int main(){vsip_init((void*)0);
{
vsip_cscalar_d alpha = vsip_cmplx_d(alpha_r,alpha_i);
vsip_cscalar_d beta = vsip_cmplx_d(beta_r,beta_i);
void VU_cmprint_d(vsip_cmview_d*);
void VU_cmfill_d(vsip_cmview_d*, vsip_cscalar_d);
vsip_cmview_d *A = vsip_cmcreate_d(M,N,VSIP_ROW,0),
*C = vsip_cmcreate_d(N,M,VSIP_ROW,0);
int row, col,i;
for(row=0;row<M;row++) for(col=0;col<N;col++)
vsip_cmput_d(A,row,col,vsip_cmplx_d(col,-row - sqrt(col*row)));
printf("\n A input \n");
VU_cmprint_d(A);
VU_cmfill_d(C,vsip_cmplx_d(0,0));
printf("\n C input \n");
VU_cmprint_d(C);
printf("alpha= %f %+fi, beta= %f %+fi,\n",
alpha_r,alpha_i,
beta_r,beta_i);
{
vsip_mat_op OpA = VSIP_MAT_HERM;
printf("OpA %i\n",OpA);
for(i=0; i<L; i++){
vsip_cgems_d(alpha,A,OpA,beta,C);
printf("C number %i\n",i); VU_cmprint_d(C);
}
}
vsip_cmalldestroy_d(A);
vsip_cmalldestroy_d(C);
}vsip_finalize((void*)0);return 0;
}
int vsip_cfirflt_d(
vsip_cfir_d *fir,
const vsip_cvview_d *xc,
const vsip_cvview_d *yc)
{
vsip_length nout,k;
vsip_cvview_d xx = *(xc),
yy = *(yc);
vsip_cvview_d H1 = *(fir->h),
H2 = *(fir->h);
vsip_cvview_d *x=&xx,*y=&yy;
vsip_cvview_d *h1=&H1,*h2=&H2;
vsip_offset oinc;
oinc = (vsip_offset)((vsip_stride)fir->D * x->stride);
/* calculate number of terms in y */
nout = (fir->N - fir->p);
nout = ((nout % fir->D) == 0) ? (nout / fir->D ) : (nout / fir->D + 1);
/* do overlap section */
k = 0;
x->length = fir->p + 1;
h1->length = fir->s->length;
h2->length = x->length;
h2->offset = h1->length;
while(x->length < fir->M){
vsip_cscalar_d a = vsip_cvdot_d(h1,fir->s);
vsip_cscalar_d b = vsip_cvdot_d(h2,x);
vsip_cvput_d(y,k++,vsip_cmplx_d(a.r + b.r,a.i + b.i));
x->length += fir->D;
fir->s->length -= fir->D;
fir->s->offset += fir->D;
h1->length = fir->s->length;
h2->length = x->length;
h2->offset = h1->length;
}
x->offset += (x->length - fir->M) * x->stride;
x->length = fir->M;
while(k < nout){ /* do the rest of the pieces */
vsip_cvput_d(y,k++,vsip_cvdot_d(fir->h,x));
x->offset += oinc;
}
{
vsip_stride temp_p = (fir->p % fir->D) - (fir->N % fir->D);
fir->p = ((temp_p < 0) ? (vsip_length)((vsip_stride)fir->D + temp_p) : (vsip_length)temp_p);
}
fir->s->offset = 0;
fir->s->length = (fir->state == VSIP_STATE_SAVE) ? fir->M - 1 - fir->p : fir->M -1;
x->length = fir->s->length;
/* fix by JMA 31/01/2000, incorrect offset calculation */
/* x->offset = xc->length - fir->s->length; */
x->offset = xc->offset + (xc->length - fir->s->length) * xc->stride;
if((fir->s->length > 0) && (fir->state == VSIP_STATE_SAVE))
VI_cvcopy_d_d(x,fir->s);
if(fir->state == VSIP_STATE_NO_SAVE) {
VI_cvfill_d(vsip_cmplx_d((vsip_scalar_d)0,(vsip_scalar_d)0),fir->s);
fir->p = 0;
}
return k;
}
int main(){vsip_init((void*)0);
{
vsip_cmview_d *A = vsip_cmcreate_d(M,P,VSIP_ROW,0),
*B = vsip_cmcreate_d(P,N,VSIP_ROW,0),
*R = vsip_cmcreate_d(M,N,VSIP_ROW,0);
int i, j;
for(i=0;i<M;i++) for(j=0;j<P;j++)
vsip_cmput_d(A,i,j,vsip_cmplx_d(i,1 + sqrt(i*j)));
for(i=0;i<P;i++) for(j=0;j<N;j++)
vsip_cmput_d(B,i,j,vsip_cmplx_d(1 , i * j));
printf("\n A input \n");
vsip_cmprodj_d(A,B,R);
{for(i=0;i<M;i++) {for(j=0;j<P;j++){
printf("%5.2f + %5.2fi%s ",
vsip_real_d(vsip_cmget_d(A,i,j)),
vsip_imag_d(vsip_cmget_d(A,i,j)),
(j == P-1) ? ";":",");
}
printf("\n");
}
}
printf("\n B input \n");
{for(i=0;i<P;i++) {for(j=0;j<N;j++){
printf("%5.2f + %5.2fi%s ",
vsip_real_d(vsip_cmget_d(B,i,j)),
vsip_imag_d(vsip_cmget_d(B,i,j)),
(j == N-1) ? ";":",");
}
printf("\n");
}
}
printf("\n R output = A * conj(B)\n");
{for(i=0;i<M;i++) {for(j=0;j<N;j++){
printf("%5.2f + %5.2fi%s ",
vsip_real_d(vsip_cmget_d(R,i,j)),
vsip_imag_d(vsip_cmget_d(R,i,j)),
(j == N-1) ? ";":",");
}
printf("\n");
}
}
vsip_cmalldestroy_d(A);
vsip_cmalldestroy_d(B);
vsip_cmalldestroy_d(R);
}vsip_finalize((void*)0);return 0;
}
int main()
{
int i;
/* define some data space */
vsip_cvview_d* dataComplex;
vsip_cvview_d* dataComplexQuotient;
vsip_init((void *)0);
dataComplex = vsip_cvcreate_d(L, VSIP_MEM_NONE);
dataComplexQuotient = vsip_cvcreate_d(L, VSIP_MEM_NONE);
/* put some complex data in dataComplex */
for(i = 0; i < L; i++)
vsip_cvput_d(dataComplex,i,
vsip_cmplx_d((double)(i * i),(double)(i+1)));
/*divide dataComplex by some denom and print the input and output */
vsip_cvrsdiv_d(dataComplex,denom,dataComplexQuotient);
for(i=0; i<L; i++)
printf("(%7.4f + %7.4fi) / %7.4f) = (%7.4f + %7.4fi)\n",
vsip_real_d(vsip_cvget_d(dataComplex,i)),
vsip_imag_d(vsip_cvget_d(dataComplex,i)),
denom,
vsip_real_d(vsip_cvget_d(dataComplexQuotient,i)),
vsip_imag_d(vsip_cvget_d(dataComplexQuotient,i)));
vsip_cblockdestroy_d(vsip_cvdestroy_d(dataComplex));
