#include <deal.II/base/hdf5.h>

Inheritance diagram for HDF5::DataSet:

Public Member Functions
template<typename Container >
Container	read ()

template<typename Container >
Container	read_selection (const std::vector< hsize_t > &coordinates)

template<typename Container >
Container	read_hyperslab (const std::vector< hsize_t > &offset, const std::vector< hsize_t > &count)

template<typename Container >
Container	read_hyperslab (const std::vector< hsize_t > &data_dimensions, const std::vector< hsize_t > &offset, const std::vector< hsize_t > &stride, const std::vector< hsize_t > &count, const std::vector< hsize_t > &block)

template<typename number >
void	read_none ()

template<typename Container >
void	write (const Container &data)

template<typename Container >
void	write_selection (const Container &data, const std::vector< hsize_t > &coordinates)

template<typename Container >
void	write_hyperslab (const Container &data, const std::vector< hsize_t > &offset, const std::vector< hsize_t > &count)

template<typename Container >
void	write_hyperslab (const Container &data, const std::vector< hsize_t > &data_dimensions, const std::vector< hsize_t > &offset, const std::vector< hsize_t > &stride, const std::vector< hsize_t > &count, const std::vector< hsize_t > &block)

template<typename number >
void	write_none ()

bool	get_query_io_mode () const

void	set_query_io_mode (const bool new_query_io_mode)

std::string	get_io_mode ()

H5D_mpio_actual_io_mode_t	get_io_mode_as_hdf5_type ()

std::string	get_local_no_collective_cause ()

uint32_t	get_local_no_collective_cause_as_hdf5_type ()

std::string	get_global_no_collective_cause ()

uint32_t	get_global_no_collective_cause_as_hdf5_type ()

std::vector< hsize_t >	get_dimensions () const

unsigned int	get_size () const

unsigned int	get_rank () const

template<typename T >
T	get_attribute (const std::string &attr_name) const

template<>
std::string	get_attribute (const std::string &attr_name) const

template<typename T >
void	set_attribute (const std::string &attr_name, const T value)

template<>
void	set_attribute (const std::string &attr_name, const std::string value)

std::string	get_name () const

Protected Member Functions
	DataSet (const std::string &name, const hid_t &parent_group_id, bool mpi)

	DataSet (const std::string &name, const hid_t &parent_group_id, const std::vector< hsize_t > &dimensions, const std::shared_ptr< hid_t > &t_type, const bool mpi)

Protected Attributes
const std::string	name

std::shared_ptr< hid_t >	hdf5_reference

const bool	mpi

Private Attributes
unsigned int	rank

std::vector< hsize_t >	dimensions

std::shared_ptr< hid_t >	dataspace

unsigned int	size

bool	query_io_mode

H5D_mpio_actual_io_mode_t	io_mode

uint32_t	local_no_collective_cause

uint32_t	global_no_collective_cause

Friends
class	Group

Detailed Description

This class implements an HDF5 DataSet.

Definition at line 426 of file hdf5.h.

Constructor & Destructor Documentation

◆ DataSet() [1/2]

HDF5::DataSet::DataSet	(	const std::string &	name,
		const hid_t &	parent_group_id,
		bool	mpi
	)

protected

Open dataset. This is an internal constructor. The function Group::open_dataset() should be used to open a dataset.

Definition at line 89 of file hdf5.cc.

◆ DataSet() [2/2]

HDF5::DataSet::DataSet	(	const std::string &	name,
		const hid_t &	parent_group_id,
		const std::vector< hsize_t > &	dimensions,
		const std::shared_ptr< hid_t > &	t_type,
		const bool	mpi
	)

protected

Create dataset. This is an internal constructor. The function Group::create_dataset() should be used to create a dataset.

Definition at line 144 of file hdf5.cc.

Member Function Documentation

◆ read()

template<typename Container >

Container HDF5::DataSet::read

Reads all the data of the dataset.

