int main(){vsip_init((void*)0);
{
int i,j;
vsip_vview_f* a = vsip_vcreate_f(M,0);
vsip_vview_f* b = vsip_vcreate_f(N,0);
vsip_mview_f* r = vsip_mcreate_f(
vsip_vgetlength_f(a),
vsip_vgetlength_f(b),
VSIP_ROW,0);
vsip_vramp_f(1.0, 1.0 , b);
vsip_vramp_f(1.0, -2.0/(double)(N-1), a);
vsip_vouter_f(1,a, b, r);
printf("a = [\n");
for(i=0; i<vsip_vgetlength_f(a); i++)
printf("%6.3f%s\n", vsip_vget_f(a,i),
(i == vsip_vgetlength_f(a) -1) ? "":";");
printf("]\n");
printf("b = [\n");
for(i=0; i<vsip_vgetlength_f(b); i++)
printf("%6.3f%s\n", vsip_vget_f(b,i),
(i == vsip_vgetlength_f(a) -1) ? "":";");
printf("]\n");
printf("\na * b' = \n");
for(i=0; i<vsip_vgetlength_f(a); i++){
for(j=0; j<vsip_vgetlength_f(b); j++)
printf("%6.3f ",vsip_mget_f(r,i,j));
printf("\n");
}
printf("]\n");
/*destroy the vector views and any associated blocks*/
vsip_valldestroy_f(b);
vsip_valldestroy_f(a);
vsip_malldestroy_f(r);
} vsip_finalize((void*)0); return 0;
}
/* Below we implement example from "help interp1" in octave 2.9.9
* xf=[0:0.05:10]; yf = sin(2*pi*xf/5);
* xp=[0:10]; yp = sin(2*pi*xp/5);
* lin=interp1(xp,yp,xf);
*/
int main (int argc, const char * argv[])
{
int retval = vsip_init((void*)0);
vsip_length N0 = 11;
vsip_length N = 201;
vsip_length M = 3;
vsip_mview_f *yf = vsip_mcreate_f(M,N,VSIP_ROW,VSIP_MEM_NONE);
vsip_vview_f *xf = vsip_vcreate_f(N,VSIP_MEM_NONE);
vsip_vview_f *xp = vsip_vcreate_f(N0,VSIP_MEM_NONE);
vsip_mview_f *yp = vsip_mcreate_f(M,N0,VSIP_COL,VSIP_MEM_NONE);
vsip_vview_f *yp0 = vsip_mrowview_f(yp,0);
vsip_vview_f *yp1 = vsip_mrowview_f(yp,1);
vsip_vview_f *yp2 = vsip_mrowview_f(yp,2);
vsip_vramp_f(0.0,1.0,xp);
vsip_vramp_f(0.0,0.05,xf);
vsip_svmul_f(2.0/5.0 * M_PI,xp,yp0);
vsip_svadd_f(M_PI/8.0,yp0,yp1);
vsip_svadd_f(M_PI/8.0,yp1,yp2);
vsip_vsin_f(yp0,yp0);
vsip_vsin_f(yp1,yp1);
vsip_vsin_f(yp2,yp2);
printf("xp = ");VPRINT(xp);
printf("yp = ");MPRINT(yp);
printf("xf = ");VPRINT(xf);
vsip_minterp_linear_f(xp,yp,VSIP_ROW,xf,yf);
printf("linear = "); MPRINT(yf);
vsip_vdestroy_f(yp0);
vsip_vdestroy_f(yp1);
vsip_vdestroy_f(yp2);
vsip_valldestroy_f(xf);
vsip_valldestroy_f(xp);
vsip_malldestroy_f(yp);
vsip_malldestroy_f(yf);
vsip_finalize((void*)0);
return retval;
}
/* sign function as defined in http://www.netlib.org/lapack/lawnspdf/lawn148.pdf */
static vsip_scalar_f sign_f(vsip_scalar_f a_in)
{
if(a_in < 0.0)
return -1.0;
else
return 1.0;
} /* same */
static void biDiagPhaseToZero_f( svdObj_f *svd)
{
vsip_mview_f *L = svd->L;
vsip_vview_f *d = svd->d;
vsip_vview_f *f = svd->f;
vsip_mview_f *R = svd->R;
vsip_scalar_f eps0 = svd->eps0;
vsip_length n_d=vsip_vgetlength_f(d);
vsip_length n_f=vsip_vgetlength_f(f);
vsip_index i,j;
vsip_scalar_f ps;
vsip_scalar_f m;
vsip_vview_f *l = svd->ls_one;
vsip_vview_f *r = svd->rs_one;
for(i=0; i<n_d; i++){
ps=vsip_vget_f(d,i);
m = vsip_mag_f(ps);
ps=sign_f(ps);
if(m > eps0){
col_sv_f(L,l,i);vsip_svmul_f(ps,l,l);
vsip_vput_f(d,i,m);
if (i < n_f)
vsip_vput_f(f,i,ps*vsip_vget_f(f,i));
} else {
vsip_vput_f(d,i,0.0);
}
}
svdZeroCheckAndSet_f(eps0,d,f);
for (i=0; i<n_f-1; i++){
