MLCommon::LinAlg::opg Namespace Reference

Functions
void	eigDC (const raft::handle_t &h, float *eigenValues, float *eigenVectors, std::vector< Matrix::Data< float > * > &inParts, Matrix::PartDescriptor &desc, int myRank, cudaStream_t stream)
	Multi-GPU version of Eigen decomposition. This function works for symmetric matrices only. Whole input matrix is gathered at rank 0 and Eigen decomposition is carried out sequentially. More...
void	eigDC (const raft::handle_t &h, double *eigenValues, double *eigenVectors, std::vector< Matrix::Data< double > * > &inParts, Matrix::PartDescriptor &desc, int myRank, cudaStream_t stream)
void	eigJacobi (const raft::handle_t &h, float *eigenValues, float *eigenVectors, std::vector< Matrix::Data< float > * > &inParts, Matrix::PartDescriptor &desc, int myRank, cudaStream_t stream)
void	eigJacobi (const raft::handle_t &h, double *eigenValues, double *eigenVectors, std::vector< Matrix::Data< double > * > &inParts, Matrix::PartDescriptor &desc, int myRank, cudaStream_t stream)
void	gemm (const raft::handle_t &h, std::vector< Matrix::Data< float > * > &outZParts, Matrix::PartDescriptor &outZDesc, std::vector< Matrix::Data< float > * > &inXParts, Matrix::PartDescriptor &inXDesc, std::vector< Matrix::Data< float > * > &inYParts, Matrix::PartDescriptor &inYDesc, int myRank, cudaStream_t stream)
	A multi gpu generalized matrix multiplication function. This function performs Z = X * Y The X and Y matrix are distributed in blocks on different ranks. First Y matrix is duplicated at each rank. It is multiplied with blocks of X local to the rank. More...
void	gemm (const raft::handle_t &h, std::vector< Matrix::Data< double > * > &outZParts, Matrix::PartDescriptor &outZDesc, std::vector< Matrix::Data< double > * > &inXParts, Matrix::PartDescriptor &inXDesc, std::vector< Matrix::Data< double > * > &inYParts, Matrix::PartDescriptor &inYDesc, int myRank, cudaStream_t stream)
void	lstsqEig (const raft::handle_t &handle, const std::vector< Matrix::Data< float > * > &A, const Matrix::PartDescriptor &ADesc, const std::vector< Matrix::Data< float > * > &b, float *w, cudaStream_t *streams, int n_streams)
	performs MNMG Least squares calculation. More...
void	lstsqEig (const raft::handle_t &handle, const std::vector< Matrix::Data< double > * > &A, const Matrix::PartDescriptor &ADesc, const std::vector< Matrix::Data< double > * > &b, double *w, cudaStream_t *streams, int n_streams)
void	meanSquaredError (double *out, const Matrix::Data< double > &in1, const Matrix::PartDescriptor &in1Desc, const Matrix::Data< double > &in2, const Matrix::PartDescriptor &in2Desc, const raft::comms::comms_t &comm, cudaStream_t stream, int root=0, bool broadcastResult=true)
	multi-gpu mean squared error More...
void	meanSquaredError (float *out, const Matrix::Data< float > &in1, const Matrix::PartDescriptor &in1Desc, const Matrix::Data< float > &in2, const Matrix::PartDescriptor &in2Desc, const raft::comms::comms_t &comm, cudaStream_t stream, int root=0, bool broadcastResult=true)
void	mm_aTa (const raft::handle_t &handle, Matrix::Data< double > &out, const std::vector< Matrix::Data< double > * > &in1, const Matrix::PartDescriptor &in1Desc, cudaStream_t *streams, int n_streams)
	performs MNMG A^T x A calculation. More...
void	mm_aTa (const raft::handle_t &handle, Matrix::Data< float > &out, const std::vector< Matrix::Data< float > * > &in1, const Matrix::PartDescriptor &in1Desc, cudaStream_t *streams, int n_streams)
void	mv_aTb (const raft::handle_t &handle, Matrix::Data< double > &out, const std::vector< Matrix::Data< double > * > &A, const Matrix::PartDescriptor &ADesc, const std::vector< Matrix::Data< double > * > &b, cudaStream_t *streams, int n_streams)
	performs MNMG A^T x b calculation. More...
void	mv_aTb (const raft::handle_t &handle, Matrix::Data< float > &out, const std::vector< Matrix::Data< float > * > &A, const Matrix::PartDescriptor &ADesc, const std::vector< Matrix::Data< float > * > &b, cudaStream_t *streams, int n_streams)
void	colNorm2 (const raft::handle_t &handle, Matrix::Data< double > &out, const std::vector< Matrix::Data< double > * > &in, const Matrix::PartDescriptor &inDesc, cudaStream_t *streams, int n_streams)
	performs MNMG Least squares calculation. More...
void	colNorm2 (const raft::handle_t &handle, Matrix::Data< float > &out, const std::vector< Matrix::Data< float > * > &in, const Matrix::PartDescriptor &inDesc, cudaStream_t *streams, int n_streams)
void	colNorm2NoSeq (const raft::handle_t &handle, Matrix::Data< double > &out, const std::vector< Matrix::Data< double > * > &in, const Matrix::PartDescriptor &inDesc, cudaStream_t *streams, int n_streams)
void	colNorm2NoSeq (const raft::handle_t &handle, Matrix::Data< float > &out, const std::vector< Matrix::Data< float > * > &in, const Matrix::PartDescriptor &inDesc, cudaStream_t *streams, int n_streams)
void	svdEig (const raft::handle_t &handle, const std::vector< Matrix::Data< float > * > &A, const Matrix::PartDescriptor &ADesc, std::vector< Matrix::Data< float > * > &U, float *S, float *V, cudaStream_t *streams, int n_streams)
	performs MNMG Least squares calculation. More...
void	svdEig (const raft::handle_t &handle, const std::vector< Matrix::Data< double > * > &A, const Matrix::PartDescriptor &ADesc, std::vector< Matrix::Data< double > * > &U, double *S, double *V, cudaStream_t *streams, int n_streams)

