On 1/28/20 5:11 PM, sarn wrote:
Admit it, "my version is very likely IO bound" is hand-wavey. The top comment on HN right now is pointing out that it doesn't make sense.
I don't think it's i/o bound. Doing the split/walk-length and then walk-length again is likely making it 2x slower.
Phobos' byLine is not the fastest you can get, but it's not a slouch either. It also tries to use as many libc tricks as it can.
I would say the key point is that it's fast enough for most purposes, no matter what the actual performance problem is.
-Steve