Thanks for clarifying. Are you certain the non-threaded IPP functions do not block? Do some of them not protect system resources with mutexes or semaphores?
I have a simple test application with mutliple threads spanning multiple processor cores, accessing a single non-threaded IPP function, and my test data suggests that the function call disallows parallel execution across these multiple threads/cores. Perhaps the problem is in my thread creation or data creation in relation to the Intel IPP model?
Any guidance is most appreciated.