Function Documentation

colNorm2() [1/2]

void MLCommon::LinAlg::opg::colNorm2	(	const raft::handle_t &	handle,
		Matrix::Data< double > &	out,
		const std::vector< Matrix::Data< double > * > &	in,
		const Matrix::PartDescriptor &	inDesc,
		cudaStream_t *	streams,
		int	n_streams
	)

performs MNMG Least squares calculation.

colNorm2() [2/2]

void MLCommon::LinAlg::opg::colNorm2	(	const raft::handle_t &	handle,
		Matrix::Data< float > &	out,
		const std::vector< Matrix::Data< float > * > &	in,
		const Matrix::PartDescriptor &	inDesc,
		cudaStream_t *	streams,
		int	n_streams
	)

colNorm2NoSeq() [1/2]

void MLCommon::LinAlg::opg::colNorm2NoSeq	(	const raft::handle_t &	handle,
		Matrix::Data< double > &	out,
		const std::vector< Matrix::Data< double > * > &	in,
		const Matrix::PartDescriptor &	inDesc,
		cudaStream_t *	streams,
		int	n_streams
	)

colNorm2NoSeq() [2/2]

void MLCommon::LinAlg::opg::colNorm2NoSeq	(	const raft::handle_t &	handle,
		Matrix::Data< float > &	out,
		const std::vector< Matrix::Data< float > * > &	in,
		const Matrix::PartDescriptor &	inDesc,
		cudaStream_t *	streams,
		int	n_streams
	)

eigDC() [1/2]

void MLCommon::LinAlg::opg::eigDC	(	const raft::handle_t &	h,
		double *	eigenValues,
		double *	eigenVectors,
		std::vector< Matrix::Data< double > * > &	inParts,
		Matrix::PartDescriptor &	desc,
		int	myRank,
		cudaStream_t	stream
	)

eigDC() [2/2]

void MLCommon::LinAlg::opg::eigDC	(	const raft::handle_t &	h,
		float *	eigenValues,
		float *	eigenVectors,
		std::vector< Matrix::Data< float > * > &	inParts,
		Matrix::PartDescriptor &	desc,
		int	myRank,
		cudaStream_t	stream
	)

Multi-GPU version of Eigen decomposition. This function works for symmetric matrices only. Whole input matrix is gathered at rank 0 and Eigen decomposition is carried out sequentially.

