ILU0Kernels.hpp
Go to the documentation of this file.
184template <int blocksize, class InputScalar, class OutputScalar, MatrixStorageMPScheme mixedPrecisionScheme>
Definition: ILU0Kernels.hpp:25 void solveLowerLevelSet(T *reorderedMat, int *rowIndices, int *colIndices, int *indexConversion, int startIdx, int rowsInLevelSet, const T *d, T *v, int threadBlockSize, cudaStream_t stream) Perform a lower solve on certain rows in a matrix that can safely be computed in parallel. void LUFactorizationSplit(InputScalar *srcReorderedLowerMat, int *lowerRowIndices, int *lowerColIndices, InputScalar *srcReorderedUpperMat, int *upperRowIndices, int *upperColIndices, InputScalar *srcDiagonal, OutputScalar *dstReorderedLowerMat, OutputScalar *dstReorderedUpperMat, OutputScalar *dstDiagonal, int *reorderedToNatural, int *naturalToReordered, int startIdx, int rowsInLevelSet, int threadBlockSize) Computes the ILU0 factorization in-place of a bcsr matrix stored in a split format (lower,... void LUFactorization(T *reorderedMat, int *rowIndices, int *columnIndices, int *naturalToReordered, int *reorderedToNatual, size_t rowsInLevelSet, int startIdx, int threadBlockSize) Computes the ILU Factorization of the input bcsr matrix, which is stored in a reordered way.... void solveUpperLevelSet(T *reorderedMat, int *rowIndices, int *colIndices, int *indexConversion, int startIdx, int rowsInLevelSet, T *v, int threadBlockSize, cudaStream_t stream) Perform a upper solve on certain rows in a matrix that can safely be computed in parallel. void solveLowerLevelSetSplit(MatrixScalar *reorderedLowerMat, int *rowIndices, int *colIndices, int *indexConversion, int startIdx, int rowsInLevelSet, const LinearSolverScalar *d, LinearSolverScalar *v, int threadBlockSize, cudaStream_t stream) Perform an lower solve on certain rows in a matrix that can safely be computed in parallel. void solveUpperLevelSetSplit(MatrixScalar *reorderedMat, int *rowIndices, int *colIndices, int *indexConversion, int startIdx, int rowsInLevelSet, const DiagonalScalar *dInv, LinearSolverScalar *v, int threadBlockSize, cudaStream_t stream) Perform an upper solve on certain rows in a matrix that can safely be computed in parallel. |