void vsip_mfloor_d_d(const vsip_mview_d* a, const vsip_mview_d* r)
{
vsip_length n_mj, /* major length */
n_mn; /* minor length */
vsip_stride ast_mj, ast_mn,
rst_mj, rst_mn;
vsip_scalar_d *rp = (r->block->array) + r->offset * r->block->rstride;
vsip_scalar_d *ap = (a->block->array) + a->offset * a->block->rstride;
vsip_scalar_d *rp0 = rp;
vsip_scalar_d *ap0 = ap;
if(r->row_stride < r->col_stride){
n_mj = r->row_length; n_mn = r->col_length;
rst_mj = r->row_stride; rst_mn = r->col_stride;
ast_mj = a->row_stride; ast_mn = a->col_stride;
ast_mj *= a->block->rstride; ast_mn *= a->block->rstride;
} else {
n_mn = r->row_length; n_mj = r->col_length;
rst_mn = r->row_stride; rst_mj = r->col_stride;
ast_mn = a->row_stride; ast_mj = a->col_stride;
ast_mn *= a->block->rstride; ast_mj *= a->block->rstride;
}
while(n_mn-- > 0){
vsip_length n = n_mj;
while(n-- >0){
*rp = VSIP_FLOOR_D(*ap);
ap += ast_mj; rp += rst_mj;
}
ap0 += ast_mn; rp0 += rst_mn;
ap = ap0; rp = rp0;
}
return;
}
void vsip_vfloor_d_d(const vsip_vview_d* a, const vsip_vview_d* r)
{
vsip_length n = r->length;
vsip_stride ast = a->stride * a->block->rstride,
rst = r->stride * r->block->rstride;
vsip_scalar_d *ap = (a->block->array) + a->offset * a->block->rstride;
vsip_scalar_d *rp = (r->block->array) + r->offset * r->block->rstride;
while(n-- > 0){
*rp = VSIP_FLOOR_D(*ap);
ap += ast; rp += rst;
}
}
void vsip_vfloor_d_i(const vsip_vview_d* a, const vsip_vview_i* r)
{
vsip_length n = r->length;
vsip_stride ast = a->stride * a->block->rstride,
rst = r->stride;
vsip_scalar_d *ap = (a->block->array) + a->offset * a->block->rstride;
vsip_scalar_i *rp = (r->block->array) + r->offset;
while(n-- > 0){
vsip_scalar_d t = VSIP_FLOOR_D(*ap);
if(t == 0.0)
*rp = 0;
else
*rp = t > 0 ? (vsip_scalar_i)(t+DBL_EPSILON):(vsip_scalar_i)(t-DBL_EPSILON);
ap += ast; rp += rst;
}
}
void vsip_mround_d_i(const vsip_mview_d* a, const vsip_mview_i* r)
{
vsip_length n_mj, /* major length */
n_mn; /* minor length */
vsip_stride ast_mj, ast_mn,
rst_mj, rst_mn;
vsip_scalar_i *rp = (r->block->array) + r->offset;
vsip_scalar_d *ap = (a->block->array) + a->offset * a->block->rstride;
vsip_scalar_i *rp0 = rp;
vsip_scalar_d *ap0 = ap;
vsip_scalar_d t;
if(r->row_stride < r->col_stride){
n_mj = r->row_length; n_mn = r->col_length;
rst_mj = r->row_stride; rst_mn = r->col_stride;
ast_mj = a->row_stride; ast_mn = a->col_stride;
ast_mj *= a->block->rstride; ast_mn *= a->block->rstride;
} else {
n_mn = r->row_length; n_mj = r->col_length;
rst_mn = r->row_stride; rst_mj = r->col_stride;
ast_mn = a->row_stride; ast_mj = a->col_stride;
ast_mn *= a->block->rstride; ast_mj *= a->block->rstride;
}
while(n_mn-- > 0){
vsip_length n = n_mj;