Parameters

	h	cuML handle object.
[out]	eigenValues	Output N Eigen values.
[out]	eigenVectors	Output N Eigen vectors of size N x 1.
[in]	inParts	Input symmetric matrix of size N x N.
[in]	desc	Descriptor of input matrix.
[in]	myRank	MPI rank of the process.
[in]	stream	CUDA stream.

eigJacobi() [1/2]

void MLCommon::LinAlg::opg::eigJacobi	(	const raft::handle_t &	h,
		double *	eigenValues,
		double *	eigenVectors,
		std::vector< Matrix::Data< double > * > &	inParts,
		Matrix::PartDescriptor &	desc,
		int	myRank,
		cudaStream_t	stream
	)

eigJacobi() [2/2]

void MLCommon::LinAlg::opg::eigJacobi	(	const raft::handle_t &	h,
		float *	eigenValues,
		float *	eigenVectors,
		std::vector< Matrix::Data< float > * > &	inParts,
		Matrix::PartDescriptor &	desc,
		int	myRank,
		cudaStream_t	stream
	)

gemm() [1/2]

void MLCommon::LinAlg::opg::gemm	(	const raft::handle_t &	h,
		std::vector< Matrix::Data< double > * > &	outZParts,
		Matrix::PartDescriptor &	outZDesc,
		std::vector< Matrix::Data< double > * > &	inXParts,
		Matrix::PartDescriptor &	inXDesc,
		std::vector< Matrix::Data< double > * > &	inYParts,
		Matrix::PartDescriptor &	inYDesc,
		int	myRank,
		cudaStream_t	stream
	)

gemm() [2/2]

void MLCommon::LinAlg::opg::gemm	(	const raft::handle_t &	h,
		std::vector< Matrix::Data< float > * > &	outZParts,
		Matrix::PartDescriptor &	outZDesc,
		std::vector< Matrix::Data< float > * > &	inXParts,
		Matrix::PartDescriptor &	inXDesc,
		std::vector< Matrix::Data< float > * > &	inYParts,
		Matrix::PartDescriptor &	inYDesc,
		int	myRank,
		cudaStream_t	stream
	)

A multi gpu generalized matrix multiplication function. This function performs Z = X * Y The X and Y matrix are distributed in blocks on different ranks. First Y matrix is duplicated at each rank. It is multiplied with blocks of X local to the rank.

Parameters

	h	cuML handle object.
[out]	outZParts	Result of the multiplication with size M x N. Distributed across ranks according to inXDesc/outZDesc descriptor.
[out]	outZDesc	Descriptor for outZParts matrix. It has to be same as inXDesc.
[in]	inXParts	Input matrix X with dimensions M x K. Distributed across ranks according to inXDesc descriptor.
[in]	inXDesc	Descriptor for X matrix.
[in]	inYParts	Input matrix Y with dimensions K x N. Distributed across ranks according to inYDesc descriptor.
[in]	inYDesc	Descriptor for Y matrix.
[in]	myRank	Rank of calling process.
[in]	stream	cuda stream on which work is launched.

lstsqEig() [1/2]

void MLCommon::LinAlg::opg::lstsqEig	(	const raft::handle_t &	handle,
		const std::vector< Matrix::Data< double > * > &	A,
		const Matrix::PartDescriptor &	ADesc,
		const std::vector< Matrix::Data< double > * > &	b,
		double *	w,
		cudaStream_t *	streams,
		int	n_streams
	)

lstsqEig() [2/2]

void MLCommon::LinAlg::opg::lstsqEig	(	const raft::handle_t &	handle,
		const std::vector< Matrix::Data< float > * > &	A,
		const Matrix::PartDescriptor &	ADesc,
		const std::vector< Matrix::Data< float > * > &	b,
		float *	w,
		cudaStream_t *	streams,
		int	n_streams
	)

performs MNMG Least squares calculation.

