jk-jeon / dtoa-benchmark Goto Github PK

Not really an issue, sorry for polluting the issue section but I couldn't find a way of contacting you and maybe you're interested in this :)

I stumbled upon your blog post and while copying/changing things I stumbled upon a (very minor but apparently consistent) performance improvement.

Basically I unrolled manually the case digits <= 8 and then used a computed goto to jump to the correct iteration. Nothing fancy, the ASM is way bigger than yours but maybe this interests you anyway.

https://quick-bench.com/q/ZeZDP1NCi4zyy_YSDR0jjKhEVDY

And thanks for the detailed explanation in the blog post :)

EDIT: only after I finished posting this I realize this is the dtoa repo and not the itoa one. Embarassing, sorry again.