eq_cm_ti_cm_dag(U_, U1_, U2_);
_co_eq_tr_cm(&w2, U_);
disc2[_GWI(mu,ix,VOLUME)+1] += hpe3_coeff * (w1.im - w2.im);
/*
_cm_eq_cm_ti_cm(U1_, g_gauge_field+_GGI(g_idn[ix][nu],nu), g_gauge_field+_GGI(ix,mu) );
_cm_eq_cm_ti_cm(U2_, g_gauge_field+_GGI(g_idn[ix][nu],mu), g_gauge_field+_GGI(g_iup[g_idn[ix][nu]][mu], nu) );
_cm_eq_cm_ti_cm_dag(U_, U1_, U2_);
_co_eq_tr_cm(&w2, U_);
disc2[_GWI(mu,ix,VOLUME)+1] += hpe3_coeff * (w1.im + w2.im);
*/
/* fprintf(stdout, "mu=%1d, ix=%5d, nu=%1d, w1=%25.16e +i %25.16e; w2=%25.16e +i %25.16e\n",
mu, ix, nu, w1.re, w1.im, w2.re, w2.im); */
} /* of nu */
/****************************************
* - in case lattice size equals 4
* calculate additional loop term
* - _NOTE_ the possible minus sign from
* the fermionic boundary conditions
****************************************/
if(dims[mu]==4) {
wilson_loop(&w, ix, mu, dims[mu]);
fnorm = -64. * g_kappa*g_kappa*g_kappa*g_kappa / onepmutilde2 / onepmutilde2 / onepmutilde2 / onepmutilde2;
disc2[_GWI(mu,ix,VOLUME)+1] += fnorm * w.im;
/* fprintf(stdout, "loop contribution: ix=%5d, mu=%2d, fnorm=%25.16e, w=%25.16e\n", ix, mu, fnorm, w.im); */
}
/*
fprintf(stdout, "-------------------------------------------\n");
fprintf(stdout, "disc2[ix=%d,mu=%d] = %25.16e +i %25.16e\n", ix, mu, disc2[_GWI(mu,ix,VOLUME)], disc2[_GWI(mu,ix,VOLUME)+1]);
fprintf(stdout, "-------------------------------------------\n");
*/
}
}
/*
sprintf(filename, "avc_disc_hpe5_3rd.%.4d", Nconf);
ofs = fopen(filename, "w");
for(ix=0; ix<VOLUME; ix++) {
for(mu=0; mu<4; mu++) {
fprintf(ofs, "%6d%3d%25.16e\t%25.16e\n", ix, mu, disc[_GWI(mu,ix,VOLUME)], disc[_GWI(mu,ix,VOLUME)+1]);
}
}
fclose(ofs);
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
*/
/*
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
ix = g_ipt[x0][x1][x2][x3];
for(mu=0; mu<4; mu++) {
dxm[0]=0; dxm[1]=0; dxm[2]=0; dxm[3]=0; dxm[mu]=1;
for(i=1; i<4; i++) {
nu = (mu+i)%4;
dxn[0]=0; dxn[1]=0; dxn[2]=0; dxn[3]=0; dxn[nu]=1;
ixpm = g_ipt[(x0+dxm[0]+T)%T][(x1+dxm[1]+LX)%LX][(x2+dxm[2]+LY)%LY][(x3+dxm[3]+LZ)%LZ];
ixpn = g_ipt[(x0+dxn[0]+T)%T][(x1+dxn[1]+LX)%LX][(x2+dxn[2]+LY)%LY][(x3+dxn[3]+LZ)%LZ];
_cm_eq_cm_ti_cm(U1_, g_gauge_field + 72*ix+18*mu, g_gauge_field + 72*ixpm+18*nu );
_cm_eq_cm_ti_cm(U2_, g_gauge_field + 72*ix+18*nu, g_gauge_field + 72*ixpn+18*mu );
_cm_eq_cm_ti_cm_dag(U_, U1_, U2_);
_co_eq_tr_cm(&w1, U_);
ixpm = g_ipt[(x0+dxm[0]-dxn[0]+T)%T][(x1+dxm[1]-dxn[1]+LX)%LX][(x2+dxm[2]-dxn[2]+LY)%LY][(x3+dxm[3]-dxn[3]+LZ)%LZ];
ixpn = g_ipt[(x0-dxn[0]+T)%T][(x1-dxn[1]+LX)%LX][(x2-dxn[2]+LY)%LY][(x3-dxn[3]+LZ)%LZ];
_cm_eq_cm_ti_cm(U1_, g_gauge_field + 72*ixpn+18*nu, g_gauge_field + 72*ix+18*mu);
_cm_eq_cm_ti_cm(U2_, g_gauge_field + 72*ixpn+18*mu, g_gauge_field + 72*ixpm+18*nu);
_cm_eq_cm_ti_cm_dag(U_, U1_, U2_);
_co_eq_tr_cm(&w2, U_);
disc2[_GWI(mu,ix,VOLUME)+1] += hpe3_coeff * (w1.im + w2.im);
fprintf(stdout, "mu=%1d, ix=%5d, nu=%1d, w1=%25.16e; w2=%25.16e\n", mu, ix, nu, w1.im, w2.im);
}
fprintf(stdout, "-------------------------------------------\n");
fprintf(stdout, "disc2[ix=%d,mu=%d] = %25.16e +i %25.16e\n", ix, mu, disc2[_GWI(mu,ix,VOLUME)], disc2[_GWI(mu,ix,VOLUME)+1]);
fprintf(stdout, "-------------------------------------------\n");
}
}
}
}
}
*/
/***********************************************
* start loop on source id.s
***********************************************/
for(sid=g_sourceid; sid<=g_sourceid2; sid+=g_sourceid_step) {
/* read the new propagator */
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid);
if(read_lime_spinor(g_spinor_field[2], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, Nconf, sid);
