Is there an explanation of why they switched away from an algorithm that is (according to their linked table) 4x faster than the new algorithm described in the post?
The article says Żmij changed from being based on Schubfach (34.58 ns) to being based on yy (16.61 ns) which with other optimizations helped Żmij get down to 3.93 ns.
rrampage | 10 hours ago
How does Zmij compare with Teju Jagua ( https://github.com/cassioneri/teju_jagua ) which is another unheard of but fast double to string algorithm ?
dtolnay | 2 hours ago
Zmij is 2–3× faster. https://github.com/dtolnay/dtoa-benchmark
jaredkrinke | a day ago
Is there an explanation of why they switched away from an algorithm that is (according to their linked table) 4x faster than the new algorithm described in the post?
[OP] fanf | a day ago
The article says Żmij changed from being based on Schubfach (34.58 ns) to being based on yy (16.61 ns) which with other optimizations helped Żmij get down to 3.93 ns.
jaredkrinke | 22 hours ago
Ha, yes, I see where I misread the intro now. Thanks!