Class cv::cann::AscendStream#

In AscendCL Stream(AscendStream) is a task queue. Stream is used to manage the parallelism of tasks. The tasks inside a Stream are executed sequentially, that is, the Stream executes sequentially according to the sent tasks; the tasks in different Streams are executed in parallel. View details

#include <opencv2/cann.hpp>

Collaboration diagram for cv::cann::AscendStream:

Friends#

Return

Name

Description

class

AscendStreamAccessor

class

DefaultDeviceInitializer

Detailed Description#

In AscendCL Stream(AscendStream) is a task queue. Stream is used to manage the parallelism of tasks. The tasks inside a Stream are executed sequentially, that is, the Stream executes sequentially according to the sent tasks; the tasks in different Streams are executed in parallel.

All Non-blocking functions should pass parameter stream, These function returns immediately after the task is submitted. Caller should wait stream until completion.

Blocking functions implicityly use the default stream, and synchronize stream before function return.

See also

cuda::Stream

Constructor & Destructor Documentation#

AscendStream()#

cv::cann::AscendStream::AscendStream()

AscendStream()#

cv::cann::AscendStream::AscendStream(const Ptr< Impl > & impl)

Member Function Documentation#

addTensorHolder()#

void cv::cann::AscendStream::addTensorHolder(const std::shared_ptr< uchar > & holder)

waitAscendEvent()#

void cv::cann::AscendStream::waitAscendEvent(const cv::cann::AscendEvent & event)

blocks the current CPU thread until event trigger.

waitForCompletion()#

void cv::cann::AscendStream::waitForCompletion()

blocks the current CPU thread until all operations in the stream are complete.

Null()#

static AscendStream & cv::cann::AscendStream::Null()

return default AscendStream object for default Acl stream.

Member Data Documentation#

impl_#

Ptr< Impl > cv::cann::AscendStream::impl_

Source file#

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