3 comments

[ 3.6 ms ] story [ 23.9 ms ] thread
> half the speed as that of a memory copy

without going into much deep analysis, id try writing the newlines in a different loop: you are hitting the cache(and memory) in a different way when memcopying and when writing directly.

Curious question: would prefilling destination buffer with newlines (\n only) be faster than sprinkling them from a list of positions?
I'm grateful that both x86 & ARM CPUs are perf-tested.