vsip_cblockdestroy_d(vsip_cvdestroy_d(dataComplexQuotient));
vsip_finalize((void *)0);
return 0;
}
static
void
VI_cgivens_d(
vsip_cscalar_d a,
vsip_cscalar_d b,
vsip_cscalar_d *c,
vsip_cscalar_d *s,
vsip_cscalar_d *r)
{
vsip_scalar_d am = vsip_cmag_d(a);
vsip_scalar_d bm = vsip_cmag_d(b);
c->i = 0.0;
if(am == 0.0){
*r = b;
c->r = 0.0;
s->r = 1; s->i = 0.0;
} else {
vsip_scalar_d scale = am + bm;
vsip_cscalar_d alpha = vsip_cmplx_d(a.r/am, a.i/am);
vsip_scalar_d scalesq = scale * scale;
vsip_scalar_d norm = scale * (vsip_scalar_d)sqrt((am*am)/scalesq + (bm * bm)/scalesq);
c->r =am/norm;
s->r = (alpha.r * b.r + alpha.i * b.i)/norm;
s->i = (-alpha.r * b.i + alpha.i * b.r)/norm;
r->r = alpha.r * norm; r->i = alpha.i * norm;
}
return;
}
int main(){vsip_init((void*)0);
{
void VU_vfprintyg_d(char*,vsip_vview_d*,char*);
void VU_vfreqswapIP_d(vsip_vview_d*);
vsip_vview_d* Cw = vsip_vcreate_cheby_d(Nlength,ripple,0);
vsip_fft_d *fft = vsip_ccfftip_create_d(Nlength,1.0,VSIP_FFT_FWD,0,0);
vsip_cvview_d* FCW = vsip_cvcreate_d(Nlength,0);
/*printf("CW = "); VU_vprintm_d("%6.8f ;\n",Cw); */
VU_vfprintyg_d("%6.8f\n",Cw,"Cheby_Window");
vsip_cvfill_d(vsip_cmplx_d(0,0),FCW);
{ vsip_vview_d *rv = vsip_vrealview_d(FCW);
vsip_vcopy_d_d(Cw,rv);
vsip_ccfftip_d(fft,FCW);
vsip_vcmagsq_d(FCW,rv);
{ vsip_index ind;
vsip_scalar_d max = vsip_vmaxval_d(rv,&ind);
vsip_scalar_d min = max/(10e12);
vsip_vclip_d(rv,min,max,min,max,rv);
}
vsip_vlog10_d(rv,rv);
vsip_svmul_d(10,rv,rv);
VU_vfreqswapIP_d(rv);
VU_vfprintyg_d("%6.8f\n",rv,"Cheby_Window_Frequency_Response");
vsip_vdestroy_d(rv);
}
vsip_fft_destroy_d(fft);
vsip_valldestroy_d(Cw);
vsip_cvalldestroy_d(FCW);
} vsip_finalize((void*)0); return 0;
}
int main(){vsip_init((void*)0);
{
vsip_cmview_d *A = vsip_cmcreate_d(M,P,VSIP_ROW,0),
*B = vsip_cmcreate_d(N,P,VSIP_ROW,0),
*R = vsip_cmcreate_d(M,N,VSIP_ROW,0);
int i, j;
for(i=0;i<M;i++) for(j=0;j<P;j++)
vsip_cmput_d(A,i,j,vsip_cmplx_d(i,1 + sqrt(i*j)));
for(i=0;i<N;i++) for(j=0;j<P;j++)
vsip_cmput_d(B,i,j,vsip_cmplx_d(1 , i * j));
printf("\n A input \n");
vsip_cmprodh_d(A,B,R);
{for(i=0;i<M;i++) {for(j=0;j<P;j++){
printf(": (%5.2f, %5.2f) ",
vsip_real_d(vsip_cmget_d(A,i,j)),
vsip_imag_d(vsip_cmget_d(A,i,j)));
}
printf(":\n");
}
}
printf("\n B input \n");
{for(i=0;i<N;i++) {for(j=0;j<P;j++){
printf(": (%5.2f, %5.2f) ",
vsip_real_d(vsip_cmget_d(B,i,j)),
vsip_imag_d(vsip_cmget_d(B,i,j)));
}
printf(":\n");
}
}
printf("\n R output \n");
{for(i=0;i<M;i++) {for(j=0;j<N;j++){
printf(": (%5.2f, %5.2f) ",
vsip_real_d(vsip_cmget_d(R,i,j)),
vsip_imag_d(vsip_cmget_d(R,i,j)));
}
printf(":\n");
}
}
vsip_cmalldestroy_d(A);
vsip_cmalldestroy_d(B);
vsip_cmalldestroy_d(R);
}vsip_finalize((void*)0);return 0;
}
int main(){vsip_init((void*)0);
{
/* vsip_cmview_d *B = vsip_cmcreate_d(P,N,VSIP_ROW,0);*/
vsip_scalar_d Re[20], Im[20];
vsip_cblock_d *Bblock = vsip_cblockbind_d(Re,Im,20,0);
vsip_cmview_d *B = vsip_cmbind_d(Bblock,0,4,5,1,4);
vsip_cvview_d *a = vsip_cvcreate_d(P,0),
*r = vsip_cvcreate_d(N,0);
int i, j;
for(i=0;i<P;i++)
vsip_cvput_d(a,i,vsip_cmplx_d(i,1 + sqrt(i)));
for(i=0;i<P;i++) for(j=0;j<N;j++)
vsip_cmput_d(B,i,j,vsip_cmplx_d(1 , i * j));
vsip_cvmprod_d(a,B,r);
printf("\n vector input \n v=[");
for(i=0;i<P;i++)
printf("(%5.2f %+5.2fi) ",
vsip_real_d(vsip_cvget_d(a,i)),
vsip_imag_d(vsip_cvget_d(a,i)));
printf("]\n");
printf("\n B input \n B =[\n");
{for(i=0;i<P;i++) {for(j=0;j<N;j++){
printf(" (%5.2f %+5.2fi) ",
vsip_real_d(vsip_cmget_d(B,i,j)),
vsip_imag_d(vsip_cmget_d(B,i,j)));
}
printf(";\n");
}
printf("]\n");
}
printf("\n vector output \n");
for(i=0;i<N;i++)
printf("(%5.2f, %5.2f) ",
vsip_real_d(vsip_cvget_d(r,i)),
vsip_imag_d(vsip_cvget_d(r,i)));
printf("\n");
vsip_cvalldestroy_d(a);
vsip_cmalldestroy_d(B);
vsip_cvalldestroy_d(r);
}vsip_finalize((void*)0);return 0;
}
static void VU_cmfillz_d(vsip_cmview_d* Z)
{
vsip_length m = vsip_cmgetcollength_d(Z);
vsip_length n = vsip_cmgetrowlength_d(Z);
vsip_stride i,j;
for(i=0; i<m; i++)
for(j=0; j<n; j++)
vsip_cmput_d(Z,i,j,vsip_cmplx_d(0.0,0.0));
return;
}
int main() {
vsip_init((void*)0);
{
vsip_cscalar_d alpha = vsip_cmplx_d(alpha_r,alpha_i),
beta = vsip_cmplx_d(beta_r,beta_i);
void VU_cmprint_d(vsip_cmview_d*);
void VU_cmfill_d(vsip_cmview_d*, vsip_cscalar_d);
vsip_cmview_d *A = vsip_cmcreate_d(M,N,VSIP_ROW,0),
*B = vsip_cmcreate_d(M,P,VSIP_ROW,0),
*C = vsip_cmcreate_d(N,P,VSIP_ROW,0);
int row, col,i;
for(row=0; row<M; row++) for(col=0; col<N; col++)
vsip_cmput_d(A,row,col,vsip_cmplx_d(col,-row));
for(row=0; row<M; row++) for(col=0; col<P; col++)
vsip_cmput_d(B,row,col,vsip_cmplx_d(row, 2*col));