Datatype conversion takes place at the time of the read operation and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

Container can be std::vector<float>, std::vector<double>, std::vector<std::complex<float>>, std::vector<std::complex<double>>, std::vector<int>, std::vector<unsigned int>, Vector<float>, Vector<double>, Vector<std::complex<float>>, Vector<std::complex<double>>, FullMatrix<float>, FullMatrix<double>, FullMatrix<std::complex<float>> or FullMatrix<std::complex<double>>.

Definition at line 1760 of file hdf5.h.

◆ read_selection()

template<typename Container >

Container HDF5::DataSet::read_selection ( const std::vector< hsize_t > & coordinates )

Reads data of a subset of the dataset.

Datatype conversion takes place at the time of the read operation and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

The selected elements can be scattered and take any shape in the dataset. For example, in the case of a dataset with rank 4 a selection of 3 points will be described by a 3-by-4 array. Note the indexing is zero-based. To select the points (1,1,1,1), (14,6,12,18), and (8,22,30,22), the point selection array would be as follows:

Parallel HDF5 supports collective I/O on point selections.

Datatype conversion takes place at the time of the read operation and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

Definition at line 1795 of file hdf5.h.

◆ read_hyperslab() [1/2]

template<typename Container >

Container HDF5::DataSet::read_hyperslab	(	const std::vector< hsize_t > &	offset,
		const std::vector< hsize_t > &	count
	)

Reads a hyperslab from the dataset. The parameters are summarized below:

offset: The starting location for the hyperslab.
count: The number of elements to select along each dimension.

When reading a hyperslab, HDF5 also allows to provide "stride" and "block" parameters (see the HDF5 documentation). These are not used by the current function and set to nullptr. However these parameters can be used with the function read_hyperslab(const std::vector<hsize_t> &, const std::vector<hsize_t> &, const std::vector<hsize_t> &, const std::vector<hsize_t> &, const std::vector<hsize_t> &)

See the Dataspaces and Data Transfer section in the HDF5 User's Guide. See as well the H5Sselect_hyperslab definition.

Datatype conversion takes place at the time of a read or write and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

Container can be std::vector<float>, std::vector<double>, std::vector<std::complex<float>>, std::vector<std::complex<double>>, std::vector<int>, std::vector<unsigned int>, Vector<float>, Vector<double>, Vector<std::complex<float>>, Vector<std::complex<double>>, FullMatrix<float>, FullMatrix<double>, FullMatrix<std::complex<float>> or FullMatrix<std::complex<double>>.

Definition at line 1847 of file hdf5.h.

◆ read_hyperslab() [2/2]

template<typename Container >

Container HDF5::DataSet::read_hyperslab	(	const std::vector< hsize_t > &	data_dimensions,
		const std::vector< hsize_t > &	offset,
		const std::vector< hsize_t > &	stride,
		const std::vector< hsize_t > &	count,
		const std::vector< hsize_t > &	block
	)

Writes a data hyperslab to the dataset. The parameters are summarized below:

dataset_dimensions: the dimensions of the data memory block.
offset: The starting location for the hyperslab.
stride: The number of elements to separate each element or block to be selected.
count: The number of elements or blocks to select along each dimension.
block: The size of the block selected from the dataspace.

See the Dataspaces and Data Transfer section in the HDF5 User's Guide. See as well the H5Sselect_hyperslab definition.

Datatype conversion takes place at the time of a read or write and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

Container can be std::vector<float>, std::vector<double>, std::vector<std::complex<float>>, std::vector<std::complex<double>>, std::vector<int>, std::vector<unsigned int>, Vector<float>, Vector<double>, Vector<std::complex<float>>, Vector<std::complex<double>>, FullMatrix<float>, FullMatrix<double>, FullMatrix<std::complex<float>> or FullMatrix<std::complex<double>>.

Definition at line 1901 of file hdf5.h.

◆ read_none()

template<typename number >

void HDF5::DataSet::read_none

This function does not read any data, but it can contribute to a collective read call. number can be float, double, std::complex<float>, std::complex<double>, int or unsigned int.

