exatn::runtime::TensorRuntime Class Reference

Public Member Functions
	TensorRuntime (const std::string &graph_executor_name="eager-dag-executor", const std::string &node_executor_name="talsh-node-executor")
	TensorRuntime (const TensorRuntime &)=delete
TensorRuntime &	operator= (const TensorRuntime &)=delete
	TensorRuntime (TensorRuntime &&) noexcept=delete
TensorRuntime &	operator= (TensorRuntime &&) noexcept=delete
void	resetLoggingLevel (int level=0)
void	openScope (const std::string &scope_name)
void	pauseScope ()
void	resumeScope (const std::string &scope_name)
void	closeScope ()
bool	currentScopeIsSet () const
VertexIdType	submit (std::shared_ptr< TensorOperation > op)
bool	sync (TensorOperation &op, bool wait=true)
bool	sync (const Tensor &tensor, bool wait=true)
std::future< std::shared_ptr< talsh::Tensor > >	getLocalTensor (std::shared_ptr< Tensor > tensor, const std::vector< std::pair< DimOffset, DimExtent >> &slice_spec)

Member Function Documentation

closeScope()

void exatn::runtime::TensorRuntime::closeScope

(

)

Closes the current scope, fully completing all tensor operations in the current execution graph.

currentScopeIsSet()

bool exatn::runtime::TensorRuntime::currentScopeIsSet ( ) const

inline

Returns TRUE if there is the current scope is set.

getLocalTensor()

std::future< std::shared_ptr< talsh::Tensor > > exatn::runtime::TensorRuntime::getLocalTensor	(	std::shared_ptr< Tensor >	tensor,
		const std::vector< std::pair< DimOffset, DimExtent >> &	slice_spec
	)

Returns a locally stored tensor slice (talsh::Tensor) providing access to tensor elements. This slice will be extracted from the exatn::numerics::Tensor implementation as a copy. The returned future becomes ready once the execution thread has retrieved the slice copy.

openScope()

void exatn::runtime::TensorRuntime::openScope

(

const std::string &

scope_name

)

Opens a new scope represented by a new execution graph (DAG).

pauseScope()

void exatn::runtime::TensorRuntime::pauseScope

(

)

Pauses the current scope by completing all outstanding tensor operations and pausing the further progress of the current execution graph until resume. Returns TRUE upon successful pausing, FALSE otherwise.

resetLoggingLevel()

void exatn::runtime::TensorRuntime::resetLoggingLevel

(

int

level = 0

)

Resets the logging level (0:none) [MAIN THREAD].

resumeScope()

void exatn::runtime::TensorRuntime::resumeScope

(

const std::string &

scope_name

)

Resumes the execution of a previously paused scope (execution graph).

submit()

VertexIdType exatn::runtime::TensorRuntime::submit

(

std::shared_ptr< TensorOperation >

op

)

Submits a tensor operation into the current execution graph and returns its integer id.

sync() [1/2]

bool exatn::runtime::TensorRuntime::sync	(	const Tensor &	tensor,
		bool	wait = true
	)

Tests for completion of all outstanding update operations on a given tensor. If wait = TRUE, it will block until completion.

sync() [2/2]

bool exatn::runtime::TensorRuntime::sync	(	TensorOperation &	op,
		bool	wait = true
	)

Tests for completion of a given tensor operation. If wait = TRUE, it will block until completion.

---

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

• /Users/aqw/exatn/src/runtime/tensor_runtime.hpp
• /Users/aqw/exatn/src/runtime/tensor_runtime.cpp