j=i+1;
ps = vsip_vget_f(f,i);
m = vsip_mag_f(ps);
ps=sign_f(ps);
col_sv_f(L, l, j);vsip_svmul_f(ps,l,l);
row_sv_f(R,r,j);vsip_svmul_f(ps,r,r);
vsip_vput_f(f,i,m);
vsip_vput_f(f,j,ps * vsip_vget_f(f,j));
}
j=n_f;
i=j-1;
ps=vsip_vget_f(f,i);
m=vsip_mag_f(ps);
ps=sign_f(ps);
vsip_vput_f(f,i,m);
col_sv_f(L, l, j);vsip_svmul_f(ps,l,l);
row_sv_f(R,r,j);vsip_svmul_f(ps,r,r);
}
static void phaseCheck_f(svdObj_f *svd)
{
biDiagPhaseToZero_f(svd);
}
void houseProd_f(vsip_vview_f *v, vsip_mview_f *A)
{
vsip_mattr_f a_atr;
vsip_vview_f *w;
vsip_mview_f *B;
vsip_mgetattrib_f(A,&a_atr);
B=vsip_mcreate_f(a_atr.col_length,a_atr.row_length,VSIP_ROW,VSIP_MEM_NONE);
w = vsip_vcreate_f(a_atr.row_length,VSIP_MEM_NONE);
vsip_scalar_f beta = 2.0/vsip_vdot_f(v,v);
vsip_vmprod_f(v,A,w);
vsip_vouter_f(beta,v,w,B);
vsip_msub_f(A,B,A);
vsip_valldestroy_f(w);
vsip_malldestroy_f(B);
} /* need to remove create */
void prodHouse_f(vsip_mview_f *A, vsip_vview_f *v)
{
vsip_mattr_f a_atr;
vsip_vview_f *w;
vsip_mview_f *B;
vsip_mgetattrib_f(A,&a_atr);
B=vsip_mcreate_f(a_atr.col_length,a_atr.row_length,VSIP_ROW,VSIP_MEM_NONE);
w = vsip_vcreate_f(a_atr.col_length,VSIP_MEM_NONE);
vsip_scalar_f beta = 2.0/vsip_vdot_f(v,v);
vsip_mvprod_f(A,v,w);
vsip_vouter_f(beta,w,v,B);
vsip_msub_f(A,B,A);
vsip_valldestroy_f(w);
vsip_malldestroy_f(B);
} /* need to remove create */
int main(int argc, char **argv)
{
vsip_vview_f *vw;
vsip_scalar_f *buf = 0;
vsip_dda_sync_policy pol = VSIP_DDA_SYNC_POLICY_IN;
vsip_length reqsz;
vsip_dda_vdata_f *data;
vsip_init((void*)0);
vw = vsip_vcreate_f(N, VSIP_MEM_NONE); /* TODO: Handle error */
if(0 != (reqsz = vsip_dda_vrequired_buffer_size_f(vw, 0, pol)))
buf = malloc(reqsz * sizeof(vsip_scalar_f));
data = vsip_dda_vdatacreate_f(vw, 0, pol, buf); /* TODO: Handle error */
/* TODO : Operate on the data object */
vsip_dda_vdatadestroy_f(data); /* TODO: Handle error */
vsip_valldestroy_f(vw);
vsip_finalize((void*)0);
return 0;
}
int main () {
vsip_init((void*)0);{
double time0, time_rc=0, time_cr=0;
int i;
/*int k; for(k=0; k<4; k++)*/{
vsip_cvview_f* vectorOut = vsip_cvcreate_f(L/2 + 1,VSIP_MEM_NONE);
vsip_vview_f* vectorIn = vsip_vcreate_f(L,VSIP_MEM_NONE);
vsip_vview_f* ivectorOut = vsip_vcreate_f(L,VSIP_MEM_NONE);
vsip_fft_f* fftplan = vsip_rcfftop_create_f(L,1.0/L,0,VSIP_ALG_SPACE);
vsip_fft_f* ifftplan = vsip_crfftop_create_f(L,(double)(1.0),0,VSIP_ALG_TIME);
char printbuffer[10][40];
/* char printbuffer[L/2][40]; */
vsip_vramp_f(0, 2 * M_PI * F, vectorIn);
vsip_vcos_f(vectorIn,vectorIn);
/* print vectorIn */
/* for(i=0; i<L/2; i++) */
for(i=0; i<10; i++)
sprintf(printbuffer[i],"%2d, %2d input-> %7.3f, %7.3f",
i, i+L/2, vsip_vget_f(vectorIn,i), vsip_vget_f(vectorIn,i+L/2));
/*find the fft*/
for(i=0; i<N; i++){
/* printf("%d\n",i);fflush(stdout); */
time0=VU_ansi_c_clock();
vsip_rcfftop_f(fftplan,vectorIn,vectorOut);
time_rc += (VU_ansi_c_clock() - time0);
/*find the inverse fft*/
time0=VU_ansi_c_clock();
vsip_crfftop_f(ifftplan,vectorOut,ivectorOut);
time_cr += (VU_ansi_c_clock() - time0);
}
/*print it */
