bool
test_fundamentals(Index const dimension)
{
bool
passed = true;
Index const
number_components = integer_power(dimension, Tensor::ORDER);
std::vector<Scalar> const
X = generate_sequence<Scalar>(number_components, 1.0, 1.0);
// Test constructor with pointer
Tensor const
A(dimension, &X[0]);
// Test copy constructor
Tensor
B = A;
Tensor
C;
// Test copy assignment
C = B - A;
Scalar
error = norm_f(C);
bool const
copy_assigned = error <= machine_epsilon<Scalar>();
passed = passed && copy_assigned;
// Test fill with pointer
B.fill(&X[0]);
C = B - A;
error = norm_f(C);
bool const
filled_pointer = error <= machine_epsilon<Scalar>();
passed = passed && filled_pointer;
std::vector<Scalar> const
Y = generate_sequence<Scalar>(number_components, -1.0, -1.0);
C.fill(&Y[0]);
// Test increment
C += A;
error = norm_f(C);
bool const
incremented = error <= machine_epsilon<Scalar>();
passed = passed && incremented;
C.fill(&X[0]);
// Test decrement
C -= A;
error = norm_f(C);
bool const
decremented = error <= machine_epsilon<Scalar>();
passed = passed && decremented;
return passed;
}
bool
test_arithmetic(Index const dimension)
{
bool
passed = true;
Index const
number_components = integer_power(dimension, Tensor::ORDER);
std::vector<Scalar> const
X = generate_sequence<Scalar>(number_components, 1.0, 1.0);
Real const
sum_squares = number_components * (number_components + 1) *
(2 * number_components + 1) / 6;
// Test addition
Tensor const
A(dimension, &X[0]);
Tensor const
B = -1.0 * A;
Tensor const
C = -1.0 * B;
Tensor const
D = A + B;
Real
error = norm_f_square(D);
bool const
added = error <= machine_epsilon<Scalar>();
passed = passed && added;
// Test subtraction
Tensor const
E = A - C;
error = norm_f_square(E);
bool const
subtracted = error <= machine_epsilon<Scalar>();
passed = passed && subtracted;
// Test scaling
error = norm_f_square(C) - sum_squares;
bool const
scaled = error <= machine_epsilon<Scalar>();
passed = passed && scaled;
Tensor const
F = C / -1.0;
error = norm_f_square(F) - sum_squares;
bool const
divided = error <= machine_epsilon<Scalar>();
passed = passed && divided;
Tensor const
G = 1.0 / C;
error = norm_f_square(G) - sum_squares;
bool const
split = error <= machine_epsilon<Scalar>();
passed = passed && split;
return passed;
}
bool
test_filling(Index const dimension)
{
bool
passed = true;
Index const
number_components = integer_power(dimension, Tensor::ORDER);
// Test construct with zeros
Tensor
A(dimension, ZEROS);
Real
error = norm_f_square(A);
bool const
zeros_constructed = error <= machine_epsilon<Scalar>();
passed = passed && zeros_constructed;
// Test construct with ones
Tensor
B(dimension, ONES);
error = norm_f_square(B) - number_components;
bool const
ones_constructed = error <= machine_epsilon<Scalar>();
passed = passed && ones_constructed;
// Test construct with random entries
Tensor
C(dimension, RANDOM_UNIFORM);
error = norm_f(C);
bool const
random_constructed = error > 0.0 && error < number_components;
passed = passed && random_constructed;
// Test fill with random components
A.fill(RANDOM_UNIFORM);
error = norm_f(A);
bool const
random_filled = error > 0.0 && error < number_components;
passed = passed && random_filled;
// Test fill with zeros
B.fill(ZEROS);
error = norm_f_square(B);
bool const
zeros_filled = error <= machine_epsilon<Scalar>();
passed = passed && zeros_filled;