Datatype conversion takes place at the time of a read or write and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

Definition at line 1954 of file hdf5.h.

◆ write()

template<typename Container >

void HDF5::DataSet::write ( const Container & data )

Writes data in the dataset. number can be float, double, std::complex<float>, std::complex<double>, int or unsigned int.

Datatype conversion takes place at the time of a read or write and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

Container can be std::vector<float>, std::vector<double>, std::vector<std::complex<float>>, std::vector<std::complex<double>>, std::vector<int>, std::vector<unsigned int>, Vector<float>, Vector<double>, Vector<std::complex<float>>, Vector<std::complex<double>>, FullMatrix<float>, FullMatrix<double>, FullMatrix<std::complex<float>> or FullMatrix<std::complex<double>>.

Definition at line 1994 of file hdf5.h.

◆ write_selection()

template<typename Container >

void HDF5::DataSet::write_selection	(	const Container &	data,
		const std::vector< hsize_t > &	coordinates
	)

Writes data to a subset of the dataset. number can be float, double, std::complex<float>, std::complex<double>, int or unsigned int.

The selected elements can be scattered and take any shape in the dataset. For example, in the case of a dataset with rank 4 a selection of 3 points will be described by a 3-by-4 array. Note the indexing is zero-based. To select the points (1,1,1,1), (14,6,12,18), and (8,22,30,22), the point selection array would be as follows:

Parallel HDF5 supports collective I/O on point selections.

Datatype conversion takes place at the time of a read or write and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

Definition at line 2026 of file hdf5.h.

◆ write_hyperslab() [1/2]

template<typename Container >

void HDF5::DataSet::write_hyperslab	(	const Container &	data,
		const std::vector< hsize_t > &	offset,
		const std::vector< hsize_t > &	count
	)

Writes a data hyperslab to the dataset. The parameters are summarized below:

offset: The starting location for the hyperslab.
count: The number of elements to select along each dimension.

When writing a hyperslab, HDF5 also allows to provide "stride" and "block" parameters (see the HDF5 documentation). These are not used by the current function and set to nullptr. However these parameters can be used with the function write_hyperslab(const Container &data, const std::vector<hsize_t> &data_dimensions, const std::vector<hsize_t> &offset, const std::vector<hsize_t> &stride, const std::vector<hsize_t> &count, const std::vector<hsize_t> &block).

See the Dataspaces and Data Transfer section in the HDF5 User's Guide. See as well the H5Sselect_hyperslab definition.

Datatype conversion takes place at the time of a read or write and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

Definition at line 2075 of file hdf5.h.

◆ write_hyperslab() [2/2]

template<typename Container >

void HDF5::DataSet::write_hyperslab	(	const Container &	data,
		const std::vector< hsize_t > &	data_dimensions,
		const std::vector< hsize_t > &	offset,
		const std::vector< hsize_t > &	stride,
		const std::vector< hsize_t > &	count,
		const std::vector< hsize_t > &	block
	)

Writes a data hyperslab to the dataset. The parameters are summarized below:

dataset_dimensions: the dimensions of the data memory block.
offset: The starting location for the hyperslab.
stride: The number of elements to separate each element or block to be selected.
count: The number of elements or blocks to select along each dimension.
block: The size of the block selected from the dataspace.

See the Dataspaces and Data Transfer section in the HDF5 User's Guide. See as well the H5Sselect_hyperslab definition.

Datatype conversion takes place at the time of a read or write and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

Container can be std::vector<float>, std::vector<double>, std::vector<std::complex<float>>, std::vector<std::complex<double>>, std::vector<int>, std::vector<unsigned int>, Vector<float>, Vector<double>, Vector<std::complex<float>>, Vector<std::complex<double>>, FullMatrix<float>, FullMatrix<double>, FullMatrix<std::complex<float>> or FullMatrix<std::complex<double>>.

Definition at line 2132 of file hdf5.h.