/* for(i=0; i<L/2; i++) */
for(i=0; i<10; i++)
printf("%s fft -> (%9.4f, %9.4f) ifft-> %7.3f, %7.3f\n" ,
printbuffer[i],
vsip_real_f(vsip_cvget_f(vectorOut,i)),
vsip_imag_f(vsip_cvget_f(vectorOut,i)),
vsip_vget_f(ivectorOut,i),
vsip_vget_f(ivectorOut,i+L/2));
printf(" fft -> (%9.4f, %9.4f)\n",
vsip_real_f(vsip_cvget_f(vectorOut,i)),
vsip_imag_f(vsip_cvget_f(vectorOut,i)));
{/* test vsip_fft_getattr_f */
vsip_fft_attr_f attr;
vsip_fft_getattr_f(fftplan, &attr);
printf("%lu %lu %d %f %d\n",attr.input,attr.output,attr.place,attr.scale,attr.dir);
vsip_fft_getattr_f(ifftplan, &attr);
printf("%lu %lu %d %f %d\n",attr.input,attr.output,attr.place,attr.scale,attr.dir);
}
vsip_cvalldestroy_f(vectorOut);
vsip_valldestroy_f(vectorIn);
vsip_fft_destroy_f(fftplan);
printf("time_rc %f; time_cr %f\n",time_rc/(double)N, time_cr/(double)N);
}
}
vsip_finalize((void*)0);
return 0;
}
int main(){vsip_init((void*)0);
{ int i,j; /* counters */
vsip_vview_f *windowt = vsip_vcreate_hanning_f(Ns,0);
vsip_vview_f *windowp = vsip_vcreate_hanning_f(Mp,0);
vsip_vview_f *kernel =
vsip_vcreate_kaiser_f(Nfilter,kaiser,0);
vsip_fir_f *fir = vsip_fir_create_f(kernel,
VSIP_NONSYM,2 * Nn,2,VSIP_STATE_SAVE,0,0);
vsip_vview_f *t =vsip_vcreate_f(Ns,0); /*time vector*/
vsip_vview_f *noise[Nnoise];
vsip_vview_f *nv = vsip_vcreate_f(2 * Nn,0);
vsip_vview_f *tt = vsip_vcreate_f(Ns,0);
vsip_mview_f *data = vsip_mcreate_f(Mp,Ns,VSIP_ROW,0),
*rmview;
vsip_vview_f *data_v, *gram_v;
vsip_cvview_f *gram_data_v;
vsip_cmview_f *gram_data =
vsip_cmcreate_f(Mp,Ns/2 + 1,VSIP_COL,0);
vsip_mview_f *gram =
vsip_mcreate_f(Mp,Ns/2 + 1,VSIP_ROW,0);
vsip_mview_f *Xim =
vsip_mcreate_f(Mp,Mp+1,VSIP_ROW,0);
vsip_scalar_f alpha = (D * Fs) / c;
vsip_vview_f *m = vsip_vcreate_f(Mp,0);
vsip_vview_f *Xi = vsip_vcreate_f(Mp + 1,0);
vsip_randstate *state =
vsip_randcreate(15,1,1,VSIP_PRNG);
vsip_scalar_f w0 = 2 * M_PI * F0/Fs;
vsip_scalar_f w1 = 2 * M_PI * F1/Fs;
vsip_scalar_f w2 = 2 * M_PI * F2/Fs;
vsip_scalar_f w3 = 2 * M_PI * F3/Fs;
vsip_scalar_f cnst1 = M_PI/Nnoise;
vsip_offset offset0 = (vsip_offset)(alpha * Mp + 1);
vsip_fftm_f *rcfftmop_obj = /* time fft */
vsip_rcfftmop_create_f(Mp,Ns,1,VSIP_ROW,0,0);
vsip_fftm_f *ccfftmip_obj =
vsip_ccfftmip_create_f(Mp,Ns/2 +
1,VSIP_FFT_FWD,1,VSIP_COL,0,0);
vsip_vramp_f(0,1,m);
vsip_vramp_f(0,M_PI/Mp,Xi);
vsip_vcos_f(Xi,Xi);
vsip_vouter_f(alpha,m,Xi,Xim);
{ vsip_vview_f *gram_v = vsip_mrowview_f(gram,0);
vsip_vputlength_f(gram_v,Mp*(Ns/2 + 1));
vsip_vfill_f(0,gram_v);
vsip_vdestroy_f(gram_v);
}
for(j=0; j<Nnoise; j++){
noise[j] = vsip_vcreate_f(Nn,0);
vsip_vrandn_f(state,nv);
vsip_firflt_f(fir,nv,noise[j]);
vsip_svmul_f(12.0/(Nnoise),noise[j],noise[j]);
vsip_vputlength_f(noise[j],Ns);
}
vsip_vramp_f(0,1.0,t); /* time vector */
for(i=0; i<Mp; i++){
vsip_scalar_f Xim_val = vsip_mget_f(Xim,i,Theta_o);
data_v = vsip_mrowview_f(data,i);
vsip_vsmsa_f(t,w0,-w0 * Xim_val,tt);
vsip_vcos_f(tt,data_v); /*F0 time series */
vsip_vsmsa_f(t,w1,-w1 * Xim_val,tt);
vsip_vcos_f(tt,tt); /*F1 time series */
vsip_vadd_f(tt,data_v,data_v);