// Test fill with ones
C.fill(ZEROS);
error = norm_f_square(C) - number_components;
bool const
ones_filled = error <= machine_epsilon<Scalar>();
passed = passed && ones_filled;
return passed;
}
void operator_integer() {
int result;
char in[USHRT_MAX];
printf("Entre com o 1º valor: ");
scanf("%s", in);
printf("\n");
result = atoi(in);
int keep_going = 1;
while (keep_going) {
switch (menu_integer()) {
case 1:
// Add
printf("Entre com o próximo valor: ");
scanf("%s", in);
printf("\n");
integer_add(&result, atoi(in));
break;
case 2:
// Subtract
printf("Entre com o próximo valor: ");
scanf("%s", in);
printf("\n");
integer_subtract(&result, atoi(in));
break;
case 3:
// Multiply
printf("Entre com o próximo valor: ");
scanf("%s", in);
printf("\n");
integer_multiply(&result, atoi(in));
break;
case 4:
// Power
printf("Entre com o próximo valor: ");
scanf("%s", in);
printf("\n");
integer_power(&result, atoi(in));
break;
case 5:
// Divide
while (1) {
printf("Entre com o próximo valor: ");
scanf("%s", in);
printf("\n");
if (atoi(in) == 0)
printf("Valor inválido!\n\n");
else
break;
}
integer_divide(&result, atoi(in));
break;
case 6:
// Module
printf("Entre com o próximo valor: ");
scanf("%s", in);
printf("\n");
integer_module(&result, atoi(in));
break;
default:
keep_going = 0;
break;
}
printf("Resultado: %d\n\n", result);
}
}
bool
test_fundamentals(Index const dimension)
{
bool
passed = true;
Index const
number_components = integer_power(dimension, Tensor::ORDER);
std::vector<Scalar> const
X = generate_sequence<Scalar>(number_components, 1.0, 1.0);
// Test constructor with pointer
Tensor const
A(dimension, &X[0]);
// Test copy constructor
Tensor
B = A;
Tensor
C;
// Test copy assignment
C = B - A;
Scalar
error = norm_f(C);
bool const
copy_assigned = error <= machine_epsilon<Scalar>();
passed = passed && copy_assigned;
// Test fill with pointer
B.fill(&X[0]);
C = B - A;
error = norm_f(C);
bool const
filled_pointer = error <= machine_epsilon<Scalar>();
passed = passed && filled_pointer;
std::vector<Scalar> const
Y = generate_sequence<Scalar>(number_components, -1.0, -1.0);
C.fill(&Y[0]);
// Test increment
C += A;
error = norm_f(C);
bool const
incremented = error <= machine_epsilon<Scalar>();
passed = passed && incremented;
C.fill(&X[0]);
// Test decrement
C -= A;
error = norm_f(C);
bool const
decremented = error <= machine_epsilon<Scalar>();
passed = passed && decremented;
#ifdef HAVE_INTREPID_KOKKOSCORE
//test Tensor fill and create for Kokkos data types
Kokkos::View<Scalar *, Kokkos::DefaultExecutionSpace>
X1("X1_kokkos", dimension);
Kokkos::View<Scalar **, Kokkos::DefaultExecutionSpace>
X2("X2_kokkos", dimension, dimension);
Kokkos::View<Scalar ***, Kokkos::DefaultExecutionSpace>
X3("X3_kokkos", dimension, dimension, dimension);
Kokkos::View<Scalar ****, Kokkos::DefaultExecutionSpace>
X4("X4_kokkos", dimension, dimension, dimension, dimension);
Kokkos::deep_copy(X1, 3.1);
Kokkos::deep_copy(X2, 3.2);
Kokkos::deep_copy(X3, 3.3);
Kokkos::deep_copy(X4, 3.4);
Tensor
Z(dimension); //(X1_k,0);
Index
rank = 0;
Index
temp = number_components;
while (temp != 1) {
temp = temp / dimension;
rank = rank + 1;
assert(temp > 0);
}
switch (rank) {
default:
assert(false);
break;
case 1:
Z.fill(X1, 0);
break;
case 2:
Z.fill(X2, 0, 0);
break;
case 3:
Z.fill(X3, 0, 0, 0);
break;
case 4:
Z.fill(X4, 0, 0, 0, 0);
break;
}
// Test copy constructor.
