Pseudo expression providing an operator = that will evaluate its argument asynchronously on the specified device. Currently only ThreadPoolDevice implements proper asynchronous execution, while the default and GPU devices just run the expression synchronously and call m_done() on completion.. More...

#include <TensorDevice.h>

Public Member Functions
	TensorAsyncDevice (const DeviceType &device, ExpressionType &expression, DoneCallback done)

template<typename OtherDerived >
EIGEN_STRONG_INLINE TensorAsyncDevice &	operator= (const OtherDerived &other)

Protected Attributes
const DeviceType &	m_device

ExpressionType &	m_expression

DoneCallback	m_done

Detailed Description

template<typename ExpressionType, typename DeviceType, typename DoneCallback>
class Eigen::TensorAsyncDevice< ExpressionType, DeviceType, DoneCallback >

Pseudo expression providing an operator = that will evaluate its argument asynchronously on the specified device. Currently only ThreadPoolDevice implements proper asynchronous execution, while the default and GPU devices just run the expression synchronously and call m_done() on completion..

Example: auto done = []() { ... expression evaluation done ... }; C.device(thread_pool_device, std::move(done)) = A + B;