qdp_scalarvecsite_sse.cc

Go to the documentation of this file.

 
 
#include "qdp.h"
 
// These SSE asm instructions are only supported under GCC/G++
#if defined(__GNUC__) && __GNUC_MINOR__ >= 2  &&  QDP_USE_SSE == 1
 
namespace QDP {
 
// Specialization to optimize the case   
//    LatticeColorMatrix[Subset] = LatticeColorMatrix * LatticeColorMatrix
template<>
void evaluate(OLattice<PScalar<PColorMatrix<RComplexFloat, 3> > >& d, 
              const OpAssign& op, 
              const QDPExpr<BinaryNode<OpMultiply, 
              Reference<QDPType<PScalar<PColorMatrix<RComplexFloat, 3> >, 
              OLattice<PScalar<PColorMatrix<RComplexFloat, 3> > > > >, 
              Reference<QDPType<PScalar<PColorMatrix<RComplexFloat, 3> >, 
              OLattice<PScalar<PColorMatrix<RComplexFloat, 3> > > > > >,
              OLattice<PScalar<PColorMatrix<RComplexFloat, 3> > > >& rhs,
              const Subset& s)
{
//  cout << "call single site QDP_M_eq_M_times_M" << std::endl;
 
  typedef OLattice<PScalar<PColorMatrix<RComplexFloat, 3> > >  C;
 
  const C& l = static_cast<const C&>(rhs.expression().left());
  const C& r = static_cast<const C&>(rhs.expression().right());
 
  if( s.hasOrderedRep()) { 
    const int istart = s.start() >> INNER_LOG;
    const int iend   = s.end()   >> INNER_LOG;
 
    for(int i=istart; i <= iend; ++i) 
      {
        float *dd = (float*)&(d.elem(i).elem());
        float *ll = (float*)&(l.elem(i).elem());
        float *rr = (float*)&(r.elem(i).elem());
        
        _inline_ssevec_mult_su3_nn(dd,ll,rr,0);
        _inline_ssevec_mult_su3_nn(dd,ll,rr,1);
        _inline_ssevec_mult_su3_nn(dd,ll,rr,2);
      }
  }
  else { 
    // Do unoptimised - hope this still workl
    const int *tab = s.siteTable().slice();
    for(int j=0; j < s.numSiteTable(); j++) { 
      i = tab[j]; 
      d.elem(i).elem() = l.elem(i).elem() * r.elem(i).elem();
    }
  }
 
}
 
} // namespace QDP;
 
#endif  // defined(__GNUC__)