Re: Pipelined processing with the, Thread-Per-Block scheduler?

A possible explanation for 1) is the following:

Your flowgraph is so simple that essentially
in the 1st case the source/sink blocks are blocking most of the time
(after having produced their data), while the resampler is using 100% of
the one of the cores?

This however does not explain the behavior in 3)

does that make sense?
Achilleas