printf("\n A input \n");
VU_cmprint_d(A);
printf("\n B input \n");
VU_cmprint_d(B);
VU_cmfill_d(C,vsip_cmplx_d(0,0));
printf("alpha= (%f %+fi), beta= (%f %+fi),\n",
vsip_real_d(alpha),vsip_imag_d(alpha),
vsip_real_d(beta),vsip_imag_d(beta));
{
vsip_mat_op OpA = VSIP_MAT_HERM;
vsip_mat_op OpB = VSIP_MAT_NTRANS;
printf("OpA %i OpB %i\n",OpA,OpB);
for(i=0; i<L; i++) {
vsip_cgemp_d(alpha,A,OpA,B,OpB,beta,C);
printf("C number %i\n",i);
VU_cmprint_d(C);
}
}
vsip_cmalldestroy_d(A);
vsip_cmalldestroy_d(B);
vsip_cmalldestroy_d(C);
}
vsip_finalize((void*)0);
return 0;
}
static vsip_cmview_d* VU_cI_d(vsip_length M)
{
vsip_cmview_d *I = vsip_cmcreate_d(M,M,VSIP_ROW,VSIP_MEM_NONE);
if(I != NULL){
vsip_cvview_d *row = vsip_cmrowview_d(I,0);
if(row != NULL){
vsip_cvputlength_d(row,(vsip_length)(M * M));
vsip_cvfill_d(vsip_cmplx_d(0.0,0.0),row);
vsip_cvputlength_d(row,M);
vsip_cvputstride_d(row,(vsip_stride) (M + 1));
vsip_cvfill_d(vsip_cmplx_d(1.0,0.0),row);
vsip_cvdestroy_d(row);
} else {
vsip_cmdestroy_d(I);
return (vsip_cmview_d*) NULL;
}
} else {
return (vsip_cmview_d*) NULL;
}
return I;
}
int main(int argc, char *argv[]){vsip_init((void*)0);
{ if(argc < 4){
printf("usage\nfirEx length decimation filterFile <indataFile >outputFile\n");
exit(0);
}
{
vsip_length N = (vsip_length)atoi(argv[1]);
int D = atoi(argv[2]);
vsip_scalar_vi n = 0,k = 0;
vsip_scalar_d hvr,hvi;
FILE *fh;
vsip_cfir_d *fir;
vsip_cvview_d *x = vsip_cvcreate_d(N,0),
*h,*y;
fh = fopen(argv[3],"r");
while(fscanf(fh,"%lf %lf",&hvr, &hvi) == 2) n++;
rewind(fh);
y = vsip_cvcreate_d(N/D + ((N % D) ? 1:0),0);
h = vsip_cvcreate_d(n,0);
n=0;
while(fscanf(fh,"%lf %lf",&hvr,&hvi) == 2){
vsip_cvput_d(h,n,vsip_cmplx_d(hvr,hvi));
n++;
}
fclose(fh);
/* fir = vsip_cfir_create_d(h,VSIP_SYM_EVEN_LEN_ODD,N,D,STATE,0,0); */
/* fir = vsip_cfir_create_d(h,VSIP_SYM_EVEN_LEN_EVEN,N,D,STATE,0,0); */
fir = vsip_cfir_create_d(h,VSIP_NONSYM,N,D,STATE,0,0);
n = VU_cscanstdin_d(x);
while(n == N){
k = vsip_cfirflt_d(fir,x,y);
VU_cprintstdout_d(y,0,k);
n = VU_cscanstdin_d(x);
}
if(n != 0){
k = vsip_cfirflt_d(fir,x,y);
VU_cprintstdout_d(y,0,k);
}
{/* test getattr */
vsip_cfir_attr attr;
vsip_cfir_getattr_d(fir,&attr);
printf("kernel %lu\n input %lu\n output %lu\ndecimation %d\n",
attr.kernel_len,attr.in_len,attr.out_len, attr.decimation);
}
vsip_cvalldestroy_d(y);
vsip_cvalldestroy_d(x);
vsip_cvalldestroy_d(h);
vsip_cfir_destroy_d(fir);
}vsip_finalize((void*)0);return 1;
}
}
vsip_cscalar_d (vsip_cvmeanval_d)(
const vsip_cvview_d* a) {
{
/* register */ vsip_length n = a->length;
vsip_stride cast = a->block->cstride;
vsip_scalar_d *apr = (vsip_scalar_d*) ((a->block->R->array) + cast * a->offset);
vsip_scalar_d *api = (vsip_scalar_d*) ((a->block->I->array) + cast * a->offset);
vsip_cscalar_d sum = vsip_cmplx_d((vsip_scalar_d)0,(vsip_scalar_d)0);
/* register */ vsip_stride ast = (cast * a->stride);
while(n-- > 0){
sum.r += *apr;
sum.i += *api;
apr += ast; api += ast;
}
sum.r /= a->length;
sum.i /= a->length;
return sum;
}
}
vsip_cscalar_d (vsip_cadd_d)(
vsip_cscalar_d x, vsip_cscalar_d y) {/* x + y*/
return vsip_cmplx_d(
(vsip_real_d(x) + vsip_real_d(y)),
(vsip_imag_d(x) + vsip_imag_d(y))); }
vsip_cscalar_d (vsip_conj_d)(vsip_cscalar_d x) { /* conj(x) */
return vsip_cmplx_d(vsip_real_d(x),
-(vsip_imag_d(x))); }
void vsip_ccorrelate1d_d(
const vsip_ccorr1d_d *cor,
vsip_bias bias,
const vsip_cvview_d *h,
const vsip_cvview_d *x,
const vsip_cvview_d *y)
{
vsip_cvview_d xx = *cor->x,
hh = *cor->h;
vsip_cvview_d *xt = &xx,
*ht = &hh;
xt->length = cor->x->length - x->length;
VI_cvfill_d(vsip_cmplx_d((vsip_scalar_d)0,(vsip_scalar_d)0),xt);
xt->offset = xt->length;
xt->length = x->length;
VI_cvcopy_d_d(x,xt);
xt->length = cor->x->length;
xt->offset = 0;
ht->length = cor->h->length - h->length;
ht->offset = h->length;
VI_cvfill_d(vsip_cmplx_d((vsip_scalar_d)0,(vsip_scalar_d)0),ht);
ht->offset = 0;
ht->length = h->length;
VI_cvcopy_d_d(h,ht);
vsip_ccfftip_d(cor->fft,cor->h);
vsip_ccfftip_d(cor->fft,cor->x);
vsip_cvjmul_d(cor->x,cor->h,cor->x);
vsip_cvconj_d(cor->x,cor->x);
vsip_rscvmul_d(1/(vsip_scalar_d)cor->N,cor->x,cor->x);
vsip_ccfftip_d(cor->fft,cor->x);
/* vsip_cvconj_d(cor->x,cor->x); */
switch(cor->support){
case VSIP_SUPPORT_FULL:
xt->offset = xt->length - cor->mn;
xt->length = y->length;
if(bias == VSIP_UNBIASED){
VI_cvunbiasfull_d(cor,xt,y);
} else {
VI_cvcopy_d_d(xt,y);
}
break;
case VSIP_SUPPORT_SAME:
xt->offset = xt->length - cor->mn + (cor->m-1)/2;
xt->length = y->length;
if(bias == VSIP_UNBIASED){
VI_cvunbiassame_d(cor,xt,y);
} else {