◆ write_none()

template<typename number >

void HDF5::DataSet::write_none

This function does not write any data, but it can contribute to a collective write call. In the context of a collective MPI write call, if a process does not write any data at all, the process should call this function because the operation is collective and all the MPI processes have to contribute to the call, even if they don't have data to write. number can be float, double, std::complex<float>, std::complex<double>, int or unsigned int.

Datatype conversion takes place at the time of a read or write and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

An example of how to use this function can be found in step-62.

Definition at line 2184 of file hdf5.h.

◆ get_query_io_mode()

bool HDF5::DataSet::get_query_io_mode ( ) const

This function returns the boolean query_io_mode.

In cases where maximum performance has to be achieved, it is important to make sure that all MPI read/write operations are collective. The HDF5 library provides API routines that can be used after the read/write I/O operations to query the I/O mode. If query_io_mode is set to true, then after every read/write operation the deal.II's HDF5 interface calls the routines H5Pget_mpio_actual_io_mode() and H5Pget_mpio_no_collective_cause(). The results are stored in io_mode, local_no_collective_cause and global_no_collective_cause. We suggest to query the I/O mode only in Debug mode because it requires calling additional HDF5 routines.

Definition at line 306 of file hdf5.cc.

◆ set_query_io_mode()

void HDF5::DataSet::set_query_io_mode ( const bool new_query_io_mode )

This function sets the boolean query_io_mode.

Definition at line 198 of file hdf5.cc.

◆ get_io_mode()

std::string HDF5::DataSet::get_io_mode ( )

This function returns the I/O mode that was used on the last parallel I/O call. See H5Pget_mpio_actual_io_mode.

The return value is a std::string and can be

Value	Meaning
H5D_MPIO_NO_COLLECTIVE	No collective I/O was performed. Collective I/O was not requested or collective I/O isn't possible on this dataset.
H5D_MPIO_CHUNK_INDEPENDENT	HDF5 performed chunk collective optimization schemes and each chunk was accessed independently.
H5D_MPIO_CHUNK_COLLECTIVE	HDF5 performed chunk collective optimization and each chunk was accessed collectively.
H5D_MPIO_CHUNK_MIXED	HDF5 performed chunk collective optimization and some chunks were accessed independently, some collectively.
H5D_MPIO_CONTIGUOUS_COLLECTIVE	Collective I/O was performed on a contiguous dataset.

Definition at line 214 of file hdf5.cc.

◆ get_io_mode_as_hdf5_type()

H5D_mpio_actual_io_mode_t HDF5::DataSet::get_io_mode_as_hdf5_type ( )

This function returns the I/O mode that was used on the last parallel I/O call. See H5Pget_mpio_actual_io_mode. The return type is H5D_mpio_actual_io_mode_t which corresponds to the value returned by H5Pget_mpio_actual_io_mode.

The return value can be

Value	Meaning
H5D_MPIO_NO_COLLECTIVE	No collective I/O was performed. Collective I/O was not requested or collective I/O isn't possible on this dataset.
H5D_MPIO_CHUNK_INDEPENDENT	HDF5 performed chunk collective optimization and each chunk was accessed independently.
H5D_MPIO_CHUNK_COLLECTIVE	HDF5 performed chunk collective optimization and each chunk was accessed collectively.
H5D_MPIO_CHUNK_MIXED	HDF5 performed chunk collective optimization and some chunks were accessed independently, some collectively.
H5D_MPIO_CONTIGUOUS_COLLECTIVE	Collective I/O was performed on a contiguous dataset.

Definition at line 248 of file hdf5.cc.

◆ get_local_no_collective_cause()

std::string HDF5::DataSet::get_local_no_collective_cause ( )

This function returns the local causes that broke collective I/O on the last parallel I/O call. See H5Pget_mpio_no_collective_cause.

