void SpinAdapted::MatrixMultiply (double d, const Matrix& a, Matrix& b)
{
// b += d * a;
#ifdef BLAS
assert ((a.Nrows () == b.Nrows ()) && (a.Ncols () == b.Ncols ()));
int n = a.Nrows () * a.Ncols ();
GAXPY (n, d, a.Store (), 1, b.Store (), 1);
#else
b += d * a;
#endif
}
/**
* Perform \f$self := scalar \times x + self\f$.
*
* @param self A vector object.
* @param x Another vector object.
* @param scalar A scalar.
*
* @return 0 on success.
*/
int parms_VecAXPY(FLOAT *self, FLOAT *x, FLOAT scalar, parms_Map map)
{
int i, lsize;
lsize = parms_MapGetLocalSize(map);
#ifdef HAS_BLAS
i = 1;
GAXPY(lsize, scalar, x, i, self, i);
#else
for (i = 0; i < lsize; i++)
{
self[i] += scalar * x[i];
}
#endif
return 0;
}
/**
* Perform \f$self = scalar \times self + x\f$.(block version)
*
* @param self A vector object.
* @param x Another vector object.
* @param scalar A scalar.
*
* @return 0 on success.
*/
int parms_VecAYPX_b(FLOAT *self, FLOAT *x, FLOAT scalar, parms_Map map)
{
int lsize, i;
lsize = parms_MapGetLocallSize(map);//should be point lsize instead of block lsize
#ifdef HAS_BLAS
FLOAT t;
i = 1;
t = 1.0;
GSCAL(lsize, scalar, self, i);
GAXPY(lsize, t, x, i, self, i);
#else
for (i = 0; i < lsize; i++) {
self[i] = scalar * self[i] + x[i];
}
#endif
return 0;
}
void SpinAdapted::MatrixTensorProduct (const Matrix& a_ref, char conjA, Real scaleA, const Matrix& b_ref, char conjB, Real scaleB, Matrix& c, int rowstride, int colstride, bool allocate)
{
#ifndef BLAS
Matrix A;
Matrix B;
#endif
Matrix& a = const_cast<Matrix&>(a_ref); // for BLAS calls
Matrix& b = const_cast<Matrix&>(b_ref);
int arows = a.Nrows();
int acols = a.Ncols();
// some specialisations
#ifdef FAST_MTP
// if ((brows == 1) && (bcols == 1))
{
double b00 = *b.Store();
if (conjA == 'n')
{
double* cptr = c.Store()+ rowstride*c.Ncols() + colstride;
for (int i=0; i< a.Nrows();i++)
DAXPY(a.Ncols(), scaleA * scaleB * b00, a.Store()+i*a.Ncols(), 1, cptr + i*c.Ncols(), 1);
return;
}
else
{
double* aptr = a.Store();
double* cptr = c.Store() + rowstride*c.Ncols() + colstride;
for (int col = 0; col < acols; ++col)
{
DAXPY(arows, scaleA * scaleB * b00, aptr, acols, cptr, 1);
++aptr;
cptr += c.Ncols();//arows;
}
return;
}
}
// else
// abort();
#else
try
{
if (conjA == 'n' && conjB == 'n')
{
if (allocate)
{
c.ReSize (a.Nrows () * b.Nrows (), a.Ncols () * b.Ncols ());
Clear (c);
}
//assert ((c.Nrows () == (a.Nrows () * b.Nrows ())) && (c.Ncols () == (a.Ncols () * b.Ncols ())));
#ifdef BLAS
int aRows = a.Nrows ();
int aCols = a.Ncols ();
int bRows = b.Nrows ();
int bCols = b.Ncols ();
for (int i = 0; i < aRows; ++i)
for (int j = 0; j < aCols; ++j)
{
Real scale = scaleA * scaleB * a (i+1,j+1);
for (int k = 0; k < bRows; ++k)
GAXPY (bCols, scale, &b (k+1,1), 1, &c (i * bRows + k+1 +rowstride,j * bCols+1+colstride), 1);
}
return;
#else
A = a;
B = b;
#endif
}
else if (conjA == 't' && conjB == 'n')
{
if (allocate)
{
c.ReSize (a.Ncols () * b.Nrows (), a.Nrows () * b.Ncols ());
Clear (c);
}
//assert ((c.Nrows () == (a.Ncols () * b.Nrows ())) && (c.Ncols () == (a.Nrows () * b.Ncols ())));
#ifdef BLAS
int aRows = a.Ncols ();
int aCols = a.Nrows ();
int bRows = b.Nrows ();
int bCols = b.Ncols ();
for (int i = 0; i < aRows; ++i)
for (int j = 0; j < aCols; ++j)
{
Real scale = scaleA * scaleB * a (j+1,i+1);
for (int k = 0; k < bRows; ++k)
GAXPY (bCols, scale, &b (k+1,1), 1, &c (i * bRows + k+1+rowstride,j * bCols+1+colstride), 1);
}
return;
#else
A = a.t ();
B = b;
#endif
}
else if (conjA == 'n' && conjB == 't')
{
if (allocate)
{
c.ReSize (a.Nrows () * b.Ncols (), a.Ncols () * b.Nrows ());
Clear (c);
}
//assert ((c.Nrows () == (a.Nrows () * b.Ncols ())) && (c.Ncols () == (a.Ncols () * b.Nrows ())));
#ifdef BLAS
int aRows = a.Nrows ();
int aCols = a.Ncols ();
int bRows = b.Ncols ();
int bCols = b.Nrows ();
for (int i = 0; i < aRows; ++i)
for (int j = 0; j < aCols; ++j)
{
Real scale = scaleA * scaleB * a (i+1,j+1);
for (int k = 0; k < bRows; ++k)
GAXPY (bCols, scale, &b (1,k+1), bRows, &c (i * bRows + k+1+rowstride,j * bCols+1+colstride), 1);
}
return;
#else
A = a;
B = b.t ();
#endif
}
else if (conjA == 't' && conjB == 't')
{
if (allocate)
{
c.ReSize (a.Ncols () * b.Ncols (), a.Nrows () * b.Nrows ());
Clear (c);
}
//assert ((c.Nrows () == (a.Ncols () * b.Ncols ())) && (c.Ncols () == (a.Nrows () * b.Nrows ())));
#ifdef BLAS
int aRows = a.Ncols ();
int aCols = a.Nrows ();
int bRows = b.Ncols ();
int bCols = b.Nrows ();
for (int i = 0; i < aRows; ++i)
for (int j = 0; j < aCols; ++j)
{
Real scale = scaleA * scaleB * a (j+1,i+1);
for (int k = 0; k < bRows; ++k)
GAXPY (bCols, scaleA * scaleB * a (j+1,i+1), &b (1,k+1), bRows, &c (i * bRows + k+1+rowstride,j * bCols+1+colstride), 1);
}
return;
#else
A = a.t ();
B = b.t ();
#endif
}
else
abort ();
#ifndef BLAS
for (int i = 1; i <= A.Nrows (); ++i)
for (int j = 1; j <= A.Ncols (); ++j)
c.SubMatrix ((i - 1) * B.Nrows () + 1, i * B.Nrows (), (j - 1) * B.Ncols () + 1, j * B.Ncols ()) += (scaleA * scaleB) * A (i,j) * B;
#endif
}
catch (Exception)
{
pout << Exception::what () << endl;
abort ();
}
#endif
}