vsip_vsmsa_f(t,w2,-w2 * Xim_val,tt);
vsip_vcos_f(tt,tt); /*F2 time series */
vsip_vadd_f(tt,data_v,data_v);
vsip_vsmsa_f(t,w3,-w3 * Xim_val,tt);
vsip_vcos_f(tt,tt); /*F3 time series */
vsip_svmul_f(3.0,tt,tt); /* scale by 3.0 */
vsip_vadd_f(tt,data_v,data_v);
vsip_svmul_f(3,data_v,data_v);
for(j=0; j<Nnoise; j++){
/* simple time delay beam forming for noise */
vsip_vputoffset_f(noise[j],offset0 +
(int)( i * alpha * cos(j * cnst1)));
vsip_vadd_f(noise[j],data_v,data_v);
}
/* need to destroy before going on to next phone */
vsip_vdestroy_f(data_v);
}
/* window the data and the array to reduce sidelobes */
vsip_vmmul_f(windowt,data,VSIP_ROW,data);
vsip_vmmul_f(windowp,data,VSIP_COL,data);
/* do ffts */
vsip_rcfftmop_f(rcfftmop_obj,data,gram_data);
vsip_ccfftmip_f(ccfftmip_obj,gram_data);
{ /* scale gram to db, min 0 max 255 */
vsip_index ind;
gram_v = vsip_mrowview_f(gram,0);
gram_data_v = vsip_cmcolview_f(gram_data,0);
rmview = vsip_mrealview_f(gram_data);
vsip_vputlength_f(gram_v,Mp*(Ns/2 + 1));
vsip_cvputlength_f(gram_data_v,Mp*(Ns/2 + 1));
data_v = vsip_vrealview_f(gram_data_v);
vsip_vcmagsq_f(gram_data_v,data_v);
vsip_mcopy_f_f(rmview,gram);
vsip_vdestroy_f(data_v);
vsip_svadd_f(1.0 -
vsip_vminval_f(gram_v,&ind),gram_v,gram_v);
vsip_vlog10_f(gram_v,gram_v);
vsip_svmul_f(256.0 / vsip_vmaxval_f(gram_v,&ind),
gram_v,gram_v);/* scale */
/* reorganize the data to place zero in the
center for direction space */
data_v = vsip_vcloneview_f(gram_v);
vsip_vputlength_f(data_v,(Mp/2) * (Ns/2 + 1));
vsip_vputoffset_f(data_v,(Mp/2) * (Ns/2 + 1));
vsip_vputlength_f(gram_v,(Mp/2) * (Ns/2 + 1));
vsip_vswap_f(data_v,gram_v);
vsip_vdestroy_f(gram_v);
vsip_vdestroy_f(data_v);
vsip_cvdestroy_f(gram_data_v);
vsip_mdestroy_f(rmview);
}
VU_mprintgram_f(gram,"gram_output");
} vsip_finalize((void*)0); return 0;
}
int main () {
vsip_init((void*)0);
{ vsip_vview_f *kernel =
vsip_vcreate_kaiser_f(128,15.0,VSIP_MEM_NONE);
vsip_randstate *r_state =
vsip_randcreate(11,1,1,VSIP_NPRNG);
vsip_conv1d_f *conv;
vsip_fir_f *fir;
vsip_vview_f *data = vsip_vcreate_f(
2 * N_data,VSIP_MEM_NONE),
*noise = vsip_vcreate_f(
3 * N_data,VSIP_MEM_NONE),
*avg = vsip_vcreate_f(
4 * N_data,VSIP_MEM_NONE);
int i; vsip_length N_len;
vsip_vputlength_f(data,
(vsip_length)((N_data-1)/dec1)+1);
vsip_vputlength_f(avg,
(vsip_length)((N_data-1)/dec1)+1);
vsip_vputstride_f(data,2);
vsip_vputstride_f(avg,4);
vsip_vputlength_f(noise,N_data); vsip_vputstride_f(noise,3);
conv = vsip_conv1d_create_f(
kernel,VSIP_NONSYM,
N_data,dec1,VSIP_SUPPORT_SAME,0,0);
fir = vsip_fir_create_f(
kernel,VSIP_NONSYM, N_data,dec1,VSIP_STATE_NO_SAVE,0,0);
vsip_vfill_f(0,avg);
for(i=0; i<10; i++){
vsip_vrandn_f(r_state,noise);
vsip_convolve1d_f(conv,noise,data);
VU_vfrdB_f(data,1e-13);
vsip_vsma_f(data,0.1,avg,avg);
}
N_len = vsip_vgetlength_f(avg);
{ vsip_vview_f *x = vsip_vcreate_f(
N_len,VSIP_MEM_NONE);
vsip_vramp_f(-.5,1.0/(vsip_scalar_f)(N_len-1),x);
VU_vfprintxyg_f("%8.6f %8.6f\n",x,avg,"conv_dec1");
vsip_vdestroy_f(x);
}
vsip_vfill_f(0,avg);
for(i=0; i<10; i++){
vsip_vrandn_f(r_state,noise);
