[FAQ] PROCESSOR-SDK-DRA8X-TDA4X: Multi channel processing in OpenVX

There are a few things to consider when creating a multi channel graph in OpenVX using pipelining.

In the case that each of the channels are being processed by the same downstream node, using the vxReplicateNode feature is the recommended approach in TIOVX.  This method reduces application code size and is generally a more scalable approach to using multiple channels.  A common case is where this occurs is when the tivxCaptureNode output is connected to the tivxVpacVissNode.  Each of the captured channels are processed by the VISS HWA and thus the replicate node feature can be used.

However, in the case listed above, the replicate node feature will not be sufficient given that the channels are processed by different nodes.  In this case, the vxGetObjectArray API must first be used as stated.  However, in addition to using the vxGetObjectArray API, the downstream node consuming the multi channel output must also be modified to consume the proper channel.  The API tivxGetObjDescElement can be used within the node process callback itself to extract the element used by the node.  An example of the usage of this API can be found in tiovx/kernels_j7/hwa/display/vx_display_target.c.

For more information on this topic, please refer to the section in the user guide below:

https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/latest/exports/docs/tiovx/docs/user_guide/TIOVX_USAGE_REC.html#TI_PIPELINING_COMPOSITE_OBJECTS