count API calls and enqueues between blocking calls

[bashbaug]

Observed Behavior

N/A, this is a new feature request.

Desired Behavior

For some types of performance analysis it can be helpful to understand how many API calls and enqueues occur between blocking commands as an estimate of application complexity.

The API calls should be collected per host thread.

The enqueues should be collected per command-queue.

Initially, the count can simply be recorded to the log, ideally in a form that can be imported into a spreadsheet for further analysis. Eventually, the count could be processed (e.g. into a histogram?).

Steps to Reproduce

N/A

[bashbaug]

In a perfect world I would love to dump a DOT file showing the enqueued commands and dependences between them between blocking commands, but a count would be useful as a first step.