vsip_firflt_f(fir,noise,data);
VU_vfrdB_f(data,1e-13);
vsip_vsma_f(data,0.1,avg,avg);
}
N_len = vsip_vgetlength_f(avg);
{ vsip_vview_f *x = vsip_vcreate_f(
N_len,VSIP_MEM_NONE);
vsip_vramp_f(-.5,1.0/(vsip_scalar_f)(N_len-1),x);
VU_vfprintxyg_f("%8.6f %8.6f\n",x,avg,"fir_dec1");
vsip_vdestroy_f(x);
}
vsip_conv1d_destroy_f(conv);
vsip_fir_destroy_f(fir);
conv = vsip_conv1d_create_f(
kernel,VSIP_NONSYM,
N_data,dec3,VSIP_SUPPORT_SAME,0,0);
fir = vsip_fir_create_f(
kernel,VSIP_NONSYM,
N_data,dec3,VSIP_STATE_NO_SAVE,0,0);
vsip_vputlength_f(data,
(vsip_length)((N_data-1)/dec3)+1);
vsip_vputlength_f(avg,
(vsip_length)((N_data-1)/dec3)+1);
vsip_vfill_f(0,avg);
for(i=0; i<10; i++){
vsip_vrandn_f(r_state,noise);
vsip_convolve1d_f(conv,noise,data);
VU_vfrdB_f(data,1e-13);
vsip_vsma_f(data,0.1,avg,avg);
}
N_len = vsip_vgetlength_f(avg);
{ vsip_vview_f *x = vsip_vcreate_f(
N_len,VSIP_MEM_NONE);
vsip_vramp_f(-.5,1.0/(vsip_scalar_f)(N_len - 1),x);
VU_vfprintxyg_f("%8.6f %8.6f\n",
x, avg,"conv_dec3");
vsip_vdestroy_f(x);
}
vsip_vfill_f(0,avg);
for(i=0; i<10; i++){
vsip_vrandn_f(r_state,noise);
vsip_firflt_f(fir,noise,data);
VU_vfrdB_f(data,1e-13);
vsip_vsma_f(data,0.1,avg,avg);
}
N_len = vsip_vgetlength_f(avg);
{ vsip_vview_f *x = vsip_vcreate_f(
N_len,VSIP_MEM_NONE);
vsip_vramp_f(-.5,1.0/(vsip_scalar_f)(N_len-1),x);
VU_vfprintxyg_f("%8.6f %8.6f\n",
x, avg,"fir_dec3");
vsip_vdestroy_f(x);
}
N_len = vsip_vgetlength_f(kernel);
{ vsip_vview_f *x = vsip_vcreate_f(
N_len,VSIP_MEM_NONE);
vsip_vramp_f(0,1,x);
VU_vfprintxyg_f("%8.6f %8.6f\n",
x,kernel,"kaiser_window");
vsip_vramp_f(-.5,1.0/(vsip_scalar_f)(N_len-1),x);
VU_vfrdB_f(kernel,1e-20);
VU_vfprintxyg_f("%8.6f %8.6f\n",
x,kernel,"Freq_Resp_Kaiser");
vsip_vdestroy_f(x);
}
vsip_randdestroy(r_state);
vsip_valldestroy_f(kernel);
vsip_conv1d_destroy_f(conv);vsip_fir_destroy_f(fir);
vsip_valldestroy_f(data); vsip_valldestroy_f(noise);
vsip_valldestroy_f(avg);
} vsip_finalize((void*)0); return 0;
}/*end of main program *******************************/
/* Declarations */
#define A_ROWS 20
#define A_COLS 10
#define NDECOMP 15
#define QR_BIG_BUFFER_SIZE 1000000
vsip_vview_f *bx;
int iD;
void *recvBuffer;
void *qrBufferA,
*qrBufferB;
size_t qrBytes;
vsip_qr_f *qrdObject = NULL;
/* Construction: allocate memory for the vectors and the buffers */
bx = vsip_vcreate_f(A_COLS, VSIP_MEM_NONE);
qrBufferA = (void *) malloc(QR_BIG_BUFFER_SIZE);
qrBufferB = (void *) malloc(QR_BIG_BUFFER_SIZE);
/* Loop over the number of decompositions to be done */
for (iD = 0; iD < NDECOMP; iD++)
{
/* Choose the buffer to be used by this loop
/* Perform double-buffering: on even iterations use buffer A,
on odd iterations use buffer B */
if ((iD % 2) == 0)
{
recvBuffer = qrBufferA;
}
else
{
recvBuffer = qrBufferB;
int main() {
vsip_init((void*)0);{
vsip_mview_f *A = vsip_mcreate_f(NN,NN,VSIP_COL,0);
vsip_vview_f *x0 = vsip_vcreate_f(NN,0);
vsip_vview_f *b = vsip_vcreate_f(NN,0);
vsip_mview_f *X = vsip_mcreate_f(NN,3,VSIP_ROW,0);
vsip_mview_f *XT = vsip_mcreate_f(NN,3,VSIP_COL,0);
vsip_vramp_f(1,1,x0); vsip_vmul_f(x0,x0,x0);
