Template Function Pennylane::LightningGPU::Util::SparseMV_cuSparseMPI

• Defined in File MPILinearAlg.hpp

Function Documentation

template<class index_type, class Precision, class CFP_t, class DevTypeID = int>
inline void Pennylane::LightningGPU::Util::SparseMV_cuSparseMPI(MPIManager &mpi_manager, const std::size_t &length_local, const index_type *csrOffsets_ptr, const int64_t csrOffsets_size, const index_type *columns_ptr, const std::complex<Precision> *values_ptr, CFP_t *X, CFP_t *Y, DevTypeID device_id, cudaStream_t stream_id, cusparseHandle_t handle)

Sparse matrix vector multiply offloaded to cuSparse (Y = alpha*SparseMat*X + beta)

Template Parameters

• index_type – Integer type for offsets, indices and number of elements (std::size_t for the moment).

• Precision – Floating data-type.

• DevTypeID – Integer type of device id.

Parameters

• mpi_manager – MPI operation wrapper.

• length_local – Length of X.

• csrOffsets_ptr – Pointer to offsets in CSR format.

• csrOffsets_size – Number of elements of offsets.

• columns_ptr – Pointer to column indices in CSR format.

• values_ptr – Pointer to value of each non-zero elements in CSR format.

• numNNZ – Number of non-zero elements.

• X – Pointer to vector.

• Y – Pointer to vector.

• device_id – Device id.

• cudaStream_t – Stream id.

• handle – cuSparse handle.

api/function_MPILinearAlg_8hpp_1a89fc0abf4b4b1998ed1ab64b284f267c

 

Download Python script

 

Download Notebook

 

View on GitHub