The return value is a string and can be

Value	Meaning
H5D_MPIO_COLLECTIVE	Collective I/O was performed successfully.
H5D_MPIO_SET_INDEPENDENT	Collective I/O was not performed because independent I/O was requested.
H5D_MPIO_DATATYPE_CONVERSION	Collective I/O was not performed because datatype conversions were required.
H5D_MPIO_DATA_TRANSFORMS	Collective I/O was not performed because data transforms needed to be applied.
H5D_MPIO_SET_MPIPOSIX	Collective I/O was not performed because the selected file driver was MPI-POSIX.
H5D_MPIO_NOT_SIMPLE_OR_SCALAR_DATASPACES	Collective I/O was not performed because one of the dataspaces was neither simple nor scalar.
H5D_MPIO_POINT_SELECTIONS	Collective I/O was not performed because there were point selections in one of the dataspaces.
H5D_MPIO_NOT_CONTIGUOUS_OR_CHUNKED_DATASET	Collective I/O was not performed because the dataset was neither contiguous nor chunked.
H5D_MPIO_FILTERS	Collective I/O was not performed because filters needed to be applied.

Definition at line 259 of file hdf5.cc.

◆ get_local_no_collective_cause_as_hdf5_type()

uint32_t HDF5::DataSet::get_local_no_collective_cause_as_hdf5_type ( )

This function returns the local causes that broke collective I/O on the last parallel I/O call. See H5Pget_mpio_no_collective_cause. The return type is uint32_t and corresponds to the value returned by H5Pget_mpio_no_collective_cause.

The return value can be

Value	Meaning
H5D_MPIO_COLLECTIVE	Collective I/O was performed successfully.
H5D_MPIO_SET_INDEPENDENT	Collective I/O was not performed because independent I/O was requested.
H5D_MPIO_DATATYPE_CONVERSION	Collective I/O was not performed because datatype conversions were required.
H5D_MPIO_DATA_TRANSFORMS	Collective I/O was not performed because data transforms needed to be applied.
H5D_MPIO_SET_MPIPOSIX	Collective I/O was not performed because the selected file driver was MPI-POSIX.
H5D_MPIO_NOT_SIMPLE_OR_SCALAR_DATASPACES	Collective I/O was not performed because one of the dataspaces was neither simple nor scalar.
H5D_MPIO_POINT_SELECTIONS	Collective I/O was not performed because there were point selections in one of the dataspaces.
H5D_MPIO_NOT_CONTIGUOUS_OR_CHUNKED_DATASET	Collective I/O was not performed because the dataset was neither contiguous nor chunked.
H5D_MPIO_FILTERS	Collective I/O was not performed because filters needed to be applied.

Definition at line 271 of file hdf5.cc.

◆ get_global_no_collective_cause()

std::string HDF5::DataSet::get_global_no_collective_cause ( )

This function retrieves the global causes that broke collective I/O on the last parallel I/O call. See H5Pget_mpio_no_collective_cause.

The return value is a std::string and can be

Value	Meaning
H5D_MPIO_COLLECTIVE	Collective I/O was performed successfully.
H5D_MPIO_SET_INDEPENDENT	Collective I/O was not performed because independent I/O was requested.
H5D_MPIO_DATATYPE_CONVERSION	Collective I/O was not performed because datatype conversions were required.
H5D_MPIO_DATA_TRANSFORMS	Collective I/O was not performed because data transforms needed to be applied.
H5D_MPIO_SET_MPIPOSIX	Collective I/O was not performed because the selected file driver was MPI-POSIX.
H5D_MPIO_NOT_SIMPLE_OR_SCALAR_DATASPACES	Collective I/O was not performed because one of the dataspaces was neither simple nor scalar.
H5D_MPIO_POINT_SELECTIONS	Collective I/O was not performed because there were point selections in one of the dataspaces.
H5D_MPIO_NOT_CONTIGUOUS_OR_CHUNKED_DATASET	Collective I/O was not performed because the dataset was neither contiguous nor chunked.
H5D_MPIO_FILTERS	Collective I/O was not performed because filters needed to be applied.

Definition at line 283 of file hdf5.cc.

◆ get_global_no_collective_cause_as_hdf5_type()

uint32_t HDF5::DataSet::get_global_no_collective_cause_as_hdf5_type ( )

This function returns the global causes that broke collective I/O on the last parallel I/O call. See H5Pget_mpio_no_collective_cause. The return type is uint32_t and corresponds to the value returned by H5Pget_mpio_no_collective_cause.