vsip_mput_f(A,0,0,-3); vsip_mput_f(A,0,1,7); vsip_mput_f(A,0,2,10); vsip_mput_f(A,0,3,12);
vsip_mput_f(A,1,0,0); vsip_mput_f(A,1,1,13); vsip_mput_f(A,1,2,18); vsip_mput_f(A,1,3,6);
vsip_mput_f(A,2,0,2); vsip_mput_f(A,2,1,-9); vsip_mput_f(A,2,2,6); vsip_mput_f(A,2,3,3);
vsip_mput_f(A,3,0,1); vsip_mput_f(A,3,1,2); vsip_mput_f(A,3,2,3); vsip_mput_f(A,3,3,4);
{int i,j; printf("A = [\n"); for(i=0; i<NN; i++){
for(j=0; j<NN; j++) printf("%9.2f%s",vsip_mget_f(A,i,j),(j == NN-1) ? "":",");
printf(";\n");
}
printf("]\n");
}
{ int k;
vsip_vview_f *x;
vsip_length L = vsip_mgetrowlength_f(X);
for(k=0; k<L; k++){
x = vsip_mcolview_f(X,k);
vsip_mvprod_f(A,x0,b);
vsip_vcopy_f_f(b,x);
vsip_svmul_f(2.0,x0,x0);
vsip_vdestroy_f(x);
}
{int i,j; printf("X = [\n");for(i=0; i<NN; i++){
for(j=0; j<3; j++) printf("%9.2f%s",vsip_mget_f(X,i,j),(j == 2) ? "":",");
printf(";\n");
}
printf("]\n");
}
{
vsip_lu_f* luAop = vsip_lud_create_f(NN);
vsip_mcopy_f_f(X,XT);
if(luAop == NULL) exit(1);
vsip_lud_f(luAop,A);
{ vsip_lu_attr_f attr;
vsip_lud_getattr_f(luAop,&attr);
printf("lud size %lu\n",attr.n);
}
vsip_lusol_f(luAop,0,X);
vsip_lusol_f(luAop,1,XT);
vsip_lud_destroy_f(luAop);
}
}
{int i,j; printf("A\\X\n"); for(i=0; i<NN; i++){
for(j=0; j<3; j++) printf("%9.2f%s",vsip_mget_f(X,i,j),(j == 2) ? "":",");
printf(";\n");
}
}
{int i,j; printf("A'\\X\n"); for(i=0; i<NN; i++){
for(j=0; j<3; j++) printf("%9.2f%s",vsip_mget_f(XT,i,j),(j == 2) ? "":",");
printf(";\n");
}
}
{
vsip_valldestroy_f(b);
vsip_valldestroy_f(x0);
vsip_malldestroy_f(X);
vsip_malldestroy_f(A);
}
}vsip_finalize((void*)0);return 1;
}
int main(){vsip_init((void*)0);
{
vsip_mview_f *Adummy = vsip_mcreate_f(5*NN,5*NN,VSIP_COL,0);
vsip_mview_f *A = vsip_msubview_f(Adummy,3,2,NN,NN);
vsip_vview_f *x0 = vsip_vcreate_f(NN,0);
vsip_mview_f *X = vsip_mcreate_f(NN,3,VSIP_ROW,0);
vsip_mview_f *XT = vsip_mcreate_f(NN,3,VSIP_COL,0);
vsip_mputrowstride_f(A,2*vsip_mgetrowstride_f(A));
vsip_mputcolstride_f(A,3*vsip_mgetcolstride_f(A));
/* matrix data */
vsip_mput_f(A,0,0,0.5); vsip_mput_f(A,0,1,7); vsip_mput_f(A,0,2,10); vsip_mput_f(A,0,3,12);
vsip_mput_f(A,0,4,-3); vsip_mput_f(A,0,5,0); vsip_mput_f(A,0,6,.05);
vsip_mput_f(A,1,0,2); vsip_mput_f(A,1,1,13); vsip_mput_f(A,1,2,18); vsip_mput_f(A,1,3,6);
vsip_mput_f(A,1,4,0); vsip_mput_f(A,1,5,130); vsip_mput_f(A,1,6,8);
vsip_mput_f(A,2,0,3); vsip_mput_f(A,2,1,-9); vsip_mput_f(A,2,2,2); vsip_mput_f(A,2,3,3);
vsip_mput_f(A,2,4,2); vsip_mput_f(A,2,5,-9); vsip_mput_f(A,2,6,6);
vsip_mput_f(A,3,0,4); vsip_mput_f(A,3,1,2); vsip_mput_f(A,3,2,2); vsip_mput_f(A,3,3,4);
vsip_mput_f(A,3,4,1); vsip_mput_f(A,3,5,2); vsip_mput_f(A,3,6,3);
vsip_mput_f(A,4,0,.2); vsip_mput_f(A,4,1,2); vsip_mput_f(A,4,2,9); vsip_mput_f(A,4,3,4);
vsip_mput_f(A,4,4,1); vsip_mput_f(A,4,5,2); vsip_mput_f(A,4,6,3);
vsip_mput_f(A,5,0,.1); vsip_mput_f(A,5,1,2); vsip_mput_f(A,5,2,.3); vsip_mput_f(A,5,3,4);
vsip_mput_f(A,5,4,1); vsip_mput_f(A,5,5,2); vsip_mput_f(A,5,6,3);
vsip_mput_f(A,6,0,.01); vsip_mput_f(A,6,1,.2); vsip_mput_f(A,6,2,3); vsip_mput_f(A,6,3,4);
