static void VI_csolve_ut_d(
vsip_cmview_d *R,
vsip_cmview_d *B)
{
vsip_length N = R->row_length;
vsip_cmview_d XX = *B;
vsip_cmview_d *X = &XX;
vsip_cvview_d xx, rr;
vsip_cvview_d *x = VI_cmrowview_d(X,(vsip_index) (N-1),&xx);
vsip_cvview_d *r_d = VI_cmrowview_d(R,(vsip_index) 0,&rr);
vsip_cvview_d rr_m = rr;
vsip_cvview_d *r_m = &rr_m;
vsip_stride d_s = R->row_stride + R->col_stride;
r_d->length = 1; r_d->offset += ((N-1) * d_s);
r_m->length = 0; r_m->offset = r_d->offset + R->row_stride;
X->col_length = 1; X->offset = x->offset;
vsip_csvmul_d(vsip_crecip_d(vsip_cvget_d(r_d,0)),x,x);
N--;
while(N-- >0){
r_d->offset -= d_s;
r_m->length++; r_m->offset -= d_s;
x->offset -= X->col_stride;
VI_cvsubvmprodIP_d(r_m,X,x);
vsip_csvmul_d(vsip_crecip_d(vsip_cvget_d(r_d,0)),x,x);
X->col_length++; X->offset = x->offset;
}
return;
}
static void VI_csolve_lt_d(
vsip_cmview_d *R,
vsip_cmview_d *B)
{
vsip_length N = R->row_length;
vsip_cmview_d XX = *B;
vsip_cmview_d *X = &XX;
vsip_cvview_d xx, rr;
vsip_cvview_d *x = VI_cmrowview_d(X,(vsip_index)0,&xx);
vsip_cvview_d *r_d = VI_cmrowview_d(R,(vsip_index) 0,&rr);
vsip_cvview_d rr_m = rr;
vsip_cvview_d *r_m = &rr_m;
vsip_stride d_s = R->row_stride + R->col_stride;
r_d->length = 1;
r_m->length = 0;
X->col_length = 1;
/* vsip_csvmul_d(vsip_crecip_d(vsip_conj_d(vsip_cvget_d(r_d,0))),x,x); */
vsip_rscvmul_d((vsip_scalar_d)1.0/(vsip_real_d(vsip_cvget_d(r_d,0))),x,x);
N--;
while(N-- >0){
r_d->offset += d_s;
r_m->length++; r_m->offset += R->col_stride;
x->offset += X->col_stride;
VI_cvjsubvmprodIP_d(r_m,X,x);
/* vsip_csvmul_d(vsip_crecip_d(vsip_conj_d(vsip_cvget_d(r_d,0))),x,x); */
vsip_rscvmul_d((vsip_scalar_d)1.0/(vsip_real_d(vsip_cvget_d(r_d,0))),x,x);
X->col_length++;
}
return;
}
void vsip_mpolar_d(const vsip_cmview_d* a, const vsip_mview_d* r, const vsip_mview_d* t){
vsip_index i;
vsip_vview_d rv; vsip_vview_d tv; vsip_cvview_d av;
if(a->row_stride < a->col_stride){
for(i=0; i < a->col_length; i++){
VI_mrowview_d(r,i,&rv); VI_mrowview_d(t,i,&tv);VI_cmrowview_d(a,i,&av);
vpolar_d(&av,&rv,&tv);
}
} else {
for(i=0; i < a->col_length; i++){
VI_mcolview_d(r,i,&rv); VI_mcolview_d(t,i,&tv);VI_cmcolview_d(a,i,&av);
vpolar_d(&av,&rv,&tv);
}
}
}
static void VI_csolve_lt_special_d(
const vsip_cmview_d *R,
const vsip_cmview_d *B)
{
vsip_length N = R->row_length;
vsip_cmview_d XX = *B;
vsip_cmview_d *X = &XX;
vsip_cvview_d xx, rr;
