Consider, please, the ability to specify the affinity and/or number of worker threads during (parent) session initialization.
It will enable more efficient computational resources distribution in the following cases:
- imsdk-applications that have their own worker threads with high computational load;
- multiple processes using imsdk running in parallel;
- promote the use of only one particular NUMA-node (jointly with a custom BufferAllocator/FrameAllocator) by processors and memory.