vsip_mput_f(A,6,4,1); vsip_mput_f(A,6,5,2); vsip_mput_f(A,6,6,3);
{int i,j; printf("A = [\n"); for(i=0; i<NN; i++){
for(j=0; j<NN; j++) printf("%9.2f%s",vsip_mget_f(A,i,j),(j == NN-1) ? "":",");
printf(";\n");
}
printf("]\n");
}
{ /* were solving for NTRANS Ax = B */
/* use a known X, calculate B using Ax */
int k;
vsip_vview_f *x;
vsip_mview_f *AT = vsip_mtransview_f(A);
vsip_length L = vsip_mgetrowlength_f(X);
vsip_vramp_f(1,1,x0);
for(k=0; k<L; k++){
x = vsip_mcolview_f(X,k);
vsip_mvprod_f(A,x0,x);
vsip_svmul_f(2.0,x0,x0);
vsip_vdestroy_f(x);
}
vsip_vramp_f(1,1,x0);
for(k=0; k<L; k++){
x = vsip_mcolview_f(XT,k);
vsip_mvprod_f(AT,x0,x);
vsip_svmul_f(2.0,x0,x0);
vsip_vdestroy_f(x);
}
vsip_mdestroy_f(AT);
printf("X = 1 2 4\n 2 4 8\n 3 6 12\n 4 8 16\n 5 10 20\n 6 12 24\n 7 14 28\n");
{
int i,j; printf("B = [\n");for(i=0; i<NN; i++){
for(j=0; j<3; j++) printf("%9.2f%s",vsip_mget_f(X,i,j),(j == 2) ? "":",");
printf(";\n");
}
printf("]\n");
}
{
/* then solve for B to see if we get X back */
vsip_lu_f* luAop = vsip_lud_create_f(NN);
if(luAop == NULL) exit(1);
vsip_lud_f(luAop,A);
{ vsip_lu_attr_f attr;
vsip_lud_getattr_f(luAop,&attr);
printf("lud size %lu\n",attr.n);
}
vsip_lusol_f(luAop,VSIP_MAT_NTRANS,X);
vsip_lusol_f(luAop,VSIP_MAT_TRANS,XT);
vsip_lud_destroy_f(luAop);
}
}
{int i,j; printf("A\\X\n"); for(i=0; i<NN; i++){
for(j=0; j<3; j++) printf("%9.2f%s",vsip_mget_f(X,i,j),(j == 2) ? "":",");
printf(";\n");
}
}
{int i,j; printf("A'\\X\n"); for(i=0; i<NN; i++){
for(j=0; j<3; j++) printf("%9.2f%s",vsip_mget_f(XT,i,j),(j == 2) ? "":",");
printf(";\n");
}
}
{
vsip_malldestroy_f(XT);
vsip_valldestroy_f(x0);
vsip_malldestroy_f(X);
vsip_mdestroy_f(A);
vsip_malldestroy_f(Adummy);
}
} vsip_finalize((void*)0); return 1;
}
int main(){
int init = vsip_init((void*)0);
int i,j, cholsol_retval,chold_retval;
double t0 = VU_ansi_c_clock(); /* for doing some timeing */
vsip_cscalar_f czero = vsip_cmplx_f((vsip_scalar_f)0.0,(vsip_scalar_f)0.0);
vsip_cmview_f *A = vsip_cmcreate_f(N,N,VSIP_COL,0);
vsip_cmview_f *RU = vsip_cmcreate_f(N,N,VSIP_COL,0);
vsip_cmview_f *RL = vsip_cmcreate_f(N,N,VSIP_COL,0);
vsip_cmview_f *XB = vsip_cmcreate_f(N,M,VSIP_ROW,0);
vsip_cchol_f* chol = vsip_cchold_create_f(UPORLO,N); /* NOTE: UPORLO macro above main() */
/* to make sure we have a valid Positive Symetric define */
/* an upper triangular (RU) with positive pivots and */
/* zero below the main diagonal. */
/* Then initialize RL with hermitian of RU */
/* finally create A as the matrix product of RL and RU */
/* Initialise matrix RU */
/* time this */
t0 = VU_ansi_c_clock();
for (i=0; i<N; i++){
for(j = i; j < N; j++){
#ifdef OBNOXIOUS
/* make up some reasonably obnoxious data */
vsip_scalar_f a = cos(1.5/((j+1)*(i+1)))+sqrt(i*j);
vsip_scalar_f b = (i + j + 1) * cos(M_PI * a);
#else
/* the above was to obnoxious for bigger than about N = 10 */
/* the following works for N > 100 */
vsip_scalar_f a = 1; vsip_scalar_f b = 1;
#endif
if(i == j) /* fill diagonal */
vsip_cmput_f(RU,i,j, vsip_cmplx_f(sqrt(N) + sqrt(i),0));
else { /* fill off diagonal */
