Event¶

class paddle.device.cuda. Event [source]

The handle of the CUDA event.

Parameters

• enable_timing (bool, optional) – Whether the event will measure time. Default: False.

• blocking (bool, optional) – Whether the wait() func will be blocking. Default: False;

• interprocess (bool, optional) – Whether the event can be shared between processes. Defalut: False.

Examples

# required: gpu
import paddle
event = paddle.device.cuda.Event()

query ( self: paddle.fluid.core_avx.CUDAEvent ) → bool

query¶

Queries the event’s status.

Returns: A boolean which indicates all work currently captured by the event has been completed.

Examples

# required: gpu
import paddle
event = paddle.device.cuda.Event()
is_done = event.query()

record ( self: paddle.fluid.core_avx.CUDAEvent, stream: paddle.fluid.core_avx.CUDAStream = None ) → None

record¶

Records the event in the given stream.

Parameters

stream (CUDAStream, optional) – The handle of CUDA stream. If None, the stream is the current stream. Default: None.

Examples

# required: gpu
import paddle
event = paddle.device.cuda.Event()
event.record()

synchronize ( self: paddle.fluid.core_avx.CUDAEvent ) → None [source]

synchronize¶

Waits for an event to complete.

Examples

# required: gpu
import paddle
event = paddle.device.cuda.Event()
event.synchronize()