public static void idx_m2squdotm2acc(idx <T> i1, idx <T> i2, idx <T> o)
{
loops.idx_checkorder3(i1, 2, i2, 2, o, 0);
#warning idx_checkdim2(i1, 0, i2.dim(0), i1, 1, i2.dim(1));
intg imax = i1.dim(0), jmax = i2.dim(1);
intg c1_m0 = i1.mod(0), c2_m0 = i2.mod(0);
intg c1_m1 = i1.mod(1), c2_m1 = i2.mod(1);
eptr <T> c1_0, c2_0, d1, c1, c2;
c1_0 = i1.idx_ptr();
c2_0 = i2.idx_ptr();
d1 = o.idx_ptr();
c1 = c1_0;
c2 = c2_0;
T f = d1.item;
for (int i = 0; i < imax; ++i)
{
c1.index = c1_0.index;
c2.index = c2_0.index;
for (int j = 0; j < jmax; ++j)
{
T ff = Dot(Dot(c1.item, c1.item), c2.item);
f = Add <T>(ff, f);
c1.index += c1_m1;
c2.index += c2_m1;
}
c1_0.index += c1_m0;
c2_0.index += c2_m0;
}
d1.item = f;
}
public static void idx_m1extm1(idx <T> i1, idx <T> i2, idx <T> o1)
{
eptr <T> c2, d1, c1, c2_0, d1_0;
intg c2_m0 = i2.mod(0), d1_m1 = o1.mod(1);
intg c1_m0 = i1.mod(0), d1_m0 = o1.mod(0);
intg j, jmax = o1.dim(1);
intg i, imax = o1.dim(0);
c1 = i1.idx_ptr();
c2_0 = i2.idx_ptr();
d1_0 = o1.idx_ptr();
for (i = 0; i < imax; i++)
{
d1 = d1_0;
c2 = c2_0;
for (j = 0; j < jmax; j++)
{
d1.item = Dot(c1.item, c2.item);
d1.index += d1_m1;
c2.index += c2_m0;
}
d1_0.index += d1_m0;
c1.index += c1_m0;
}
}
/* multiply M2 by M1, result in M1 */
/* matrix - vector product */
public static void idx_m2dotm1(idx <T> i1, idx <T> i2, idx <T> o1)
{
eptr <T> c1, c2, c1_0, ker;
intg c1_m1 = i1.mod(1), c2_m0 = i2.mod(0);
intg jmax = i2.dim(0);
intg c1_m0 = i1.mod(0), d1_m0 = o1.mod(0);
eptr <T> d1;
T f;
intg imax = o1.dim(0);
c1_0 = i1.idx_ptr();
ker = i2.idx_ptr();
d1 = o1.idx_ptr();
c1 = i1.idx_ptr();
c2 = i2.idx_ptr();
for (int ii = 0; ii < imax; ii++)
{
f = default(T);
c1.index = c1_0.index;
c2.index = ker.index;
if (c1_m1 == 1 && c2_m0 == 1)
{
for (int j = 0; j < jmax; j++)
{
f = Add <T>(f, Dot(c1.item, c2.item));
//f += (c1.item) * (c2.item);
c1++; c2++;
}
}
else
{
for (int j = 0; j < jmax; j++)
{
f = Add <T>(f, Dot(c1.item, c2.item));
// f += (c1.item) * (c2.item);
c1.index += c1_m1;
c2.index += c2_m0;
}
}
d1.item = f;
d1.index += d1_m0;
c1_0.index += c1_m0; //+ c1_0.index;
}
}
public static void idx_m4dotm2(idx <T> i1, idx <T> i2, idx <T> o1)
{
loops.idx_checkorder3(i1, 4, i2, 2, o1, 2); // check for compatible orders
if ((i1.dim(0) != o1.dim(0)) || (i1.dim(1) != o1.dim(1)) ||
(i1.dim(2) != i2.dim(0)) || (i1.dim(3) != i2.dim(1)))
{
loops.idx_compatibility_error3(i1, i2, o1, "incompatible dimensions");
}
eptr <T> c1, c1_2;
eptr <T> c2, c2_0;
eptr <T> c1_0, c1_1;
eptr <T> ker;
intg c1_m2 = (i1).mod(2), c2_m0 = (i2).mod(0);
intg c1_m3 = (i1).mod(3), c2_m1 = (i2).mod(1);
intg k, l, kmax = (i2).dim(0), lmax = (i2).dim(1);