if(read_cmi(g_spinor_field[2], filename) != 0) break;
}
xchange_field(g_spinor_field[2]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "time to read prop.: %e seconds\n", retime-ratime);
if(do_gt==1) {
/******************************************
* gauge transform the propagators for sid
******************************************/
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_cm_ti_fv(spinor1, gauge_trafo+18*ix, g_spinor_field[2]+_GSI(ix));
_fv_eq_fv(g_spinor_field[2]+_GSI(ix), spinor1);
}
xchange_field(g_spinor_field[2]);
}
count++;
/************************************************
* calculate the source: apply Q_phi_tbc
************************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_phi_tbc(g_spinor_field[0], g_spinor_field[2]);
xchange_field(g_spinor_field[0]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to calculate source: %e seconds\n", retime-ratime);
/************************************************
* HPE: apply BH5
************************************************/
BH5(g_spinor_field[1], g_spinor_field[2]);
/* add new contractions to (existing) disc */
# ifdef MPI
ratime = MPI_Wtime();
# else
ratime = (double)clock() / CLOCKS_PER_SEC;
# endif
for(mu=0; mu<4; mu++) { /* loop on Lorentz index of the current */
iix = _GWI(mu,0,VOLUME);
for(ix=0; ix<VOLUME; ix++) { /* loop on lattice sites */
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
/* first contribution */
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
/* second contribution */
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
iix += 2;
} /* of ix */
} /* of mu */
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to contract cvc: %e seconds\n", retime-ratime);
/************************************************
* save results for count = multiple of Nsave
************************************************/
if(count%Nsave == 0) {
if(g_cart_id == 0) fprintf(stdout, "save results for count = %d\n", count);
fnorm = 1. / ( (double)count * g_prop_normsqr );
if(g_cart_id==0) fprintf(stdout, "# X-fnorm = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(ix=0; ix<VOLUME; ix++) {
work[_GWI(mu,ix,VOLUME) ] = disc[_GWI(mu,ix,VOLUME) ] * fnorm + disc2[_GWI(mu,ix,VOLUME) ];
work[_GWI(mu,ix,VOLUME)+1] = disc[_GWI(mu,ix,VOLUME)+1] * fnorm + disc2[_GWI(mu,ix,VOLUME)+1];
}
}
/* save the result in position space */
sprintf(filename, "cvc_hpe5_X.%.4d.%.4d", Nconf, count);
sprintf(contype, "cvc-disc-all-hpe-05-X");
write_lime_contraction(work, filename, 64, 4, contype, Nconf, count);
/*
sprintf(filename, "cvc_hpe5_Xascii.%.4d.%.4d", Nconf, count);
write_contraction(work, NULL, filename, 4, 2, 0);
*/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/* Fourier transform data, copy to work */
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(work+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
memcpy((void*)(work+_GWI(4+mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
memcpy((void*)in, (void*)(work+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(work+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
} /* of mu =0 ,..., 3*/
fnorm = 1. / (double)(T_global*LX*LY*LZ);
if(g_cart_id==0) fprintf(stdout, "# P-fnorm = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(work+_GWI(mu,0,VOLUME));
cp2 = (complex*)(work+_GWI(4+nu,0,VOLUME));
cp3 = (complex*)(work+_GWI(8+4*mu+nu,0,VOLUME));
for(x0=0; x0<T; x0++) {
q[0] = (double)(x0+Tstart) / (double)T_global;
for(x1=0; x1<LX; x1++) {
q[1] = (double)(x1) / (double)LX;
for(x2=0; x2<LY; x2++) {
q[2] = (double)(x2) / (double)LY;
for(x3=0; x3<LZ; x3++) {
q[3] = (double)(x3) / (double)LZ;