VI_cvcopy_d_d(xt,y);
}
break;
case VSIP_SUPPORT_MIN:
xt->offset = xt->length - cor->mn + cor->m - 1;
xt->length = y->length;
if(bias == VSIP_UNBIASED){
vsip_rscvmul_d(1.0/(vsip_scalar_d)cor->m,xt,y);
} else {
VI_cvcopy_d_d(xt,y);
}
break;
}
return;
}
/* note that matrix products may not be done in place */
void (vsip_cmprod_d)(
const vsip_cmview_d* a,
const vsip_cmview_d* b,
const vsip_cmview_d* c) {
/* Note that c => column major, r => row major */
/* Below decide input matrix majors ccc => 0 ccr=> 1, crc => 2, etc to rrr => 7 */
/* the resulting method is used to calculate */
unsigned method = (unsigned)(c->row_stride <= c->col_stride) +
(unsigned)(b->row_stride <= b->col_stride) * 2 +
(unsigned)(a->row_stride <= a->col_stride) * 4;
/* get the stride info */
vsip_stride a_st_r = a->row_stride * a->block->cstride,
a_st_c = a->col_stride * a->block->cstride,
b_st_r = b->row_stride * b->block->cstride,
b_st_c = b->col_stride * b->block->cstride,
c_st_r = c->row_stride * c->block->cstride,
c_st_c = c->col_stride * c->block->cstride;
/* get the length info */
vsip_length a_c_l = a->col_length, /* i_size */
c_r_l = c->row_length, /* j_size */
a_r_l = a->row_length; /* k_size */
/* define some scalars used in the calculations below */
vsip_cscalar_d a_scalar, b_scalar, temp = vsip_cmplx_d(0.0f,0.0f);
/* get the pointers to the input and output data spaces */
vsip_scalar_d *ap_r = (a->block->R->array) + a->offset * a->block->cstride,
*ap_i = (a->block->I->array) + a->offset * a->block->cstride,
*bp_r = (b->block->R->array) + b->offset * b->block->cstride,
*bp_i = (b->block->I->array) + b->offset * b->block->cstride,
*cp_r = (c->block->R->array) + c->offset * c->block->cstride,
*cp_i = (c->block->I->array) + c->offset * c->block->cstride;
/* some additional pointers to store initial data */
vsip_scalar_d *ap0_r = ap_r, *ap0_i = ap_i,
*bp0_r = bp_r, *bp0_i = bp_i,
*cp0_r = cp_r, *cp0_i = cp_i;
vsip_scalar_d *bp1_r = bp_r, *bp1_i = bp_i;
/* initialize output matrix to zero */
VI_cmfill_d(temp,c); /* initilize output matrix to zero */
/* select multiply routine bassed on majors of input matrices */
switch(method){
case 0 : { /* ccc */
vsip_scalar_d *ap_ik_r, *ap_0k_r = ap0_r, *ap_ik_i, *ap_0k_i = ap0_i;
vsip_scalar_d *bp_kj_r, *bp_0j_r = bp0_r, *bp_kj_i, *bp_0j_i = bp0_i;
vsip_scalar_d *cp_ij_r, *cp_0j_r = cp0_r, *cp_ij_i, *cp_0j_i = cp0_i;
while(c_r_l-- > 0){ /* j */
cp_ij_r = cp_0j_r; cp_ij_i = cp_0j_i;
bp_kj_r = bp_0j_r; bp_kj_i = bp_0j_i;
while(a_r_l-- > 0){ /* k */
b_scalar.r = *bp_kj_r; b_scalar.i = *bp_kj_i;
bp_kj_r += b_st_c; bp_kj_i += b_st_c;
ap_ik_r = ap_0k_r; ap_ik_i = ap_0k_i;
while(a_c_l-- > 0){ /* i */
*cp_ij_r = *ap_ik_r * b_scalar.r - *ap_ik_i * b_scalar.i + *cp_ij_r;
*cp_ij_i = *ap_ik_r * b_scalar.i + *ap_ik_i * b_scalar.r + *cp_ij_i;
cp_ij_r += c_st_c; cp_ij_i += c_st_c;
ap_ik_r += a_st_c; ap_ik_i += a_st_c;
} a_c_l = a->col_length;
ap_0k_r += a_st_r; ap_0k_i += a_st_r;
cp_ij_r = cp_0j_r; cp_ij_i = cp_0j_i;
} a_r_l = a->row_length;
cp_0j_r += c_st_r; cp_0j_i += c_st_r;
bp_0j_r += b_st_r; bp_0j_i += b_st_r;
ap_0k_r = ap0_r; ap_0k_i = ap0_i;
}
} break;
case 1 : { /* ccr */
vsip_scalar_d *ap_ik_r, *ap_i0_r = ap0_r, *ap_ik_i, *ap_i0_i = ap0_i;
vsip_scalar_d *bp_kj_r, *bp_0j_r, *bp_kj_i, *bp_0j_i;
vsip_scalar_d *cp_ij_r, *cp_i0_r = cp0_r, *cp_ij_i, *cp_i0_i = cp0_i;
while(a_c_l-- > 0){ /* i */
cp_ij_r = cp_i0_r; cp_ij_i = cp_i0_i;
bp_0j_r = bp0_r; bp_0j_i = bp0_i;
while(c_r_l-- > 0){ /* j */
ap_ik_r = ap_i0_r;
ap_ik_i = ap_i0_i;
temp = vsip_cmplx_d(0.0f,0.0f);
bp_kj_r = bp_0j_r; bp_kj_i = bp_0j_i;
while(a_r_l-- > 0){ /* k */
temp.r += *ap_ik_r * *bp_kj_r - *ap_ik_i * *bp_kj_i;
temp.i += *ap_ik_r * *bp_kj_i + *ap_ik_i * *bp_kj_r;
ap_ik_r += a_st_r; ap_ik_i += a_st_r;
bp_kj_r += b_st_c; bp_kj_i += b_st_c;
} a_r_l = a->row_length;
*cp_ij_r = temp.r; *cp_ij_i = temp.i;
cp_ij_r += c_st_r; cp_ij_i += c_st_r;
bp_0j_r += b_st_r; bp_0j_i += b_st_r;
} c_r_l = c->row_length;
cp_i0_r += c_st_c; cp_i0_i += c_st_c;
ap_i0_r += a_st_c; ap_i0_i += a_st_c;
}
} break;
case 2 : { /* crc */
vsip_scalar_d *ap_ik_r, *ap_0k_r = ap0_r, *ap_ik_i, *ap_0k_i = ap0_i;
vsip_scalar_d *bp_kj_r, *bp_0j_r = bp0_r, *bp_kj_i, *bp_0j_i = bp0_i;
vsip_scalar_d *cp_ij_r, *cp_0j_r = cp0_r, *cp_ij_i, *cp_0j_i = cp0_i;
while(c_r_l-- > 0){ /* j */
bp_kj_r = bp_0j_r; bp_kj_i = bp_0j_i;
while(a_r_l-- > 0){ /* k */
b_scalar.r = *bp_kj_r; b_scalar.i = *bp_kj_i;
ap_ik_r = ap_0k_r; ap_ik_i = ap_0k_i;
cp_ij_r = cp_0j_r; cp_ij_i = cp_0j_i;
while(a_c_l-- > 0){ /* i */
*cp_ij_r += *ap_ik_r * b_scalar.r - *ap_ik_i * b_scalar.i;