Tensor const
U = Z;
// Test copy assignment.
Tensor
V;
V = U - Z;
error = norm_f(V);
bool const
tensor_create_from_1d_kokkos = error <= machine_epsilon<Scalar>();
passed = passed && tensor_create_from_1d_kokkos;
#endif
return passed;
}
void calculate_cell_reconstruction_matrices(int n_variables, double *weight_exponent, int *maximum_order, struct FACE *face, int n_cells, struct CELL *cell, struct ZONE *zone)
{
int c, u, i, j, k, l;
int order, n_powers, n_stencil;
//find the overall maximum order
int maximum_maximum_order = 0;
for(u = 0; u < n_variables; u ++) if(maximum_order[u] > maximum_maximum_order) maximum_maximum_order = maximum_order[u];
//cell structure allocation
for(c = 0; c < n_cells; c ++) exit_if_false(cell_matrix_new(n_variables, &cell[c]),"allocating cell matrices");
//numerics values
double **matrix, *weight;
int n_constraints, *constraint;
exit_if_false(allocate_double_matrix(&matrix,ORDER_TO_POWERS(maximum_maximum_order),MAX_STENCIL),"allocating matrix");
exit_if_false(allocate_integer_vector(&constraint,MAX_STENCIL),"allocating constraints");
exit_if_false(allocate_double_vector(&weight,MAX_STENCIL),"allocating weights");
//stencil element properties
int s_id, s_index;
struct ZONE *s_zone;
char s_location, *s_condition;
double s_area, *s_centroid, s_weight;
//integration
double x[2];
int differential[2], d;
//CV polygon
int n_polygon;
double ***polygon;
exit_if_false(allocate_double_pointer_matrix(&polygon,MAX(MAX_CELL_FACES,4),2),"allocating polygon memory");
for(c = 0; c < n_cells; c ++)
{
for(u = 0; u < n_variables; u ++)
{
//problem size
order = cell[c].order[u];
n_powers = ORDER_TO_POWERS(order);
n_stencil = cell[c].n_stencil[u];
n_constraints = 0;
for(i = 0; i < n_stencil; i ++)
{
//stencil element properties
s_id = cell[c].stencil[u][i];
s_index = ID_TO_INDEX(s_id);
s_zone = &zone[ID_TO_ZONE(s_id)];
s_location = s_zone->location;
s_condition = s_zone->condition;
if(s_location == 'f') {
s_centroid = face[s_index].centroid;
s_area = face[s_index].area;
} else if(s_location == 'c') {
s_centroid = cell[s_index].centroid;
s_area = cell[s_index].area;
} else exit_if_false(0,"recognising zone location");
s_weight = (s_centroid[0] - cell[c].centroid[0])*(s_centroid[0] - cell[c].centroid[0]);
s_weight += (s_centroid[1] - cell[c].centroid[1])*(s_centroid[1] - cell[c].centroid[1]);
s_weight = 1.0/pow(s_weight,0.5*weight_exponent[u]);
if(s_location == 'c' && s_index == c) s_weight = 1.0;
weight[i] = s_weight;
//unknown and dirichlet conditions have zero differentiation
differential[0] = differential[1] = 0;
//other conditions have differential determined from numbers of x and y-s in the condition string
if(s_condition[0] != 'u' && s_condition[0] != 'd')
{
j = 0;
while(s_condition[j] != '\0')
{
differential[0] += (s_condition[j] == 'x');
differential[1] += (s_condition[j] == 'y');
j ++;
