Thanks for the reply and benchmark insights.
What OS are you running?
Running your benchmark on my laptop with macOS 12.1 with perl 5.34 I get similar percentage differences to my original benchmark code:
macOS 12.1, your benchmark code:
Rate methodic blessy direct
methodic 16445180/s -- -27% -35%
blessy 22503930/s 37% -- -11%
direct 25252404/s 54% 12% --
macOS 12.1, rerun of my original benchmark code:
Rate method blessed sub
method 17857143/s -- -35% -44%
blessed 27397260/s 53% -- -14%
sub 31746032/s 78% 16% --