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI * (q[mu]-q[nu]) );
w.im = sin( M_PI * (q[mu]-q[nu]) );
_co_eq_co_ti_co(&w1, cp1, cp2);
_co_eq_co_ti_co(cp3, &w1, &w);
_co_ti_eq_re(cp3, fnorm);
cp1++; cp2++; cp3++;
}
}
}
}
}
}
/* save the result in momentum space */
sprintf(filename, "cvc_hpe5_P.%.4d.%.4d", Nconf, count);
sprintf(contype, "cvc-disc-all-hpe-05-P");
write_lime_contraction(work+_GWI(8,0,VOLUME), filename, 64, 16, contype, Nconf, count);
/*
sprintf(filename, "cvc_hpe5_Pascii.%.4d.%.4d", Nconf, count);
write_contraction(work+_GWI(8,0,VOLUME), NULL, filename, 16, 2, 0);
*/
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to save cvc results: %e seconds\n", retime-ratime);
} /* of count % Nsave == 0 */
} /* of loop on sid */
/***********************************************
* free the allocated memory, finalize
***********************************************/
free(g_gauge_field);
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
free_geometry();
fftw_free(in);
free(disc);
free(work);
#ifdef MPI
fftwnd_mpi_destroy_plan(plan_p);
fftwnd_mpi_destroy_plan(plan_m);
MPI_Finalize();
#else
fftwnd_destroy_plan(plan_p);
fftwnd_destroy_plan(plan_m);
#endif
return(0);
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix;
int sid;
double *disc = (double*)NULL;
double *work = (double*)NULL;
double *disc_diag = (double*)NULL;
double phase[4];
int verbose = 0;
int do_gt = 0;
char filename[100];
double ratime, retime;
double plaq;
double spinor1[24], spinor2[24], U_[18];
complex w, w1, psi1[4], psi2[4];
FILE *ofs;
fftw_complex *in=(fftw_complex*)NULL;
#ifdef MPI
fftwnd_mpi_plan plan_p, plan_m;
int *status;
#else
fftwnd_plan plan_p, plan_m;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vgf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'g':
do_gt = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
set_default_input_values();
if(filename_set==0) strcpy(filename, "cvc.input");
/* read the input file */
read_input(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
/* initialize MPI parameters */
mpi_init(argc, argv);
#ifdef MPI
if((status = (int*)calloc(g_nproc, sizeof(int))) == (int*)NULL) {
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(7);
}
#endif
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifdef MPI
plan_p = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_BACKWARD, FFTW_MEASURE);
plan_m = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_FORWARD, FFTW_MEASURE);
fftwnd_mpi_local_sizes(plan_p, &T, &Tstart, &l_LX_at, &l_LXstart_at, &FFTW_LOC_VOLUME);
#else
plan_p = fftwnd_create_plan(4, dims, FFTW_BACKWARD, FFTW_MEASURE | FFTW_IN_PLACE);
plan_m = fftwnd_create_plan(4, dims, FFTW_FORWARD, FFTW_MEASURE | FFTW_IN_PLACE);
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] fftw parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(2);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
/* read the gauge field */
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
#ifdef MPI
xchange_gauge();
#endif
/* measure the plaquette */
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value: %25.16e\n", plaq);
/* allocate memory for the spinor fields */
no_fields = 2;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/* allocate memory for the contractions */
disc = (double*)calloc(8*VOLUME, sizeof(double));
work = (double*)calloc(20*VOLUME, sizeof(double));
if( (disc==(double*)NULL) || (work==(double*)NULL) ) {
fprintf(stderr, "could not allocate memory for disc/work\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
if(g_subtract == 1) {
/* allocate memory for disc_diag */
disc_diag = (double*)calloc(20*VOLUME, sizeof(double));
