sse_blas_vaypx3_g5.h

Go to the documentation of this file.

// $Id: sse_blas_vaypx3_g5.h,v 1.4 2007-08-20 17:08:14 uid4709 Exp $

 
#ifndef QDP_SSE_BLAS_VAYPX3_G5
#define QDP_SSE_BLAS_VAYPX3_G5
 
#if defined(__GNUC__)
 
#include "qdp_config.h"
#include "qdp_sse_intrin.h"
 
namespace QDP {
 
#if BASE_PRECISION==32
 
 
 
// (Vector) out = Add + (Scalar) (*scalep) * P_{+} (Vector) InScale 
inline
void xpayz_g5ProjPlus(REAL32 *Out,REAL32 *scalep,REAL32 *Add, REAL32 *InScale,int n_4vec)
{
  // GNUC vector type
  
 
  // Load Vscalep
  v4sf vscalep = _mm_load_ss(scalep);
  vscalep = _mm_shuffle_ps(vscalep, vscalep,0);
 
  for(int i=0; i < n_4vec; i++) {
 
    // Spin Component 0: z0r, z0i, z1r, z1i
    _mm_store_ps(Out+ 0, _mm_add_ps(_mm_mul_ps(vscalep, _mm_load_ps(InScale+ 0)), _mm_load_ps(Add+ 0)));
     
    // Spin Component 0: z2r, z2i, SpinComponent 1: z0r, z0i
    _mm_store_ps(Out+ 4, _mm_add_ps(_mm_mul_ps(vscalep, _mm_load_ps(InScale+ 4)), _mm_load_ps(Add+ 4)));
 
    // Spin Component 1: z1r, z1i, z2r, z2i
    _mm_store_ps(Out+ 8, _mm_add_ps(_mm_mul_ps(vscalep, _mm_load_ps(InScale+ 8)), _mm_load_ps(Add+ 8)));
 
    // Spin Component 2: z0r, z0i, z1r, z1i
    _mm_store_ps(Out+12, _mm_load_ps(Add+ 12));
 
    // Spin Component 2: z2r, z2i, z0r, z0r
    _mm_store_ps(Out+16, _mm_load_ps(Add+ 16));
 
    // Spin Component 3: z1r, z1i, z2r, z2i
    _mm_store_ps(Out+20, _mm_load_ps(Add+ 20));
 
    // Update offsets
    Out += 24; InScale += 24; Add += 24;
  }
 
}
 
// (Vector) out = (Scalar) (*scalep) * (Vector) InScale + (Vector) P{+} Add
inline
void xpayz_g5ProjMinus(REAL32 *Out,REAL32 *scalep,REAL32 *Add, REAL32 *InScale,int n_4vec)
{
  // GNUC vector type
  
 
  // Load Vscalep
  v4sf vscalep = _mm_load_ss(scalep);
  vscalep = _mm_shuffle_ps(vscalep, vscalep, 0);
 
  for(int i=0; i < n_4vec; i++) {
    // Spin Component 0: z0r, z0i, z1r, z1i
    _mm_store_ps(Out+0,_mm_load_ps(Add+ 0));
 
    // Spin Component 0: z2r, z2i, Spin Component 1: z0r, z0r
    _mm_store_ps(Out+4,_mm_load_ps(Add+ 4));
 
    // Spin Component 1: z1r, z1i, z2r, z2i
    _mm_store_ps(Out+8, _mm_load_ps(Add+ 8));
 
 
    // Spin Component 2: z0r, z0i, z1r, z1i
    _mm_store_ps(Out+12, _mm_add_ps(_mm_mul_ps(vscalep, _mm_load_ps(InScale+ 12)), _mm_load_ps(Add+ 12)));
     
    // Spin Component 2: z2r, z2i, SpinComponent 3: z0r, z0i
    _mm_store_ps(Out+ 16, _mm_add_ps(_mm_mul_ps(vscalep, _mm_load_ps(InScale+ 16)), _mm_load_ps(Add+ 16)));
 
