I wanted to test this and found a surprising result, although with shuf instead of tail... ( head -n40000 filelist_h.txt | grep 'v\|$' >> /dev/null; )

I wanted to test this and found a surprising result, although with shuf instead of tail... ( head -n40000 filelist_h.txt | grep 'v\|$' >> /dev/null; ) 0.00s user 0.02s system 43% cpu 0.045 total ( tail -n40000 filelist_h.txt | grep 'v\|$' >> /dev/null; ) 0.00s user 0.01s system 22% cpu 0.066 total ( shuf -n40000 filelist_h.txt | grep 'v\|$' >> /dev/null; ) 0.71s user 1.37s system 12% cpu 16.874 total ( cat filelist_h.txt | head -n40000 | grep 'v\|$' >> /dev/null; 0.00s user 0.01s system 101% cpu 0.017 total ( cat filelist_h.txt | tail -n40000 | grep 'v\|$' >> /dev/null; 0.05s user 0.45s system 54% cpu 0.930 total ( cat filelist_h.txt | shuf -n40000 | grep 'v\|$' >> /dev/null; 0.60s user 0.91s system 96% cpu 1.565 total The results are fairly repeatable: - as expected, head and tail alone are equally quick alone - shuf is ridiculously slow due to overhead compared to cat | shuf - cat | shuf is just a bit slower than cat | tail. - cat | tail is slower than tail. - cat | head is faster than head (but only because of overhead) Caveats are that this is WSL2 and the file is 480 MB (5 million lines) in a mounted Windows directory, although that helps magnify that slow I/O can influence how you pipe commands.