if( disc_diag == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc_diag\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(8);
}
for(ix=0; ix<20*VOLUME; ix++) disc_diag[ix] = 0.;
}
/* prepare Fourier transformation arrays */
in = (fftw_complex*)malloc(FFTW_LOC_VOLUME*sizeof(fftw_complex));
if(in==(fftw_complex*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(4);
}
if(g_resume==1) { /* read current disc from file */
sprintf(filename, ".outcvc_current.%.4d", Nconf);
c = read_contraction(disc, &count, filename, 4);
if( (g_subtract==1) && (c==0) ) {
sprintf(filename, ".outcvc_diag_current.%.4d", Nconf);
c = read_contraction(disc_diag, (int*)NULL, filename, 10);
}
#ifdef MPI
MPI_Gather(&c, 1, MPI_INT, status, 1, MPI_INT, 0, g_cart_grid);
if(g_cart_id==0) {
/* check the entries in status */
for(i=0; i<g_nproc; i++)
if(status[i]!=0) { status[0] = 1; break; }
}
MPI_Bcast(status, 1, MPI_INT, 0, g_cart_grid);
if(status[0]==1) {
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
count = 0;
}
#else
if(c != 0) {
fprintf(stdout, "could not read current disc; start new\n");
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
if(g_subtract==1) for(ix=0; ix<20*VOLUME; ix++) disc_diag[ix] = 0.;
count = 0;
}
#endif
if(g_cart_id==0) fprintf(stdout, "starting with count = %d\n", count);
} /* of g_resume == 1 */
/* start loop on source id.s */
for(sid=g_sourceid; sid<=g_sourceid2; sid++) {
/* read the new propagator */
/* sprintf(filename, "%s.%.4d.%.2d", filename_prefix, Nconf, sid); */
sprintf(filename, "source.%.4d.%.2d.inverted", Nconf, sid);
if(format==0) {
if(read_lime_spinor(g_spinor_field[1], filename, 0) != 0) break;
}
else if(format==1) {
if(read_cmi(g_spinor_field[1], filename) != 0) break;
}
count++;
xchange_field(g_spinor_field[1]);
/* calculate the source: apply Q_phi_tbc */
Q_phi_tbc(g_spinor_field[0], g_spinor_field[1]);
xchange_field(g_spinor_field[0]);
/*
sprintf(filename, "%s.source.%.2d", filename, g_cart_id);
ofs = fopen(filename, "w");
printf_spinor_field(g_spinor_field[0], ofs);
fclose(ofs);
*/
/* add new contractions to (existing) disc */
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
for(ix=0; ix<VOLUME; ix++) { /* loop on lattice sites */
for(mu=0; mu<4; mu++) { /* loop on Lorentz index of the current */
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
/* first contribution */
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[_GJI(ix, mu) ] -= 0.25 * w.re;
disc[_GJI(ix, mu)+1] -= 0.25 * w.im;
if(g_subtract==1) {
work[_GWI(mu,ix,VOLUME) ] = -0.25 * w.re;
work[_GWI(mu,ix,VOLUME)+1] = -0.25 * w.im;
}
/* second contribution */
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc[_GJI(ix, mu) ] -= 0.25 * w.re;
disc[_GJI(ix, mu)+1] -= 0.25 * w.im;
if(g_subtract==1) {
work[_GWI(mu,ix,VOLUME) ] -= 0.25 * w.re;
work[_GWI(mu,ix,VOLUME)+1] -= 0.25 * w.im;
}
}
}
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "[%2d] contractions in %e seconds\n", g_cart_id, retime-ratime);
if(g_subtract==1) {
/* add current contribution to disc_diag */
for(mu=0; mu<4; mu++) {
for(i=0; i<4; i++) phase[i] = (double)(i==mu);
memcpy((void*)in, (void*)&work[_GWI(mu,0,VOLUME)], 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
w.im = -sin( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
_co_eq_co_ti_co(&w1, &in[ix], &w);
work[_GWI(4+mu,ix,VOLUME) ] = w1.re;
work[_GWI(4+mu,ix,VOLUME)+1] = w1.im;
}
}
}
}
memcpy((void*)in, (void*)&work[_GWI(mu,0,VOLUME)], 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
w.im = sin( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
_co_eq_co_ti_co(&w1, &in[ix], &w);
work[_GWI(mu,ix,VOLUME) ] = w1.re;