}
}
//index for the determined differential
d = differential_index[differential[0]][differential[1]];
//unknowns
if(s_condition[0] == 'u')
{
/*//fit unknowns to centroid points
x[0] = s_centroid[0] - cell[c].centroid[0];
x[1] = s_centroid[1] - cell[c].centroid[1];
for(j = 0; j < n_powers; j ++)
{
matrix[j][i] = polynomial_coefficient[d][j]*
integer_power(x[0],polynomial_power_x[d][j])*
integer_power(x[1],polynomial_power_y[d][j])*
s_weight;
}*/
//fit unknowns to CV average
n_polygon = generate_control_volume_polygon(polygon, s_index, s_location, face, cell);
for(j = 0; j < n_powers; j ++) matrix[j][i] = 0.0;
for(j = 0; j <= order; j ++)
{
for(k = 0; k < n_polygon; k ++)
{
x[0] = 0.5*polygon[k][0][0]*(1.0 - gauss_x[order][j]) +
0.5*polygon[k][1][0]*(1.0 + gauss_x[order][j]) -
cell[c].centroid[0];
x[1] = 0.5*polygon[k][0][1]*(1.0 - gauss_x[order][j]) +
0.5*polygon[k][1][1]*(1.0 + gauss_x[order][j]) -
cell[c].centroid[1];
for(l = 0; l < n_powers; l ++)
{
//[face integral of polynomial integrated wrt x] * [x normal] / [CV area]
matrix[l][i] += polynomial_coefficient[d][l] *
(1.0 / (polynomial_power_x[d][l] + 1.0)) *
integer_power(x[0],polynomial_power_x[d][l]+1) *
integer_power(x[1],polynomial_power_y[d][l]) *
s_weight * gauss_w[order][j] * 0.5 *
(polygon[k][1][1] - polygon[k][0][1]) / s_area;
}
}
}
}
//knowns
else
{
//known faces fit to face average
if(s_location == 'f')
{
for(j = 0; j < n_powers; j ++) matrix[j][i] = 0.0;
for(j = 0; j < order; j ++)
{
x[0] = 0.5*face[s_index].node[0]->x[0]*(1.0 - gauss_x[order-1][j]) +
0.5*face[s_index].node[1]->x[0]*(1.0 + gauss_x[order-1][j]) -
cell[c].centroid[0];
x[1] = 0.5*face[s_index].node[0]->x[1]*(1.0 - gauss_x[order-1][j]) +
0.5*face[s_index].node[1]->x[1]*(1.0 + gauss_x[order-1][j]) -
cell[c].centroid[1];
for(k = 0; k < n_powers; k ++)
{
matrix[k][i] += polynomial_coefficient[d][k] *
integer_power(x[0],polynomial_power_x[d][k]) *
integer_power(x[1],polynomial_power_y[d][k]) *
s_weight*gauss_w[order-1][j]*0.5;
}
}
}
//cells need implementing
//if(s_location == 'c')
//{
//}
}
//constraints are the centre cell and any dirichlet boundaries
if((s_location == 'c' && s_index == c) || s_condition[0] == 'd') constraint[n_constraints++] = i;
}
//solve
if(n_constraints > 0)
exit_if_false(constrained_least_squares(n_stencil,n_powers,matrix,n_constraints,constraint) == LS_SUCCESS, "doing CLS calculation");
else
exit_if_false(least_squares(n_stencil,n_powers,matrix) == LS_SUCCESS,"doing LS calculation");
//multiply by the weights
for(i = 0; i < n_powers; i ++) for(j = 0; j < n_stencil; j ++) matrix[i][j] *= weight[j];
//store in the cell structure
for(i = 0; i < n_powers; i ++) for(j = 0; j < n_stencil; j ++) cell[c].matrix[u][i][j] = matrix[i][j];
}
}
//clean up
free_matrix((void**)matrix);
free_vector(constraint);
free_vector(weight);
free_matrix((void**)polygon);
}