vsip_cmput_f(RU,i,j,vsip_cmplx_f(b,a));
vsip_cmput_f(RU,j,i,czero);
}
}
}
/* initialize RL */
vsip_cmherm_f(RU,RL);
#ifdef PRINT
VU_cmprintm_f("7.4",RU);
VU_cmprintm_f("7.4",RL);
#endif
printf("Matrix initialize for RU and RL = %f seconds\n",VU_ansi_c_clock() - t0);
/* initialize A */
/* this step will take a long time so time it */
t0 = VU_ansi_c_clock();
vsip_cmprod_f(RL,RU,A);
#ifdef OBNOXIOUS
for(i=0; i<N; i++){
vsip_cvview_f *aview = vsip_cmrowview_f(A,i);
vsip_cvrsdiv_f(aview,vsip_cmag_f(vsip_cvmeanval_f(aview)),aview);
vsip_cvdestroy_f(aview);
}
#endif
printf("Matrix multiply for initialization of A = %f seconds\n",VU_ansi_c_clock() - t0);
/* print A */
/* we only want to do this if A is something reasonable to print */
/* selected as an option in the make file */
#ifdef PRINT
printf("Matrix A =\n");
VU_cmprintm_f("4.2",A);
fflush(stdout);
#endif
/* initialise rhs */
/* start out with XB = {1,2,3,...,M} */
/* calculate what B must be using A */
/* then solve to see if we get XB back */
{ vsip_index i;
vsip_vview_f *y = vsip_vcreate_f(vsip_cmgetcollength_f(A),VSIP_MEM_NONE);
vsip_vview_f *x_r,*x_i;
vsip_cvview_f *x;
vsip_mview_f *A_r = vsip_mrealview_f(A),
*A_i = vsip_mimagview_f(A);
/* time this */
t0 = VU_ansi_c_clock();
for(i=0; i<M; i++){
vsip_vfill_f((vsip_scalar_f)i+1.0,y);
x = vsip_cmcolview_f(XB,i);
x_r = vsip_vrealview_f(x);
x_i = vsip_vimagview_f(x);
vsip_mvprod_f(A_r,y,x_r);
vsip_mvprod_f(A_i,y,x_i);
vsip_cvdestroy_f(x);
vsip_vdestroy_f(x_r);
vsip_vdestroy_f(x_i);
}
vsip_mdestroy_f(A_r);
vsip_mdestroy_f(A_i);
printf("Matrix init for B = %f seconds\n",VU_ansi_c_clock() - t0);
}
/* print XB */
/* we only want to do this if XB is something reasonable to print */
/* selected as an option in the make file */
#ifdef PRINT
printf("Matrix B = \n");
VU_cmprintm_f("7.4",XB);
fflush(stdout);
#endif
if(chol != NULL){
t0 = VU_ansi_c_clock(); /* we want to time the decomposition */
chold_retval = vsip_cchold_f(chol,A);
printf("time decomp %f\n",VU_ansi_c_clock() - t0);
printf("decompostion returns %d\n",chold_retval);
/* now do the solution */
t0 = VU_ansi_c_clock(); /* we want to time the solution */
cholsol_retval=vsip_ccholsol_f(chol,XB);
printf("time solution %f\n",VU_ansi_c_clock() - t0);
printf("cholsol returns %d\n",cholsol_retval);
/* print XB */
/* we only want to do this if XB is something reasonable to print */
/* selected as an option in the make file; otherwise */
/* we print a single row of XB if the matrix is to large since */
/* M is usally reasonable. Printed as a column vector */
#ifdef PRINT
printf("Matrix X = \n");
VU_cmprintm_f("7.4",XB);
fflush(stdout);
#else
{ /* pick a row in the middle */
vsip_cvview_f *x = vsip_cmrowview_f(XB,N/2);
printf("This output sould be 1,2,...,M\n");
VU_cvprintm_f("7.4",x);
fflush(stdout);
vsip_cvdestroy_f(x);
}
#endif
} else {
printf("failed to create cholesky object \n");
}
vsip_cmalldestroy_f(XB);
vsip_cmalldestroy_f(A);
vsip_cmalldestroy_f(RL);
vsip_cmalldestroy_f(RU);
vsip_cchold_destroy_f(chol);
vsip_finalize((void*)0);
return 1;
}