*cp_ij_i += *ap_ik_i * b_scalar.r + *ap_ik_r * b_scalar.i;
cp_ij_r += c_st_c; cp_ij_i += c_st_c;
ap_ik_r += a_st_c; ap_ik_i += a_st_c;
} a_c_l = a->col_length;
ap_0k_r += a_st_r; ap_0k_i += a_st_r;
bp_kj_r += b_st_c; bp_kj_i += b_st_c;
} a_r_l = a->row_length;
cp_0j_r += c_st_r; cp_0j_i += c_st_r;
bp_0j_r += b_st_r; bp_0j_i += b_st_r;
ap_0k_r = ap0_r; ap_0k_i = ap0_i;
}
} break;
case 3 : { /* crr */
vsip_scalar_d *ap_ik_r, *ap_i0_r = ap0_r, *ap_ik_i, *ap_i0_i = ap0_i;
vsip_scalar_d *bp_kj_r, *bp_k0_r = bp0_r, *bp_kj_i, *bp_k0_i = bp0_i;
vsip_scalar_d *cp_ij_r, *cp_i0_r = cp0_r, *cp_ij_i, *cp_i0_i = cp0_i;
while(a_c_l-- > 0){ /* i */
ap_ik_r = ap_i0_r;
ap_ik_i = ap_i0_i;
while(a_r_l-- > 0){ /* k */
a_scalar.r = *ap_ik_r; a_scalar.i = *ap_ik_i;
ap_ik_r += a_st_r; ap_ik_i += a_st_r;
bp_kj_r = bp_k0_r; bp_kj_i = bp_k0_i;
cp_ij_r = cp_i0_r; cp_ij_i = cp_i0_i;
while(c_r_l-- > 0){ /* j */
*cp_ij_r = *bp_kj_r * a_scalar.r - *bp_kj_i * a_scalar.i + *cp_ij_r;
*cp_ij_i = *bp_kj_i * a_scalar.r + *bp_kj_r * a_scalar.i + *cp_ij_i;
cp_ij_r += c_st_r; cp_ij_i += c_st_r;
bp_kj_r += b_st_r; bp_kj_i += b_st_r;
} c_r_l = c->row_length;
bp_k0_r += b_st_c; bp_k0_i += b_st_c;
}a_r_l = a->row_length;
bp_k0_r = bp0_r; bp_k0_i = bp0_i;
cp_i0_r += c_st_c; cp_i0_i += c_st_c;
ap_i0_r += a_st_c; ap_i0_i += a_st_c;
}
} break;
case 4 : { /* rcc */
while(a_c_l-- > 0){
while(c_r_l-- > 0){
temp = vsip_cmplx_d(0.0f,0.0f);
while(a_r_l-- > 0){
temp.r += *ap_r * *bp_r - *ap_i * *bp_i;
temp.i += *ap_r * *bp_i + *ap_i * *bp_r;
ap_r += a_st_r; ap_i += a_st_r;
bp_r += b_st_c; bp_i += b_st_c;
}
*cp_r = temp.r; *cp_i = temp.i;
cp_r += c_st_r; cp_i += c_st_r;
ap_r = ap0_r; ap_i = ap0_i;
bp0_r += b_st_r; bp0_i += b_st_r;
bp_r = bp0_r; bp_i = bp0_i;
a_r_l = a->row_length;
}
ap0_r += a_st_c; ap0_i += a_st_c;
ap_r = ap0_r; ap_i = ap0_i;
bp0_r = bp1_r; bp0_i = bp1_i;
bp_r = bp0_r; bp_i = bp0_i;
cp0_r += c_st_c; cp0_i += c_st_c;
cp_r = cp0_r; cp_i = cp0_i;
c_r_l = c->row_length;
}
} break;
case 5 : { /* rcr */
while(a_c_l-- > 0){
while(c_r_l-- > 0){
temp = vsip_cmplx_d(0.0f,0.0f);
while(a_r_l-- > 0){
temp.r += *ap_r * *bp_r - *ap_i * *bp_i;
temp.i += *ap_r * *bp_i + *ap_i * *bp_r;
ap_r += a_st_r; ap_i += a_st_r;
bp_r += b_st_c; bp_i += b_st_c;
}
*cp_r = temp.r;
*cp_i = temp.i;
cp_r += c_st_r; cp_i += c_st_r;
ap_r = ap0_r; ap_i = ap0_i;
bp0_r += b_st_r; bp_r = bp0_r;
bp0_i += b_st_r; bp_i = bp0_i;
a_r_l = a->row_length;
}
ap0_r += a_st_c; ap_r = ap0_r;
ap0_i += a_st_c; ap_i = ap0_i;
bp0_r = bp1_r; bp_r = bp0_r;
bp0_i = bp1_i; bp_i = bp0_i;
cp0_r += c_st_c; cp_r = cp0_r;
cp0_i += c_st_c; cp_i = cp0_i;
c_r_l = c->row_length;
}
} break;
case 6 : { /* rrc */
vsip_scalar_d *ap_ik_r, *ap_i0_r, *ap_ik_i, *ap_i0_i;
vsip_scalar_d *bp_kj_r, *bp_0j_r = bp0_r, *bp_kj_i, *bp_0j_i = bp0_i;
vsip_scalar_d *cp_ij_r, *cp_0j_r = cp0_r, *cp_ij_i, *cp_0j_i = cp0_i;
/* jik */
while(c_r_l-- > 0){ /* j */
cp_ij_r = cp_0j_r; cp_ij_i = cp_0j_i;
ap_i0_r = ap0_r; ap_i0_i = ap0_i;
while(a_c_l-- > 0){ /* i */
temp = vsip_cmplx_d(0.0f, 0.0f);
ap_ik_r = ap_i0_r; ap_ik_i = ap_i0_i;
bp_kj_r = bp_0j_r; bp_kj_i = bp_0j_i;
while(a_r_l-- > 0){ /* k */
temp.r += *ap_ik_r * *bp_kj_r - *ap_ik_i * *bp_kj_i;;
temp.i += *ap_ik_r * *bp_kj_i + *ap_ik_i * *bp_kj_r;
ap_ik_r += a_st_r; ap_ik_i += a_st_r;
bp_kj_r += b_st_c; bp_kj_i += b_st_c;
} a_r_l = a->row_length;
*cp_ij_r = temp.r; *cp_ij_i = temp.i;
cp_ij_r +=c_st_c; cp_ij_i +=c_st_c;
ap_i0_r += a_st_c; ap_i0_i += a_st_c;
} a_c_l = a->col_length;
cp_0j_r += c_st_r; cp_0j_i += c_st_r;
bp_0j_r += b_st_r; bp_0j_i += b_st_r;
}
} break;
case 7 : { /* rrr */
vsip_scalar_d *ap_ik_r, *ap_i0_r = ap0_r, *ap_ik_i, *ap_i0_i = ap0_i;
vsip_scalar_d *bp_kj_r, *bp_k0_r = bp0_r, *bp_kj_i, *bp_k0_i = bp0_i;
vsip_scalar_d *cp_ij_r, *cp_i0_r = cp0_r, *cp_ij_i, *cp_i0_i = cp0_i;
while(a_c_l-- > 0){ /* i */
ap_ik_r = ap_i0_r; ap_ik_i = ap_i0_i;
while(a_r_l-- > 0){ /* k */
a_scalar.r = *ap_ik_r; a_scalar.i = *ap_ik_i;
ap_ik_r += a_st_r; ap_ik_i += a_st_r;
bp_kj_r = bp_k0_r; bp_kj_i = bp_k0_i;
cp_ij_r = cp_i0_r; cp_ij_i = cp_i0_i;
while(c_r_l-- > 0){ /* j */
*cp_ij_r = *bp_kj_r * a_scalar.r - *bp_kj_i * a_scalar.i + *cp_ij_r;
*cp_ij_i = *bp_kj_r * a_scalar.i + *bp_kj_i * a_scalar.r + *cp_ij_i;
cp_ij_r += c_st_r; cp_ij_i += c_st_r;
bp_kj_r += b_st_r; bp_kj_i += b_st_r;
} c_r_l = c->row_length;
bp_k0_r += b_st_c; bp_k0_i += b_st_c;
}a_r_l = a->row_length;
bp_k0_r = bp0_r; bp_k0_i = bp0_i;
cp_i0_r += c_st_c; cp_i0_i += c_st_c;
ap_i0_r += a_st_c; ap_i0_i += a_st_c;
}
}
}
}
void
test_ccorr_d(vsip_support_region support, vsip_bias bias,
vsip_length ref_size, vsip_length input_size)
