because
- GC running at every step
- one loop takes much longer time than my code.
%10 of 10s is still 1s of GC time - which is still A LOT.
because
right, but the thing that was really surprising about your original example was that GC was running every 4 times even though you were barely allocating. In this example, you are allocating 34gb per run and GC is running multiple times per loop.
Ah, I see. Let me see how to reproduce the same issue with the examples I gave. But thanks for the help so far - really appreciated.