The return value can be

Value	Meaning
H5D_MPIO_COLLECTIVE	Collective I/O was performed successfully.
H5D_MPIO_SET_INDEPENDENT	Collective I/O was not performed because independent I/O was requested.
H5D_MPIO_DATATYPE_CONVERSION	Collective I/O was not performed because datatype conversions were required.
H5D_MPIO_DATA_TRANSFORMS	Collective I/O was not performed because data transforms needed to be applied.
H5D_MPIO_SET_MPIPOSIX	Collective I/O was not performed because the selected file driver was MPI-POSIX.
H5D_MPIO_NOT_SIMPLE_OR_SCALAR_DATASPACES	Collective I/O was not performed because one of the dataspaces was neither simple nor scalar.
H5D_MPIO_POINT_SELECTIONS	Collective I/O was not performed because there were point selections in one of the dataspaces.
H5D_MPIO_NOT_CONTIGUOUS_OR_CHUNKED_DATASET	Collective I/O was not performed because the dataset was neither contiguous nor chunked.
H5D_MPIO_FILTERS	Collective I/O was not performed because filters needed to be applied.

Definition at line 295 of file hdf5.cc.

◆ get_dimensions()

std::vector< hsize_t > HDF5::DataSet::get_dimensions ( ) const

This function returns the dimensions of the dataset. The vector dimensions is a one-dimensional array of size rank specifying the size of each dimension of the dataset.

Definition at line 206 of file hdf5.cc.

◆ get_size()

unsigned int HDF5::DataSet::get_size ( ) const

This function returns the total number of elements in the dataset.

Definition at line 314 of file hdf5.cc.

◆ get_rank()

unsigned int HDF5::DataSet::get_rank ( ) const

This function returns the rank of the dataset.

Definition at line 322 of file hdf5.cc.

◆ get_attribute() [1/2]

template<typename T >

T HDF5::HDF5Object::get_attribute ( const std::string & attr_name ) const

inherited

Reads an attribute. T can be float, double, std::complex<float>, std::complex<double>, int, unsigned int, bool or std::string. Note that the encoding of std::string is UTF8 in order to be compatible with python3.

Datatype conversion takes place at the time of a read or write and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

Definition at line 1588 of file hdf5.h.

◆ get_attribute() [2/2]

template<>

std::string HDF5::HDF5Object::get_attribute ( const std::string & attr_name ) const

inlineinherited

Definition at line 1611 of file hdf5.h.

◆ set_attribute() [1/2]

template<typename T >

void HDF5::HDF5Object::set_attribute	(	const std::string &	attr_name,
		const T	value
	)

inherited

Writes an attribute. T can be float, double, std::complex<float>, std::complex<double>, int, unsigned int, bool or std::string. Note that the encoding of std::string is UTF8 in order to be compatible with python3.

Datatype conversion takes place at the time of a read or write and is automatic. See the Data Transfer: Datatype Conversion and Selection section in the HDF5 User's Guide.

Definition at line 1665 of file hdf5.h.

◆ set_attribute() [2/2]

template<>

void HDF5::HDF5Object::set_attribute	(	const std::string &	attr_name,
		const std::string	value
	)

inlineinherited

Definition at line 1705 of file hdf5.h.

◆ get_name()

std::string HDF5::HDF5Object::get_name ( ) const

inherited

Returns the name of the object. In the case of File, name corresponds to the file name. In the case of Group and DataSet, name corresponds to the name of the object in the HDF5 file.

Definition at line 82 of file hdf5.cc.

Friends And Related Function Documentation

◆ Group

friend class Group

friend

Definition at line 428 of file hdf5.h.

Member Data Documentation

◆ rank

unsigned int HDF5::DataSet::rank

private

Rank of the DataSet

Definition at line 924 of file hdf5.h.