vsip_cvview_d *x = VI_cmrowview_d(X,(vsip_index)0,&xx);
vsip_cvview_d *r_d = VI_cmrowview_d(R,(vsip_index) 0,&rr);
vsip_cvview_d rr_m = rr;
vsip_cvview_d *r_m = &rr_m;
r_d->length = 1;
r_m->length = 0;
X->col_length = 1;
N--;
while(N-- >0){
r_m->length++; r_m->offset += R->col_stride;
x->offset += X->col_stride;
VI_cvsubvmprodIP_d(r_m,X,x);
X->col_length++;
}
return;
}
void (vsip_cgemp_d)(vsip_cscalar_d alpha,
const vsip_cmview_d *AA,
vsip_mat_op OpA,
const vsip_cmview_d *BB,
vsip_mat_op OpB,
vsip_cscalar_d beta,
const vsip_cmview_d *C){
vsip_cmview_d At = *AA,
Bt = *BB;
vsip_cmview_d *A = &At,
*B = &Bt;
vsip_cscalar_d (*dot)(const vsip_cvview_d*,const vsip_cvview_d*) = vsip_cvdot_d;
if((OpA == 1) || (OpA == 2)){
A->row_length = AA->col_length;
A->col_length = AA->row_length;
A->row_stride = AA->col_stride;
A->col_stride = AA->row_stride;
}
if((OpB == 1) || (OpB == 2)){
B->row_length = BB->col_length;
B->col_length = BB->row_length;
B->row_stride = BB->col_stride;
B->col_stride = BB->row_stride;
}
{
vsip_length M = A->col_length,
N = B->row_length;
vsip_length i,j;
vsip_cvview_d aa,bb,cc,*a,*b,*c,*af,*bs;
vsip_offset Boffset = B->offset;
vsip_stride Brow_stride = B->row_stride,
Acol_stride = A->col_stride;
a = VI_cmrowview_d(A,0,&aa);
b = VI_cmcolview_d(B,0,&bb);
af = a; bs = b;
if( ((OpA == 2) || (OpA == 3)) && ((OpB !=2) || (OpB !=3))){
dot = vsip_cvjdot_d;
af = b;
bs = a;
}
if( ((OpA != 2) || (OpA != 3)) && ((OpB ==2) || (OpB ==3))){
dot = vsip_cvjdot_d;
}
if( ((OpA == 2) || (OpA == 3)) && ((OpB ==2) || (OpB ==3))){
dot = VI_cvjjdot_d;
}
c = VI_cmrowview_d(C,0,&cc);
{
vsip_cscalar_d temp;
vsip_scalar_d temp_r;
vsip_stride str = c->stride * c->block->cstride;
vsip_stride str1 = C->col_stride * C->block->cstride;
vsip_scalar_d *c_pr_r =(vsip_scalar_d*) (c->block->R->array + c->offset * c->block->cstride),
*c_pr1_r = c_pr_r;
vsip_scalar_d *c_pr_i =(vsip_scalar_d*) (c->block->I->array + c->offset * c->block->cstride),
*c_pr1_i = c_pr_i;
for(i = 0; i < M; i++){
c_pr_r = c_pr1_r;
c_pr_i = c_pr1_i;
b->offset = Boffset;
for(j =0; j < N; j++){
temp_r = beta.r * *c_pr_r - beta.i * *c_pr_i;
*c_pr_i = beta.i * *c_pr_r + beta.r * *c_pr_i;
*c_pr_r = temp_r;
temp = dot(af,bs);
*c_pr_r += (alpha.r * temp.r - alpha.i * temp.i);
*c_pr_i += (alpha.i * temp.r + alpha.r * temp.i);
c_pr_r += str;
c_pr_i += str;
b->offset += Brow_stride;
}
a->offset += Acol_stride;
c_pr1_r += str1;
c_pr1_i += str1;
}
}
}
}
