tf::cudaStream class

#include <taskflow/cuda/cuda_stream.hpp>

tf::cudaStream::cudaStream(cudaStream_t stream) explicit

constructs an RAII-styled object from the given CUDA stream

Constructs a cudaStream object which owns stream.

void tf::cudaStream::synchronize() const

synchronizes the associated stream

Equivalently calling cudaStreamSynchronize to block until this stream has completed all operations.

void tf::cudaStream::begin_capture(cudaStreamCaptureMode m = cudaStreamCaptureModeGlobal) const

begins graph capturing on the stream

When a stream is in capture mode, all operations pushed into the stream will not be executed, but will instead be captured into a graph, which will be returned via cudaStream::end_capture.

A thread's mode can be one of the following:

• cudaStreamCaptureModeGlobal: This is the default mode. If the local thread has an ongoing capture sequence that was not initiated with cudaStreamCaptureModeRelaxed at cuStreamBeginCapture, or if any other thread has a concurrent capture sequence initiated with cudaStreamCaptureModeGlobal, this thread is prohibited from potentially unsafe API calls.
• cudaStreamCaptureModeThreadLocal: If the local thread has an ongoing capture sequence not initiated with cudaStreamCaptureModeRelaxed, it is prohibited from potentially unsafe API calls. Concurrent capture sequences in other threads are ignored.
• cudaStreamCaptureModeRelaxed: The local thread is not prohibited from potentially unsafe API calls. Note that the thread is still prohibited from API calls which necessarily conflict with stream capture, for example, attempting cudaEventQuery on an event that was last recorded inside a capture sequence.

cudaGraph_t tf::cudaStream::end_capture() const

ends graph capturing on the stream

Equivalently calling cudaStreamEndCapture to end capture on stream and returning the captured graph. Capture must have been initiated on stream via a call to cudaStream::begin_capture. If capture was invalidated, due to a violation of the rules of stream capture, then a NULL graph will be returned.

void tf::cudaStream::record(cudaEvent_t event) const

records an event on the stream

Equivalently calling cudaEventRecord to record an event on this stream, both of which must be on the same CUDA context.

void tf::cudaStream::wait(cudaEvent_t event) const

waits on an event

Equivalently calling cudaStreamWaitEvent to make all future work submitted to stream wait for all work captured in event.