An execution stream.
More...
#include <dnnl.hpp>
An execution stream.
- Examples:
- cnn_inference_f32.cpp, cnn_inference_int8.cpp, cnn_training_f32.cpp, cpu_cnn_training_bf16.cpp, cpu_rnn_inference_f32.cpp, cpu_rnn_inference_int8.cpp, cross_engine_reorder.cpp, getting_started.cpp, gpu_opencl_interop.cpp, memory_format_propagation.cpp, performance_profiling.cpp, and rnn_training_f32.cpp.
◆ flags
Stream flags.
Enumerator |
---|
default_order | Default order execution.
Either in-order or out-of-order depending on the engine runtime
|
in_order | In-order execution.
|
out_of_order | Out-of-order execution.
|
default_flags | Default stream configuration.
|
◆ stream()
dnnl::stream::stream |
( |
const engine & |
eng, |
|
|
cl_command_queue |
queue |
|
) |
| |
|
inline |
Constructs a stream associated with the engine eng
and with the OpenCL command queue queue
.
◆ handle() [1/2]
Empty constructor.
Allows declaring an object before actual initialization (mostly for convenience).
- Warning
- Uninitialized object cannot be used in any library calls. Any attempt to use its methods or passing it to the other library function will lead to a thrown exception.
◆ handle() [2/2]
Constructs a C handle wrapper from a C handle.
- Parameters
-
t | The C handle to wrap. |
weak | A flag to specify whether to construct a weak wrapper. |
The documentation for this struct was generated from the following file: