max/tug
1
0
Fork 0
mirror of https://git.gfz-potsdam.de/naaice/tug.git synced 2025-12-16 10:58:23 +01:00
Code Issues Packages Projects Releases Wiki Activity

perf: Minimize copy operations

Browse Source
This commit is contained in:
Max Luebke 2024-06-10 16:02:37 +02:00
parent c01d8e8607
commit 74b46f111b
1 changed files with 11 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
include/tug/Core/BTCS.hpp
View File

@ -369,7 +369,7 @@ static void BTCS_1D(Grid<T> &grid, Boundary<T> &bc, T timestep,
const auto inner_bc = bc.getInnerBoundaryRow(0); const auto inner_bc = bc.getInnerBoundaryRow(0);
Eigen::MatrixX<T> concentrations = grid.getConcentrations(); RowMajMat<T> &concentrations = grid.getConcentrations();
int rowIndex = 0; int rowIndex = 0;
A = createCoeffMatrix(alpha, bcLeft, bcRight, inner_bc, length, rowIndex, A = createCoeffMatrix(alpha, bcLeft, bcRight, inner_bc, length, rowIndex,
sx); // this is exactly same as in 2D sx); // this is exactly same as in 2D
@ -385,13 +385,13 @@ static void BTCS_1D(Grid<T> &grid, Boundary<T> &bc, T timestep,
b(length - 1) += 2 * sx * alpha(0, length - 1) * bcRight[0].getValue(); b(length - 1) += 2 * sx * alpha(0, length - 1) * bcRight[0].getValue();
} }
concentrations_t1 = solverFunc(A, b); concentrations = solverFunc(A, b);
for (int j = 0; j < length; j++) { // for (int j = 0; j < length; j++) {
concentrations(0, j) = concentrations_t1(j); // concentrations(0, j) = concentrations_t1(j);
} // }
grid.setConcentrations(concentrations); // grid.setConcentrations(concentrations);
} }
// BTCS solution for 2D grid // BTCS solution for 2D grid
@ -405,8 +405,7 @@ static void BTCS_2D(Grid<T> &grid, Boundary<T> &bc, T timestep,
T sx = timestep / (2 * grid.getDeltaCol() * grid.getDeltaCol()); T sx = timestep / (2 * grid.getDeltaCol() * grid.getDeltaCol());
T sy = timestep / (2 * grid.getDeltaRow() * grid.getDeltaRow()); T sy = timestep / (2 * grid.getDeltaRow() * grid.getDeltaRow());
RowMajMat<T> concentrations_t1 = RowMajMat<T>::Constant(rowMax, colMax, 0); RowMajMat<T> concentrations_t1(rowMax, colMax);
Eigen::VectorX<T> row_t1(colMax);
Eigen::SparseMatrix<T> A; Eigen::SparseMatrix<T> A;
Eigen::VectorX<T> b; Eigen::VectorX<T> b;
@ -421,7 +420,7 @@ static void BTCS_2D(Grid<T> &grid, Boundary<T> &bc, T timestep,
RowMajMat<T> &concentrations = grid.getConcentrations(); RowMajMat<T> &concentrations = grid.getConcentrations();
#pragma omp parallel for num_threads(numThreads) private(A, b, row_t1) #pragma omp parallel for num_threads(numThreads) private(A, b)
for (int i = 0; i < rowMax; i++) { for (int i = 0; i < rowMax; i++) {
auto inner_bc = bc.getInnerBoundaryRow(i); auto inner_bc = bc.getInnerBoundaryRow(i);
@ -429,17 +428,14 @@ static void BTCS_2D(Grid<T> &grid, Boundary<T> &bc, T timestep,
b = createSolutionVector(concentrations, alphaX, alphaY, bcLeft, bcRight, b = createSolutionVector(concentrations, alphaX, alphaY, bcLeft, bcRight,
bcTop, bcBottom, inner_bc, colMax, i, sx, sy); bcTop, bcBottom, inner_bc, colMax, i, sx, sy);
row_t1 = solverFunc(A, b); concentrations_t1.row(i) = solverFunc(A, b);
concentrations_t1.row(i) = row_t1;
} }
concentrations_t1.transposeInPlace(); concentrations_t1.transposeInPlace();
concentrations.transposeInPlace();
alphaX.transposeInPlace(); alphaX.transposeInPlace();
alphaY.transposeInPlace(); alphaY.transposeInPlace();
#pragma omp parallel for num_threads(numThreads) private(A, b, row_t1) #pragma omp parallel for num_threads(numThreads) private(A, b)
for (int i = 0; i < colMax; i++) { for (int i = 0; i < colMax; i++) {
auto inner_bc = bc.getInnerBoundaryCol(i); auto inner_bc = bc.getInnerBoundaryCol(i);
// swap alphas, boundary conditions and sx/sy for column-wise calculation // swap alphas, boundary conditions and sx/sy for column-wise calculation
@ -447,14 +443,8 @@ static void BTCS_2D(Grid<T> &grid, Boundary<T> &bc, T timestep,
b = createSolutionVector(concentrations_t1, alphaY, alphaX, bcTop, bcBottom, b = createSolutionVector(concentrations_t1, alphaY, alphaX, bcTop, bcBottom,
bcLeft, bcRight, inner_bc, rowMax, i, sy, sx); bcLeft, bcRight, inner_bc, rowMax, i, sy, sx);
row_t1 = solverFunc(A, b); concentrations.col(i) = solverFunc(A, b);
concentrations.row(i) = row_t1;
} }
concentrations.transposeInPlace();
grid.setConcentrations(concentrations);
} }
// entry point for EigenLU solver; differentiate between 1D and 2D grid // entry point for EigenLU solver; differentiate between 1D and 2D grid