{
vsip_length const n_loop = 3;
vsip_length const output_size = ref_corr_output_size(support, ref_size, input_size);
vsip_ccorr1d_d *corr = vsip_ccorr1d_create_d(ref_size, input_size, support, 0, VSIP_ALG_SPACE);
vsip_ccorr1d_attr attr;
vsip_ccorr1d_getattr_d(corr, &attr);
test_assert(attr.support == support);
test_assert(attr.ref_len == ref_size);
test_assert(attr.data_len == input_size);
test_assert(attr.lag_len == output_size);
vsip_randstate *rand = vsip_randcreate(0, 1, 1, VSIP_PRNG);
vsip_cvview_d *ref = vsip_cvcreate_d(ref_size, VSIP_MEM_NONE);
vsip_cvview_d *in = vsip_cvcreate_d(input_size, VSIP_MEM_NONE);
vsip_cvview_d *out = vsip_cvcreate_d(output_size, VSIP_MEM_NONE);
vsip_cvfill_d(vsip_cmplx_d(100,0), out);
vsip_cvview_d *chk = vsip_cvcreate_d(output_size, VSIP_MEM_NONE);
vsip_cvfill_d(vsip_cmplx_d(101,0), chk);
vsip_index loop;
for (loop=0; loop<n_loop; ++loop)
{
if (loop == 0)
{
vsip_cvfill_d(vsip_cmplx_d(1,0), ref);
vsip_cvramp_d(vsip_cmplx_d(0,0), vsip_cmplx_d(1,0), in);
}
else if (loop == 1)
{
vsip_cvrandu_d(rand, ref);
vsip_cvramp_d(vsip_cmplx_d(0,0), vsip_cmplx_d(1,0), in);
}
else
{
vsip_cvrandu_d(rand, ref);
vsip_cvrandu_d(rand, in);
}
vsip_ccorrelate1d_d(corr, bias, ref, in, out);
ref_ccorr_d(bias, support, ref, in, chk);
double error = cverror_db_d(out, chk);
#if VERBOSE
if (error > -100)
{
vsip_index i;
for (i=0; i<output_size; ++i)
{
vsip_cscalar_d out_value = vsip_cvget_d(out, i);
vsip_cscalar_d chk_value = vsip_cvget_d(chk, i);
printf("%d : out = (%f, %f), chk = (%f, %f)\n", i, out_value.r, out_value.i, chk_value.r, chk_value.i);
}
printf("error = %f\n", error);
}
#endif
test_assert(error < -100);
}
}
void
ref_ccorr_d(vsip_bias bias, vsip_support_region sup,
vsip_cvview_d const *ref,
vsip_cvview_d const *in,
vsip_cvview_d const *out)
{
vsip_length M = vsip_cvgetlength_d(ref);
vsip_length N = vsip_cvgetlength_d(in);
vsip_length P = vsip_cvgetlength_d(out);
vsip_length expected_P = ref_corr_output_size(sup, M, N);
vsip_stride shift = ref_expected_shift(sup, M);
assert(expected_P == P);
vsip_cvview_d *sub = vsip_cvcreate_d(M, VSIP_MEM_NONE);
// compute correlation
vsip_index i;
for (i=0; i<P; ++i)
{
vsip_cvfill_d(vsip_cmplx_d(0,0), sub);
vsip_stride pos = (vsip_stride)i + shift;
double scale;
if (pos < 0)
{
vsip_cvview_d *subsub = vsip_cvsubview_d(sub, -pos, M + pos);
vsip_cvview_d *insub = vsip_cvsubview_d(in, 0, M + pos);
vsip_cvcopy_d_d(insub, subsub);
vsip_cvdestroy_d(subsub);
vsip_cvdestroy_d(insub);
scale = M + pos;
}
else if (pos + M > N)
{
vsip_cvview_d *subsub = vsip_cvsubview_d(sub, 0, N - pos);
vsip_cvview_d *insub = vsip_cvsubview_d(in, pos, N - pos);
vsip_cvcopy_d_d(insub, subsub);
vsip_cvdestroy_d(subsub);
vsip_cvdestroy_d(insub);
scale = N - pos;
}
else
{
vsip_cvview_d *insub = vsip_cvsubview_d(in, pos, M);
vsip_cvcopy_d_d(insub, sub);
vsip_cvdestroy_d(insub);
scale = M;
}
#if VSIP_IMPL_CORR_CORRECT_SAME_SUPPORT_SCALING
#else
if (sup == VSIP_SUPPORT_SAME)
{
if (i < (M/2)) scale = i + (M+1)/2; // i + ceil(M/2)
else if (i < N - (M/2)) scale = M; // M
else scale = N - 1 + (M+1)/2 - i; // N-1+ceil(M/2)-i
}
#endif
vsip_cscalar_d val = vsip_cvjdot_d(ref, sub);
if (bias == VSIP_UNBIASED)
{
val.r /= scale;
val.i /= scale;
}
vsip_cvput_d(out, i, val);
}
}
vsip_cscalar_d (vsip_rcmul_d)(
vsip_scalar_d x, vsip_cscalar_d y) { /* x*y */
return vsip_cmplx_d(
x * vsip_real_d(y),
x * vsip_imag_d(y)); }
int main(){vsip_init((void*)0);
{
vsip_cmview_d *Adummy = vsip_cmcreate_d(5*NN,5*NN,VSIP_COL,0);
vsip_cmview_d *A = vsip_cmsubview_d(Adummy,3,2,NN,NN);
/* vsip_cmview_d *A= vsip_cmcreate_d(NN,NN,VSIP_COL,0); */
vsip_cvview_d *x0 = vsip_cvcreate_d(NN,0);
vsip_vview_d *x0_r = vsip_vrealview_d(x0);
vsip_vview_d *x0_i = vsip_vimagview_d(x0);
vsip_cmview_d *X = vsip_cmcreate_d(NN,3,VSIP_ROW,0);
vsip_cmview_d *XT = vsip_cmcreate_d(NN,3,VSIP_COL,0);
vsip_cmputrowstride_d(A,2*vsip_cmgetrowstride_d(A));
vsip_cmputcolstride_d(A,3*vsip_cmgetcolstride_d(A));
/* matrix data */
vsip_cmput_d(A,0,0,vsip_cmplx_d(0.5,0.1)); vsip_cmput_d(A,0,1,vsip_cmplx_d(7,0.1));
vsip_cmput_d(A,0,2,vsip_cmplx_d(10,0.1)); vsip_cmput_d(A,0,3,vsip_cmplx_d(12,0.1));
vsip_cmput_d(A,0,4,vsip_cmplx_d(-3,0.1)); vsip_cmput_d(A,0,5,vsip_cmplx_d(0,0.1)); vsip_cmput_d(A,0,6,vsip_cmplx_d(.05,0.1));
vsip_cmput_d(A,1,0,vsip_cmplx_d(2,0.1)); vsip_cmput_d(A,1,1,vsip_cmplx_d(13,0.1));
vsip_cmput_d(A,1,2,vsip_cmplx_d(18,0.1)); vsip_cmput_d(A,1,3,vsip_cmplx_d(6,0.1));
vsip_cmput_d(A,1,4,vsip_cmplx_d(0,0.1)); vsip_cmput_d(A,1,5,vsip_cmplx_d(130,0.1)); vsip_cmput_d(A,1,6,vsip_cmplx_d(8,0.1));
