void vsip_mcumsum_f(
const vsip_mview_f *a,
vsip_major major,
const vsip_mview_f *r) {
{
vsip_vview_f *va,vaa;
vsip_vview_f *vr,vrr;
vsip_index i;
vsip_length m = a->col_length;
vsip_length n = a->row_length;
va = &vaa; vr = &vrr;
if(major == VSIP_ROW){
for(i=0; i<m; i++){
VI_mrowview_f(a,i,va);
VI_mrowview_f(r,i,vr);
VI_vcumsum_f(va,vr);
}
} else { /* must be VSIP_COL */
for(i=0; i<n; i++){
VI_mcolview_f(a,i,va);
VI_mcolview_f(r,i,vr);
VI_vcumsum_f(va,vr);
}
}
}
return;
}
static void VI_qprodm_f(
vsip_mview_f *C0,
vsip_qr_f *qr)
{
vsip_mview_f CC = *C0;
vsip_mview_f *C = &CC;
vsip_mview_f HH = *(qr->A);
vsip_mview_f *H = &HH;
vsip_scalar_f *beta = qr->beta + qr->N - 1;
vsip_vview_f *h, hh;
vsip_stride k,
h_s = H->row_stride + H->col_stride;
vsip_vview_f vv = *(qr->v), ww = *(qr->w);
vsip_vview_f *v = &vv, *w = &ww;
h = VI_mcolview_f(H,0,&hh);
h->offset += (qr->N - 1) * h_s;
h->length -= (qr->N - 1);
w->length = C->row_length;
C->col_length -= (qr->N - 1);
C->offset += (qr->N -1) * C->col_stride;
for(k = qr->N -1; k >= 0; k--){
v->length = h->length;
VI_vcopy_f_f(h,v);
VI_VPUT_F(v,(vsip_index)0,(vsip_scalar_f)1);
v->length = C->col_length;
VI_svmprod_f(- *beta,v,C,w);
v->length = h->length;
VI_opu_f(C,v,w);
C->offset -= C->col_stride; C->col_length++;
h->length++; h->offset -= h_s;
beta--;
}
return;
}
static void VI_mprodq_f(
vsip_mview_f *C0,
vsip_qr_f *qr)
{
vsip_mview_f CC = *C0;
vsip_mview_f *C = &CC;
vsip_mview_f HH = *(qr->A);
vsip_mview_f *H = &HH;
vsip_scalar_f *beta = qr->beta;
vsip_vview_f *h, hh;
vsip_stride k,
h_s = H->row_stride + H->col_stride;
vsip_vview_f vv = *(qr->v), ww = *(qr->w);
vsip_vview_f *v = &vv, *w = &ww;
h = VI_mcolview_f(H,0,&hh);
v->length = h->length;
w->length = C->col_length;
for(k= 0; k < (vsip_stride)H->row_length; k++){
VI_vcopy_f_f(h,v);
VI_VPUT_F(v,(vsip_index) 0, (vsip_scalar_f)1.0);
v->length = C->row_length;
VI_smvprod_f(- *beta,C,v,w);
VI_opu_f(C,w,v);
h->offset += h_s; h->length --;
C->offset += C->row_stride; C->row_length--;
v->length = h->length; beta++;
}
return;
}
void vsip_mrandu_f(
vsip_randstate *state,
const vsip_mview_f *A)
{
vsip_index i;
vsip_vview_f v;
vsip_length M=A->col_length, N=A->row_length;
vsip_stride CS=A->col_stride, RS=A->row_stride;
if (CS < RS){ /* do by column */
for (i=0; i<N; i++){
VI_mcolview_f(A,i,&v);
vrandu_f(state,&v);
}
} else { /* do by row */
for (i=0; i<M; i++){
VI_mrowview_f(A,i,&v);
vrandu_f(state,&v);
}
}
}