◆ dimensions

std::vector<hsize_t> HDF5::DataSet::dimensions

private

The vector dimensions is a one-dimensional array of size rank specifying the size of each dimension of the dataset.

Definition at line 930 of file hdf5.h.

◆ dataspace

std::shared_ptr<hid_t> HDF5::DataSet::dataspace

private

HDF5 dataspace identifier.

Definition at line 935 of file hdf5.h.

◆ size

unsigned int HDF5::DataSet::size

private

Total number of elements in the dataset.

Definition at line 940 of file hdf5.h.

◆ query_io_mode

bool HDF5::DataSet::query_io_mode

private

If query_io_mode is set to true, then after every read/write operation the deal.II's HDF5 interface calls the routines H5Pget_mpio_actual_io_mode() and H5Pget_mpio_no_collective_cause(). The results are stored in io_mode, local_no_collective_cause and global_no_collective_cause.

Definition at line 951 of file hdf5.h.

◆ io_mode

H5D_mpio_actual_io_mode_t HDF5::DataSet::io_mode

private

I/O mode that was performed on the last parallel I/O call.

Definition at line 956 of file hdf5.h.

◆ local_no_collective_cause

uint32_t HDF5::DataSet::local_no_collective_cause

private

Local causes that broke collective I/O on the last parallel I/O call. See H5Pget_mpio_no_collective_cause.

Definition at line 963 of file hdf5.h.

◆ global_no_collective_cause

uint32_t HDF5::DataSet::global_no_collective_cause

private

Global causes that broke collective I/O on the last parallel I/O call. See H5Pget_mpio_no_collective_cause.

Definition at line 970 of file hdf5.h.

◆ name

const std::string HDF5::HDF5Object::name

protectedinherited

Name of the HDF5Oject. In the case of File, name corresponds to the file name. In the case of Group and DataSet name corresponds to the name of the object in the HDF5 file.

Definition at line 406 of file hdf5.h.

◆ hdf5_reference

std::shared_ptr<hid_t> HDF5::HDF5Object::hdf5_reference

protectedinherited

HDF5 identifier for the objects File, Group and DataSet. The std::shared_ptr<> pointer allows the object to be copied. For example several parts of the program can share and access the same group; when all the functions that access the group are closed, the HDF5 resources of the group will be automatically released.

Definition at line 415 of file hdf5.h.

◆ mpi

const bool HDF5::HDF5Object::mpi

protectedinherited

If true use parallel HDF5, if false use serial HDF5.

Definition at line 420 of file hdf5.h.

The documentation for this class was generated from the following files:

include/deal.II/base/hdf5.h
source/base/hdf5.cc

Public Member Functions

Protected Member Functions

Protected Attributes

Private Attributes

Friends

Detailed Description

Constructor & Destructor Documentation

◆ DataSet() [1/2]

◆ DataSet() [2/2]

Member Function Documentation

◆ read()

◆ read_selection()

◆ read_hyperslab() [1/2]

◆ read_hyperslab() [2/2]

◆ read_none()

◆ write()

◆ write_selection()

◆ write_hyperslab() [1/2]

◆ write_hyperslab() [2/2]

◆ write_none()

◆ get_query_io_mode()

◆ set_query_io_mode()

◆ get_io_mode()

◆ get_io_mode_as_hdf5_type()

◆ get_local_no_collective_cause()

◆ get_local_no_collective_cause_as_hdf5_type()

◆ get_global_no_collective_cause()

◆ get_global_no_collective_cause_as_hdf5_type()

◆ get_dimensions()

◆ get_size()

◆ get_rank()

◆ get_attribute() [1/2]

◆ get_attribute() [2/2]

◆ set_attribute() [1/2]

◆ set_attribute() [2/2]

◆ get_name()

Friends And Related Function Documentation

◆ Group

Member Data Documentation

◆ rank

◆ dimensions

◆ dataspace

◆ size

◆ query_io_mode

◆ io_mode

◆ local_no_collective_cause

◆ global_no_collective_cause

◆ name

◆ hdf5_reference

◆ mpi