eptr <T> d1_0, d1;
T f;
intg c1_m0 = (i1).mod(0), d1_m0 = (o1).mod(0);
intg c1_m1 = (i1).mod(1), d1_m1 = (o1).mod(1);
intg i, j, imax = (o1).dim(0), jmax = (o1).dim(1);
c1_0 = i1.idx_ptr();
ker = i2.idx_ptr();
d1_0 = o1.idx_ptr();
c1_1 = c1_0;
d1 = d1_0;
c1_2 = c1_1;
c2_0 = ker;
for (i = 0; i < imax; i++)
{
c1_1.index = c1_0.index;
d1.index = d1_0.index;
for (j = 0; j < jmax; j++)
{
f = default(T);
c1_2.index = c1_1.index;
c2_0.index = ker.index;
for (k = 0; k < kmax; k++)
{
c1 = c1_2;
c2 = c2_0;
for (l = 0; l < lmax; l++)
{
f = Add(Dot(c1.item, c2.item), f);
c1.index += c1_m3;
c2.index += c2_m1;
}
c1_2.index += c1_m2;
c2_0.index += c2_m0;
}
d1.item = f;
d1.index += d1_m1;
c1_1.index += c1_m1;
}
d1_0.index += d1_m0;
c1_0.index += c1_m0;
}
}
public static void idx_m2dotm2(idx <T> i1, idx <T> i2, idx <T> o)
{
eptr <T> c1, c2, c1_0, ker;
intg c1_m1 = i1.mod(1), c2_m0 = i2.mod(0);
intg j, jmax = i2.dim(0);
intg c1_m0 = i1.mod(0), d1_m0 = o.mod(0);
eptr <T> d1;
T f;
intg i, imax = o.dim(0);
intg k, kmax = o.dim(1);
c1 = new eptr <T>(i1.srgptr);
c2 = new eptr <T>(i2.srgptr);
// loop on o.dim(1)
for (k = 0; k < kmax; ++k)
{
c1_0 = i1.idx_ptr();
d1 = o.idx_ptr(); d1.index += k;
ker = i2.idx_ptr(); ker.index += k;
// loop on o.dim(0)
for (i = 0; i < imax; i++)
{
f = default(T);
c1.index = c1_0.index;
c2.index = ker.index;
// loop on
if (c1_m1 == 1 && c2_m0 == 1)
{
for (j = 0; j < jmax; j++)
{
f = Add <T>(f, Dot(c1.item, c2.item));
c1++; c2++;
}
}
else
{
for (j = 0; j < jmax; j++)
{
f = Add <T>(f, Dot(c1.item, c2.item));// (*c1) * (*c2);
c1.index += c1_m1;
c2.index += c2_m0;
}
}
d1.item = f;
d1.index += d1_m0;
c1_0.index += c1_m0;
}
}
}
public static void idx_m2extm2acc(idx <T> i1, idx <T> i2, idx <T> o1)
{
eptr <T> c2_0, c2_1;
eptr <T> d1_2, d1_3;
eptr <T> d1_0, d1_1;
eptr <T> c1_0, c1_1;
eptr <T> ker;
intg c2_m0 = (i2).mod(0), c2_m1 = (i2).mod(1);
intg d1_m2 = (o1).mod(2), d1_m3 = (o1).mod(3);
intg c1_m0 = (i1).mod(0), c1_m1 = (i1).mod(1);
intg d1_m0 = (o1).mod(0), d1_m1 = (o1).mod(1);
intg k, l, lmax = (o1).dim(3), kmax = (o1).dim(2);
intg i, j, imax = (o1).dim(0), jmax = (o1).dim(1);
c1_0 = i1.idx_ptr();
ker = i2.idx_ptr();
d1_0 = o1.idx_ptr();
d1_2 = d1_0;
c2_0 = ker;
for (i = 0; i < imax; i++)
{
d1_1 = d1_0;
c1_1 = c1_0;
for (j = 0; j < jmax; j++)
{
d1_2.index = d1_1.index;
c2_0.index = ker.index;
for (k = 0; k < kmax; k++)
{
d1_3 = d1_2;
c2_1 = c2_0;
for (l = 0; l < lmax; l++)
{
d1_3.item = Add(Dot(c1_1.item, c2_1.item), d1_3.item);
d1_3.index += d1_m3;
c2_1.index += c2_m1;
}
d1_2.index += d1_m2;
c2_0.index += c2_m0;
}
d1_1.index += d1_m1;
c1_1.index += c1_m1;
}
d1_0.index += d1_m0;
c1_0.index += c1_m0;
}
}
