This patch adds a new member, owner_is_backend, to
inference_engine_tensor_buffer structure, which
is used to check if a backend should release
tensor buffer or not.
If owner_is_backend is set to 1 then upper framework,
Inference framework in case of Tizen, will not release
the tensor buffer.
Change-Id: I9db254daeb84fd4bc71123acaf6295fbf47d53f8
Signed-off-by: Inki Dae <inki.dae@samsung.com>
void *buffer; /**< a buffer which contains tensor data. */
inference_tensor_data_type_e data_type; /**< a tensor type of the layer. */
size_t size; /**< actual tensor buffer size in bytes. The size should be height * width * channel count * bytes per pixel. */
+ int owner_is_backend; /** < it indicates whether tensor buffer owner is backend or not. setting to 1 means that backend has to allocate and release the buffer. */
// TODO.
} inference_engine_tensor_buffer;