    // Spin Component 3: z1r, z1i, z2r, z2i
    _mm_store_ps(Out+ 20, _mm_add_ps(_mm_mul_ps(vscalep, _mm_load_ps(InScale+ 20)), _mm_load_ps(Add+ 20)));
    // Update offsets
    Out += 24; InScale += 24; Add += 24;
  }
 
}
 
 
 
// AXMY  versions
// (Vector) out = (Scalar) (*scalep) * (Vector) InScale - (Vector) P{+} Add
inline
void xmayz_g5ProjPlus(REAL32 *Out,REAL32 *scalep,REAL32 *Add, REAL32 *InScale,int n_4vec)
{
  // GNUC vector type
  
 
  // Load Vscalep
 
  v4sf vscalep = _mm_load_ss(scalep);
  vscalep = _mm_shuffle_ps(vscalep, vscalep, 0);
 
  for(int i=0; i < n_4vec; i++) {
    
    // Spin Component 0: z0r, z0i, z1r, z1i
    _mm_store_ps(Out+ 0, _mm_sub_ps(_mm_load_ps(Add+ 0),_mm_mul_ps(vscalep, _mm_load_ps(InScale+ 0))));
    
     
    // Spin Component 0: z2r, z2i, SpinComponent 1: z0r, z0i
    _mm_store_ps(Out+ 4, _mm_sub_ps(_mm_load_ps(Add+ 4),_mm_mul_ps(vscalep, _mm_load_ps(InScale+ 4))));
    
     // Spin Component 1: z1r, z1i, z2r, z2i
    _mm_store_ps(Out+ 8, _mm_sub_ps(_mm_load_ps(Add+ 8), _mm_mul_ps(vscalep, _mm_load_ps(InScale+ 8))));
  
    // Spin Component 2: z0r, z0i, z1r, z1i
    _mm_store_ps(Out+12, _mm_load_ps(Add + 12));
    
    
    
    // Spin Component 2: z2r, z2i, z0r, z0r
    _mm_store_ps(Out+16, _mm_load_ps(Add+ 16));
 
    // Spin Component 3: z1r, z1i, z2r, z2i
    _mm_store_ps(Out+20, _mm_load_ps(Add+20));
 
    // Update offsets
    Out += 24; InScale += 24; Add += 24;
  }
 
 
}
 
// (Vector) out = (Scalar) (*scalep) * (Vector) InScale - (Vector) P{-} Add
inline
void xmayz_g5ProjMinus(REAL32 *Out,REAL32 *scalep,REAL32 *Add, REAL32 *InScale,int n_4vec)
{
  // GNUC vector type
  
 
  // Load Vscalep
  v4sf vscalep = _mm_load_ss(scalep);
  vscalep = _mm_shuffle_ps(vscalep, vscalep,0);
 
  for(int i=0; i < n_4vec; i++) {
    // Spin Component 0: z0r, z0i, z1r, z1i
    _mm_store_ps(Out+0, _mm_load_ps(Add+0));
 
    // Spin Component 0: z2r, z2i, Spin Component 1: z0r, z0r
    _mm_store_ps(Out+4, _mm_load_ps(Add+4));
 
 
    // Spin Component 1: z1r, z1i, z2r, z2i
    _mm_store_ps(Out+8, _mm_load_ps(Add+8));
 
 
    // Spin Component 2: z0r, z0i, z1r, z1i
    _mm_store_ps(Out+12, _mm_sub_ps(_mm_load_ps(Add+ 12),_mm_mul_ps(vscalep, _mm_load_ps(InScale+ 12))));;
     
    // Spin Component 2: z2r, z2i, SpinComponent 3: z0r, z0i
    _mm_store_ps(Out+ 16, _mm_sub_ps(_mm_load_ps(Add+ 16),_mm_mul_ps(vscalep, _mm_load_ps(InScale+ 16))));
 
    // Spin Component 3: z1r, z1i, z2r, z2i
    _mm_store_ps(Out+ 20, _mm_sub_ps(_mm_load_ps(Add+ 20),_mm_mul_ps(vscalep, _mm_load_ps(InScale+ 20))));
    // Update offsets
    Out += 24; InScale += 24; Add += 24;
  }
 
}
 
 
#endif
 
} // namespace QDP;
 
 
#endif // GNUC
 
#endif // guard