Hooray, my IPv6 parser is mostly working, and only missing the final cursed case.

It's done. By god, it's done.

I mean, parseIPv6 is also awful and embarassing and much worse than the stdlib implementation right now. But is it at least faster than the stdlib?
Ope never mind, other tests are failing still. Bah.
lol, fuck. Not only is it 50% slower than the stdlib's parser, it's also not alloc-free.
oh, it's because it's benchmarking parsing a zone, and we have to intern the zone... I think.
Yess. For IPs without a zone, my parser is faster and alloc-free. Not hugely faster mind you.
I guess if you squint, it's about as good as the IPv4 parser. It has to touch 4x more bytes, so just under 4x slower isn't that bad.
That said, I'm pretty sure my code is a branch predictor's nightmare right now, not to mention it's entirely unreadable even by my low standards. Needs another round of improving, now that it passes tests.
I think it's time to checkpoint and play Elite Dangerous tonight, but the plan for tomorrow is to see if I can't turn this into a more explicit state machine, so that the code is considering fewer possibilities at any one time.
In general, I think all IPv6s parse in 4 stages:
- Hex before ellipsis
- Hex after ellipsis
- Embedded IPv4
- Zone suffix

All but the first are optional, and the conditionals to decide which new state to transition to is a bit ugly...
... But I think within each state the "happy loop" condition is simple and should keep the CPU and branch predictors pretty happy.
Current WIP, as I pack in for tonight to go play Elite: https://t.co/acwEFavxQi

WARNING: side-effects may include eye bleeding and stomach ulcers.
It also still has a bunch of debugging code in it, the comments may not match the code any more, and I haven't yet iterated on the benchmarks and SSA to see what hand-inlining was actually productive vs. harmful. So it's all hand-inlined, in the worst possible way.
The code's definitely not getting submitted like this, but that's my hacking process I guess :)

More from Internet

You May Also Like

**Thread on Bravery of Sikhs**
(I am forced to do this due to continuous hounding of Sikh Extremists since yesterday)

Rani Jindan Kaur, wife of Maharaja Ranjit Singh had illegitimate relations with Lal Singh (PM of Ranjit Singh). Along with Lal Singh, she attacked Jammu, burnt - https://t.co/EfjAq59AyI


Hindu villages of Jasrota, caused rebellion in Jammu, attacked Kishtwar.

Ancestors of Raja Ranjit Singh, The Sansi Tribe used to give daughters as concubines to Jahangir.


The Ludhiana Political Agency (Later NW Fronties Prov) was formed by less than 4000 British soldiers who advanced from Delhi and reached Ludhiana, receiving submissions of all sikh chiefs along the way. The submission of the troops of Raja of Lahore (Ranjit Singh) at Ambala.

Dabistan a contemporary book on Sikh History tells us that Guru Hargobind broke Naina devi Idol Same source describes Guru Hargobind serving a eunuch
YarKhan. (ref was proudly shared by a sikh on twitter)
Gobind Singh followed Bahadur Shah to Deccan to fight for him.


In Zafarnama, Guru Gobind Singh states that the reason he was in conflict with the Hill Rajas was that while they were worshiping idols, while he was an idol-breaker.

And idiot Hindus place him along Maharana, Prithviraj and Shivaji as saviours of Dharma.
The YouTube algorithm that I helped build in 2011 still recommends the flat earth theory by the *hundreds of millions*. This investigation by @RawStory shows some of the real-life consequences of this badly designed AI.


This spring at SxSW, @SusanWojcicki promised "Wikipedia snippets" on debated videos. But they didn't put them on flat earth videos, and instead @YouTube is promoting merchandising such as "NASA lies - Never Trust a Snake". 2/


A few example of flat earth videos that were promoted by YouTube #today:
https://t.co/TumQiX2tlj 3/

https://t.co/uAORIJ5BYX 4/

https://t.co/yOGZ0pLfHG 5/