while(n-- >0){
vsip_scalar_f tf=VSIP_FLOOR_D(*ap);
vsip_scalar_f tc=VSIP_CEIL_D(*ap);
if(tf == tc) t = tf;
else{
if( .5 < *ap - tf) t = tc;
else if( .5 > *ap - tf) t = tf;
else t = (int)tf & 1 ? tc:tf;
}
if(t == 0.0)
*rp = 0;
else
*rp = t > 0 ? (vsip_scalar_i)(t+FLT_EPSILON):(vsip_scalar_i)(t-FLT_EPSILON);
ap += ast_mj; rp += rst_mj;
}
ap0 += ast_mn; rp0 += rst_mn;
ap = ap0; rp = rp0;
}
return;
}
void vsip_vround_d_d(const vsip_vview_d* a, const vsip_vview_d* r)
{
vsip_length n = r->length;
vsip_stride ast = a->stride * a->block->rstride,
rst = r->stride * r->block->rstride;
vsip_scalar_d *ap = (a->block->array) + a->offset * a->block->rstride;
vsip_scalar_d *rp = (r->block->array) + r->offset * r->block->rstride;
while(n-- > 0){
vsip_scalar_d tf=VSIP_FLOOR_D(*ap);
vsip_scalar_d tc=VSIP_CEIL_D(*ap);
if(tf == tc) *rp = tf;
else{
if( .5 < *ap - tf) *rp = tc;
else if( .5 > *ap - tf) *rp = tf;
else *rp = (int)tf & 1 ? tc:tf;
}
ap += ast; rp += rst;
}
}
void vsip_mfloor_d_i(const vsip_mview_d* a, const vsip_mview_i* r)
{
vsip_length n_mj, /* major length */
n_mn; /* minor length */
vsip_stride ast_mj, ast_mn,
rst_mj, rst_mn;
vsip_scalar_i *rp = (r->block->array) + r->offset;
vsip_scalar_d *ap = (a->block->array) + a->offset * a->block->rstride;
vsip_scalar_i *rp0 = rp;
vsip_scalar_d *ap0 = ap;
if(r->row_stride < r->col_stride){
n_mj = r->row_length; n_mn = r->col_length;
rst_mj = r->row_stride; rst_mn = r->col_stride;
ast_mj = a->row_stride; ast_mn = a->col_stride;
ast_mj *= a->block->rstride; ast_mn *= a->block->rstride;
} else {
n_mn = r->row_length; n_mj = r->col_length;
rst_mn = r->row_stride; rst_mj = r->col_stride;
ast_mn = a->row_stride; ast_mj = a->col_stride;
ast_mn *= a->block->rstride; ast_mj *= a->block->rstride;
}
while(n_mn-- > 0){
vsip_length n = n_mj;
while(n-- >0){
vsip_scalar_d t = VSIP_FLOOR_D(*ap);
if(t == 0.0)
*rp = 0;
else
*rp = t > 0 ? (vsip_scalar_i)(t+DBL_EPSILON):(vsip_scalar_i)(t-DBL_EPSILON);
ap += ast_mj; rp += rst_mj;
}
ap0 += ast_mn; rp0 += rst_mn;
ap = ap0; rp = rp0;
}
return;
}
void vsip_vround_d_i(const vsip_vview_d* a, const vsip_vview_i* r)
{
vsip_length n = r->length;
vsip_stride ast = a->stride * a->block->rstride,
rst = r->stride;
vsip_scalar_d *ap = (a->block->array) + a->offset * a->block->rstride;
vsip_scalar_i *rp = (r->block->array) + r->offset;
vsip_scalar_d t;
while(n-- > 0){
vsip_scalar_d tf=VSIP_FLOOR_D(*ap);
vsip_scalar_d tc=VSIP_CEIL_D(*ap);
if(tf == tc) t = tf;
else{
if( .5 < *ap - tf) t = tc;
else if( .5 > *ap - tf) t = tf;
else t = (int)tf & 1 ? tc:tf;
}
if(t == 0.0)
*rp = 0;
else
*rp = t > 0 ? (vsip_scalar_i)(t+DBL_EPSILON):(vsip_scalar_i)(t-DBL_EPSILON);
ap += ast; rp += rst;
}
}
