In general you can have multiple graphs in the same context, and objects can be shared between graphs in the same context. For example, the output of one graph can be used as the input to a different graph in the same context. You are correct in your statement that objects can NOT be shared between different contexts.
Regarding device management, the OpenVX implementation generally handles that internally, automatically scheduling processing on the available compute resources, including CPUs and GPUs. Depending on your implementation, you may have some control over this via the vxSetNodeTarget API. Consult the documentation of your implementation to determine what targets are defined.
I have to achieve multigraph context and multi device management for my application.
I am referring CPU and GPU devices.
Can you please give me some suggestions which way is good to achieve this and what all things need to be taken care.
One could use multiple contexts with different of graphs within each of the context too — in such case, objects can’t be shared across contexts, so the application has to be careful.
Thanks for your reply.[/QUOTE]