vsip_cmput_d(A,2,0,vsip_cmplx_d(3,0.1)); vsip_cmput_d(A,2,1,vsip_cmplx_d(-9,0.1));
vsip_cmput_d(A,2,2,vsip_cmplx_d(2,0.1)); vsip_cmput_d(A,2,3,vsip_cmplx_d(3,0.2));
vsip_cmput_d(A,2,4,vsip_cmplx_d(2,0.2)); vsip_cmput_d(A,2,5,vsip_cmplx_d(-9,0.2)); vsip_cmput_d(A,2,6,vsip_cmplx_d(6,0.2));
vsip_cmput_d(A,3,0,vsip_cmplx_d(4,0.2)); vsip_cmput_d(A,3,1,vsip_cmplx_d(2,0.2));
vsip_cmput_d(A,3,2,vsip_cmplx_d(2,0.2)); vsip_cmput_d(A,3,3,vsip_cmplx_d(4,0.2));
vsip_cmput_d(A,3,4,vsip_cmplx_d(1,0.2)); vsip_cmput_d(A,3,5,vsip_cmplx_d(2,0.2)); vsip_cmput_d(A,3,6,vsip_cmplx_d(3,0.2));
vsip_cmput_d(A,4,0,vsip_cmplx_d(.2,0.3)); vsip_cmput_d(A,4,1,vsip_cmplx_d(2,0.3));
vsip_cmput_d(A,4,2,vsip_cmplx_d(9,0.3)); vsip_cmput_d(A,4,3,vsip_cmplx_d(4,0.3));
vsip_cmput_d(A,4,4,vsip_cmplx_d(1,0.3)); vsip_cmput_d(A,4,5,vsip_cmplx_d(2,0.3));
vsip_cmput_d(A,4,6,vsip_cmplx_d(3,0.3));
vsip_cmput_d(A,5,0,vsip_cmplx_d(.1,0.4)); vsip_cmput_d(A,5,1,vsip_cmplx_d(2,0.4));
vsip_cmput_d(A,5,2,vsip_cmplx_d(.3,0.4)); vsip_cmput_d(A,5,3,vsip_cmplx_d(4,0.4));
vsip_cmput_d(A,5,4,vsip_cmplx_d(1,0.4)); vsip_cmput_d(A,5,5,vsip_cmplx_d(2,0.4)); vsip_cmput_d(A,5,6,vsip_cmplx_d(3,0.4));
vsip_cmput_d(A,6,0,vsip_cmplx_d(.01,0.4)); vsip_cmput_d(A,6,1,vsip_cmplx_d(.2,0.4));
vsip_cmput_d(A,6,2,vsip_cmplx_d(3,0.4)); vsip_cmput_d(A,6,3,vsip_cmplx_d(4,0.4));
vsip_cmput_d(A,6,4,vsip_cmplx_d(1,0.4)); vsip_cmput_d(A,6,5,vsip_cmplx_d(2,0.4)); vsip_cmput_d(A,6,6,vsip_cmplx_d(3,0.4));
{ /* were solving for NTRANS Ax = B */
/* use a known X, calculate B using Ax */
int k;
vsip_cvview_d *x;
vsip_cmview_d *AT = vsip_cmcreate_d(NN,NN,VSIP_ROW,VSIP_MEM_NONE);
vsip_length L = vsip_cmgetrowlength_d(X);
vsip_cmherm_d(A,AT);
printf("A = "); VU_cmprintm_d("7.4",A);
printf("AT = "); VU_cmprintm_d("7.4",AT);
vsip_vramp_d(1,1,x0_r);
vsip_vramp_d(1,-1,x0_i);
for(k=0; k<L; k++){
x = vsip_cmcolview_d(X,k);
vsip_cmvprod_d(A,x0,x);
vsip_rscvmul_d(2.0,x0,x0);
vsip_cvdestroy_d(x);
}
vsip_vramp_d(1,1,x0_r);
vsip_vramp_d(1,-1,x0_i);
for(k=0; k<L; k++){
x = vsip_cmcolview_d(XT,k);
vsip_cmvprod_d(AT,x0,x);
VU_cvprintm_d("7.4",x0);
vsip_rscvmul_d(2.0,x0,x0);
vsip_cvdestroy_d(x);
}
vsip_cmalldestroy_d(AT);
printf("B = "); VU_cmprintm_d("7.4",X);
printf("BT = "); VU_cmprintm_d("7.4",XT);
{
/* then solve to see if we get X back */
vsip_clu_d* luAop = vsip_clud_create_d(NN);
if(luAop == NULL) exit(1);
vsip_clud_d(luAop,A);
{ vsip_clu_attr_d attr;
vsip_clud_getattr_d(luAop,&attr);
printf("lud size %lu\n",attr.n);
}
vsip_clusol_d(luAop,VSIP_MAT_NTRANS,X);
vsip_clusol_d(luAop,VSIP_MAT_HERM,XT);
vsip_clud_destroy_d(luAop);
}
}
printf("A\\X = "); VU_cmprintm_d("9.6",X);
printf("A'\\X = "); VU_cmprintm_d("9.6",XT);
{
vsip_vdestroy_d(x0_r);vsip_vdestroy_d(x0_i);
vsip_cvalldestroy_d(x0);
vsip_cmalldestroy_d(X);
vsip_cmalldestroy_d(A);
}
}vsip_finalize((void*)0);return 1;
}
void (vsip_cvouter_d)(
vsip_cscalar_d alpha,
const vsip_cvview_d* a,
const vsip_cvview_d* b,
const vsip_cmview_d* R) { /* R_ij = a_i * conj(b_j) */
if((a == b) && (alpha.i == 0)){ /* covariance matrix with real multiplier */
vsip_length n = a->length;
vsip_stride cRst = R->block->cstride,
cast = a->block->cstride;
vsip_scalar_d *a_pr = (vsip_scalar_d*)(a->block->R->array + cast * a->offset),
*a_pi = (vsip_scalar_d*)(a->block->I->array + cast * a->offset);
vsip_length i,j;
vsip_stride stRu = cRst * R->row_stride, /* upper stride (above diagonal) */
stRl = cRst * R->col_stride, /* lower stride (below diagonal) */
sta = cast * a->stride; /* stride of a (and b) */
vsip_offset Ro = cRst * R->offset,
Rdiag = cRst * (R->col_stride + R->row_stride);
vsip_scalar_d *R_pr = (vsip_scalar_d*)(R->block->R->array + Ro),
*R_pi = (vsip_scalar_d*)(R->block->I->array + Ro);
for(i=0; i<n; i++){
/* prep for next loop */
vsip_scalar_d *R_plr = R_pr + stRl, /* pointer to lower real */
*R_pli = R_pi + stRl, /* pointer to lower imaginary */
*R_pur = R_pr + stRu, /* pointer to upper real */
*R_pui = R_pi + stRu; /* pointer to upper imaginary */
vsip_scalar_d atmp_r = *a_pr * alpha.r;
vsip_scalar_d atmp_i = *a_pi * alpha.r;
vsip_scalar_d *b_pr = a_pr + sta, *b_pi = a_pi + sta;
/* do diagonal member */
*(R_pr) = (atmp_r * *a_pr + atmp_i * *a_pi);
*(R_pi) = 0;
R_pr += Rdiag; R_pi += Rdiag;
for(j=i+1; j<n; j++){
/* do other members, along row for upper, column for lower */
vsip_scalar_d tmp = (atmp_r * *b_pr + atmp_i * *b_pi);
*R_plr = tmp; *R_pur = tmp;
tmp = (atmp_i * *b_pr - atmp_r * *b_pi);
*R_pli = -tmp; *R_pui = tmp;
R_plr += stRl; R_pli += stRl;
R_pur += stRu; R_pui += stRu;
b_pr += sta; b_pi += sta;
}
a_pr += sta, a_pi += sta;
}
} else {
vsip_length n = a->length,
