Anonymous user
Talk:Sieve of Eratosthenes: Difference between revisions
→Trial division sieves should be moved from here
Line 227:
:: I don't consider myself active enough to remove the draft thing too... but of course I think that it's the right thing. --[[User:Elibarzilay|Elibarzilay]] ([[User talk:Elibarzilay|talk]]) 19:32, 13 September 2014 (UTC)
::: WillNess: Sorry, but continuing down here... What I did now is: (1) Removed the sequence versions from the PbTD page since they're on the new page; (2) I first edited your version on the SoE page, but then realized that my attempt at doing the same to the Bird code is shorter and faster, so I dropped both in *instead* of your code -- I don't like removing other people's code, but I decided on doing that because it's better on speed, clarity, and being more idiomatic (feel free to tweak it however you like, specifically, the timings might be off); (3) On the new page, I've added a version that is optimized with the postponement trick.
::: FWIW, I'm getting results that still don't go in-line with the conclusion that a proper SoE should be faster: the "unfaithful" plain sieve with the postponement trick <1> is about twice faster than my Bird implementation with the same trick <3>, and that is a tiny bit faster than my Bird implementation without the trick <2>, while your version of the proper SoE with the trick is about three times slower than <1>. Based on the paper and what you said, I'd expect <1> and and <2> to be roughly the same, and <3>/<4> to be much faster. --[[User:Elibarzilay|Elibarzilay]] ([[User talk:Elibarzilay|talk]]) 20:16, 13 September 2014 (UTC)
| |||