Refactoring thread manager, mainly extracting thread pool

[hkaiser]

This patch separates the thread-pool functionality from the thread-manager. This is the first step towards creating an executor which has full control over processing units. The existing executors are built on top of the underlying main scheduler, which causes them to compete for processing time. The new executor will not have that disadvantage.

[pagrubel]

See error: https://www.dropbox.com/s/54jhspdlos7g597/makeerr?dl=0

[hkaiser]

See error: https://www.dropbox.com/s/54jhspdlos7g597/makeerr?dl=0

Thanks, that's fixed now.

[pagrubel]

Running some performance tests

[pagrubel]

Performance unchanged on host computers but the mic is much slower, let me see if this is the branch or somewhere else I have not checked this in a while

[hkaiser]

Performance unchanged on host computers but the mic is much slower, let me see if this is the branch or somewhere else I have not checked this in a while

Thanks. Since I have not changed anything besides splitting the two classes I would have expected to see no performance difference.

[pagrubel]

Looks good to me. Just wondering why the change to integer from unsigned int

[hkaiser]

Just wondering why the change to integer from unsigned int

This is to cleanup the return value for performance counters. It always was int64_t. However, for some reason some of the performance counter functions did return uint64_t. This is more of a fly-by change, unrelated to the refactoring itself.

[pagrubel]

I figured it wasn’t part of the refactoring and was just curious. Thanks