m = b->length;
vsip_stride cRst = R->block->cstride, cast = a->block->cstride, cbst = b->block->cstride;
vsip_scalar_d *a_pr = (vsip_scalar_d*)(a->block->R->array + cast * a->offset),
*a_pi = (vsip_scalar_d*)(a->block->I->array + cast * a->offset);
vsip_length i,j;
vsip_stride stR = cRst * R->row_stride,
sta = cast * a->stride,
stb = cbst * b->stride;
vsip_offset Ro = cRst * R->offset,
Rco = cRst * R->col_stride,
bo = cbst * b->offset;
for(i=0; i<n; i++){
vsip_scalar_d *R_pr = (vsip_scalar_d*)(R->block->R->array + Ro + i * Rco),
*R_pi = (vsip_scalar_d*)(R->block->I->array + Ro + i * Rco),
*b_pr = (vsip_scalar_d*)(b->block->R->array + bo),
*b_pi = (vsip_scalar_d*)(b->block->I->array + bo);
vsip_cscalar_d temp = vsip_cmul_d(alpha,vsip_cmplx_d(*a_pr,*a_pi));
for(j=0; j<m; j++){
*R_pr = (temp.r * *b_pr + temp.i * *b_pi);
*R_pi = (temp.i * *b_pr - temp.r * *b_pi);
R_pr += stR; b_pr += stb;
R_pi += stR; b_pi += stb;
}
a_pr += sta;
a_pi += sta;
}
}
}
vsip_cscalar_d vsip_crandn_d(
vsip_randstate *state)
{
vsip_scalar_d real,imag;
if(state->type)
{ /* nonportable generator */
vsip_scalar_ue32 a = state->a,
c = state->c,
X = state->X;
vsip_scalar_d t2;
X = a * X + c;
real = (vsip_scalar_d)X/4294967296.0;
X = a * X + c;
real += (vsip_scalar_d)X/4294967296.0;
X = a * X + c;
real += (vsip_scalar_d)X/4294967296.0;
X = a * X + c;
t2 = (vsip_scalar_d)X/4294967296.0;
X = a * X + c;
t2 += (vsip_scalar_d)X/4294967296.0;
X = a * X + c;
t2 += (vsip_scalar_d)X/4294967296.0;
imag = real - t2;
real = 3 - t2 - real;
state->X = X;
} else { /* portable generator */
vsip_scalar_ue32 itemp;
vsip_scalar_d t2;
state->X = state->X * state->a + state->c;
state->X1 = state->X1 * state->a1 + state->c1;
itemp = state->X - state->X1;
if(state->X1 == state->X2){
state->X1++;
state->X2++;
}
real = (vsip_scalar_d)itemp/4294967296.0;
state->X = state->X * state->a + state->c;
state->X1 = state->X1 * state->a1 + state->c1;
itemp = state->X - state->X1;
if(state->X1 == state->X2){
state->X1++;
state->X2++;
}
real += (vsip_scalar_d)itemp/4294967296.0;
state->X = state->X * state->a + state->c;
state->X1 = state->X1 * state->a1 + state->c1;
itemp = state->X - state->X1;
if(state->X1 == state->X2){
state->X1++;
state->X2++;
}
real += (vsip_scalar_d)itemp/4294967296.0;
/* end t1 */
state->X = state->X * state->a + state->c;
state->X1 = state->X1 * state->a1 + state->c1;
itemp = state->X - state->X1;
if(state->X1 == state->X2){
state->X1++;
state->X2++;
}
t2 = (vsip_scalar_d)itemp/4294967296.0;
state->X = state->X * state->a + state->c;
state->X1 = state->X1 * state->a1 + state->c1;
itemp = state->X - state->X1;
if(state->X1 == state->X2){
state->X1++;
state->X2++;
}
t2 += (vsip_scalar_d)itemp/4294967296.0;
state->X = state->X * state->a + state->c;
state->X1 = state->X1 * state->a1 + state->c1;
itemp = state->X - state->X1;
if(state->X1 == state->X2){
state->X1++;
state->X2++;
}
t2 += (vsip_scalar_d)itemp/4294967296.0;
/* end t2 */
imag = real - t2;
real = 3 - t2 - real;
}
return vsip_cmplx_d(real,imag);
}
vsip_cscalar_d vsip_rect_d(
vsip_scalar_d r, vsip_scalar_d t) {
return vsip_cmplx_d(r * VSIP_COS_D(t), r * VSIP_SIN_D(t));
}
vsip_cfir_d *vsip_cfir_create_d(
const vsip_cvview_d *kernel,
vsip_symmetry symm,
vsip_length N,
vsip_length D,
vsip_obj_state state,
unsigned int ntimes,
vsip_alg_hint hint)
{
vsip_length klength = 0;
vsip_cfir_d *fir = (vsip_cfir_d*)malloc(sizeof(vsip_cfir_d));
if(fir == NULL){ return (vsip_cfir_d*) NULL;
} else {
switch(symm){
case 0: klength = kernel->length;
break;
case 1: klength = 2 * kernel->length - 1;
break;
case 2: klength = 2 * kernel->length;
break;
}
fir->h = vsip_cvcreate_d(klength,VSIP_MEM_NONE);
fir->s = vsip_cvcreate_d(klength-1,VSIP_MEM_NONE);
}
if((fir->h == NULL) | (fir->s == NULL)){
vsip_cfir_destroy_d(fir);
return (vsip_cfir_d*) NULL;
} else {
fir->N = N;
fir->M = klength;
fir->D = D;
fir->p = 0;
fir->ntimes = ntimes;
fir->symm = symm;
fir->hint = hint;
switch(symm){
case 0:{
fir->h->offset = fir->h->length - 1;
fir->h->stride = -1;
VI_cvcopy_d_d(kernel,fir->h);
}
break;
case 1: {
fir->h->length = kernel->length;
vsip_cvconj_d(kernel,fir->h);
fir->h->offset = klength - 1;
fir->h->stride = -1;
VI_cvcopy_d_d(kernel,fir->h);
fir->h->length = klength;
}
break;
case 2: {
fir->h->length = kernel->length;
vsip_cvconj_d(kernel,fir->h);
fir->h->offset = klength - 1;
fir->h->stride = -1;
VI_cvcopy_d_d(kernel,fir->h);
fir->h->length = klength;
}
break;
}
fir->state = state;
fir->h->offset = 0;
fir->h->stride = 1;
VI_cvfill_d(vsip_cmplx_d((vsip_scalar_d)0,(vsip_scalar_d)0),fir->s);
return fir;
}
}