meanSquaredError() [1/2]

void MLCommon::LinAlg::opg::meanSquaredError	(	double *	out,
		const Matrix::Data< double > &	in1,
		const Matrix::PartDescriptor &	in1Desc,
		const Matrix::Data< double > &	in2,
		const Matrix::PartDescriptor &	in2Desc,
		const raft::comms::comms_t &	comm,
		cudaStream_t	stream,
		int	root = 0,
		bool	broadcastResult = true
	)

multi-gpu mean squared error

Parameters

[out]	out	the output data (device pointer)
[in]	in1	the first input data
[in]	in1Desc	descriptor for the first input data
[in]	in2	the second input data
[in]	in2Desc	descriptor for the second input data
[in]	comm	the communicator object
[in]	stream	cuda stream where to launch work
[in]	root	worker ID which is supposed to be considered as root
[in]	broadcastResult	if false, only root process will have the result, else all ranks

meanSquaredError() [2/2]

void MLCommon::LinAlg::opg::meanSquaredError	(	float *	out,
		const Matrix::Data< float > &	in1,
		const Matrix::PartDescriptor &	in1Desc,
		const Matrix::Data< float > &	in2,
		const Matrix::PartDescriptor &	in2Desc,
		const raft::comms::comms_t &	comm,
		cudaStream_t	stream,
		int	root = 0,
		bool	broadcastResult = true
	)

mm_aTa() [1/2]

void MLCommon::LinAlg::opg::mm_aTa	(	const raft::handle_t &	handle,
		Matrix::Data< double > &	out,
		const std::vector< Matrix::Data< double > * > &	in1,
		const Matrix::PartDescriptor &	in1Desc,
		cudaStream_t *	streams,
		int	n_streams
	)

performs MNMG A^T x A calculation.

Parameters

[in]	handle	cuML handle object
[out]	out	resulting A^T x A matrix
[in]	in1	input data
[in]	in1Desc	MNMG description of the input
[in]	streams	cuda streams
[in]	n_streams	number of streams

mm_aTa() [2/2]

void MLCommon::LinAlg::opg::mm_aTa	(	const raft::handle_t &	handle,
		Matrix::Data< float > &	out,
		const std::vector< Matrix::Data< float > * > &	in1,
		const Matrix::PartDescriptor &	in1Desc,
		cudaStream_t *	streams,
		int	n_streams
	)

mv_aTb() [1/2]

void MLCommon::LinAlg::opg::mv_aTb	(	const raft::handle_t &	handle,
		Matrix::Data< double > &	out,
		const std::vector< Matrix::Data< double > * > &	A,
		const Matrix::PartDescriptor &	ADesc,
		const std::vector< Matrix::Data< double > * > &	b,
		cudaStream_t *	streams,
		int	n_streams
	)

performs MNMG A^T x b calculation.

Parameters

[in]	handle	cuML handle object
[out]	out	resulting A^T x b matrix
[in]	A	input data matrix
[in]	ADesc	MNMG description of A
[in]	b	input vector
[in]	streams	cuda streams
[in]	n_streams	number of streams

mv_aTb() [2/2]

void MLCommon::LinAlg::opg::mv_aTb	(	const raft::handle_t &	handle,
		Matrix::Data< float > &	out,
		const std::vector< Matrix::Data< float > * > &	A,
		const Matrix::PartDescriptor &	ADesc,
		const std::vector< Matrix::Data< float > * > &	b,
		cudaStream_t *	streams,
		int	n_streams
	)

svdEig() [1/2]

void MLCommon::LinAlg::opg::svdEig	(	const raft::handle_t &	handle,
		const std::vector< Matrix::Data< double > * > &	A,
		const Matrix::PartDescriptor &	ADesc,
		std::vector< Matrix::Data< double > * > &	U,
		double *	S,
		double *	V,
		cudaStream_t *	streams,
		int	n_streams
	)

svdEig() [2/2]

void MLCommon::LinAlg::opg::svdEig	(	const raft::handle_t &	handle,
		const std::vector< Matrix::Data< float > * > &	A,
		const Matrix::PartDescriptor &	ADesc,
		std::vector< Matrix::Data< float > * > &	U,
		float *	S,
		float *	V,
		cudaStream_t *	streams,
		int	n_streams
	)

performs MNMG Least squares calculation.