work[_GWI(mu,ix,VOLUME)+1] = w1.im;
}
}
}
}
} /* of mu */
for(ix=0; ix<VOLUME; ix++) {
i=-1;
for(mu=0; mu<4; mu++) {
for(nu=mu; nu<4; nu++) {
i++;
_co_eq_co_ti_co(&w, (complex*)&work[_GWI(mu,ix,VOLUME)], (complex*)&work[_GWI(4+nu,ix,VOLUME)]);
disc_diag[_GWI(ix,i,10) ] += w.re;
disc_diag[_GWI(ix,i,10)+1] += w.im;
}
}
}
} /* of g_subtract == 1 */
/* save results for count = multiple of Nsave */
if(count%Nsave == 0) {
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id == 0) fprintf(stdout, "save results for count = %d\n", count);
/* save the result in position space */
sprintf(filename, "outcvc_X.%.4d.%.4d", Nconf, count);
write_contraction(disc, NULL, filename, 4, 1, 0);
/* Fourier transform data, copy to work */
for(mu=0; mu<4; mu++) {
for(i=0; i<4; i++) phase[i] = (double)(i==mu);
for(ix=0; ix<VOLUME; ix++) {
in[ix].re = disc[_GJI(ix,mu) ];
in[ix].im = disc[_GJI(ix,mu)+1];
}
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
w.im = -sin( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
_co_eq_co_ti_co(&w1, &in[ix], &w);
work[_GWI(ix,4+mu,8) ] = w1.re / (double)count;
work[_GWI(ix,4+mu,8)+1] = w1.im / (double)count;
}
}
}
}
for(ix=0; ix<VOLUME; ix++) {
in[ix].re = disc[_GJI(ix, mu) ];
in[ix].im = disc[_GJI(ix, mu)+1];
}
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
w.im = sin( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
_co_eq_co_ti_co(&w1, &in[ix], &w);
work[_GWI(ix,mu,8) ] = w1.re / (double)count;
work[_GWI(ix,mu,8)+1] = w1.im / (double)count;
}
}
}
}
} /* of mu =0 ,..., 3*/
/* save the result in momentum space */
sprintf(filename, "outcvc_P.%.4d.%.4d", Nconf, count);
write_contraction(work, NULL, filename, 8, 1, 0);
/* calculate the correlations 00, 01, 02, 03, 11, 12, ..., 23, 33 */
for(ix=VOLUME-1; ix>=0; ix--) {
/* copy current data to auxilliary vector */
memcpy((void*)psi1, (void*)&work[_GWI(ix,0,8)], 8*sizeof(double));
memcpy((void*)psi2, (void*)&work[_GWI(ix,4,8)], 8*sizeof(double));
i = -1;
for(mu=0; mu<4; mu++) {
for(nu=mu; nu<4; nu++) {
i++;
_co_eq_co_ti_co(&w,&psi1[mu],&psi2[nu]);
if(g_subtract !=1 ) {
work[_GWI(ix,i,10) ] = w.re / (double)(T_global*LX*LY*LZ);
work[_GWI(ix,i,10)+1] = w.im / (double)(T_global*LX*LY*LZ);
}
else {
work[_GWI(ix,i,10) ] =
( w.re - disc_diag[_GWI(ix,i,10) ]/(double)(count*count) ) /
(double)(T_global*LX*LY*LZ);
work[_GWI(ix,i,10)+1] =
( w.im - disc_diag[_GWI(ix,i,10)+1]/(double)(count*count) ) /
(double)(T_global*LX*LY*LZ);
}
}
}
}
/* save current results to file */
sprintf(filename, "outcvc_final.%.4d.%.4d", Nconf, count);
write_contraction(work, (int*)NULL, filename, 10, 1, 0);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "[%2d] time to save results: %e seconds\n", g_cart_id, retime-ratime);
} /* of count % Nsave == 0 */
} /* of loop on sid */
/* write current disc to file */
sprintf(filename, ".outcvc_current.%.4d", Nconf);
write_contraction(disc, &count, filename, 4, 0, 0);
if(g_subtract == 1) {
/* write current disc_diag to file */
sprintf(filename, ".outcvc_diag_current.%.4d", Nconf);
write_contraction(disc_diag, (int*)NULL, filename, 10, 0, 0);
}
/* free the allocated memory, finalize */
free(g_gauge_field); g_gauge_field=(double*)NULL;
for(i=0; i<no_fields; i++) {
free(g_spinor_field[i]);
g_spinor_field[i] = (double*)NULL;
}
free(g_spinor_field); g_spinor_field=(double**)NULL;
free_geometry();
fftw_free(in);
free(disc);
free(work);
if(g_subtract==1) free(disc_diag);
#ifdef MPI
fftwnd_mpi_destroy_plan(plan_p);
fftwnd_mpi_destroy_plan(plan_m);
free(status);
MPI_Finalize();
#else
fftwnd_destroy_plan(plan_p);
fftwnd_destroy_plan(plan_m);
#endif
return(0);
}
