AllReduce

Fill each subcone with the reduction (in some cases, sum) of the members of the subcone.

C++ API

Note

The cutoff parameter only effects parallel performance and is used to decide whether a subcone is sufficiently large to be treated separately.

void cone::AllReduce(Matrix<Real> &x, const Matrix<Int> &orders, const Matrix<Int> &firstInds, mpi::Op op = mpi::SUM)

void cone::AllReduce(ElementalMatrix<Real> &x, const ElementalMatrix<Int> &orders, const ElementalMatrix<Int> &firstInds, mpi::Op op = mpi::SUM, Int cutoff = 1000)

void cone::AllReduce(DistMultiVec<Real> &x, const DistMultiVec<Int> &orders, const DistMultiVec<Int> &firstInds, mpi::Op op = mpi::SUM, Int cutoff = 1000)

C API

TODO

Python API

TODO