Thread overview
[OT] Fast Deterministic Selection
May 18, 2017
Stanislav Blinov
May 19, 2017
Nordlöw
May 20, 2017
Jon Degenhardt
May 18, 2017
The implementation is an improved version of what we now have in the D standard library. I'll take up the task of updating phobos at a later time.

https://www.reddit.com/r/programming/comments/6bwsjn/fast_deterministic_selection_sea_2017_now_with/


Andrei
May 18, 2017
On Thursday, 18 May 2017 at 15:14:17 UTC, Andrei Alexandrescu wrote:
> The implementation is an improved version of what we now have in the D standard library. I'll take up the task of updating phobos at a later time.
>
> https://www.reddit.com/r/programming/comments/6bwsjn/fast_deterministic_selection_sea_2017_now_with/
>
>
> Andrei

Nice, congrats!

...and predictably, poison fumes are already in on reddit...
May 19, 2017
On Thursday, 18 May 2017 at 15:14:17 UTC, Andrei Alexandrescu wrote:
> https://www.reddit.com/r/programming/comments/6bwsjn/fast_deterministic_selection_sea_2017_now_with/

Great work.
May 20, 2017
On Thursday, 18 May 2017 at 15:14:17 UTC, Andrei Alexandrescu wrote:
> The implementation is an improved version of what we now have in the D standard library. I'll take up the task of updating phobos at a later time.
>
> https://www.reddit.com/r/programming/comments/6bwsjn/fast_deterministic_selection_sea_2017_now_with/
>
>
> Andrei

Very nice! Is this materially faster than what is currently in Phobos (PR 4815)? That update was a substantial performance win by itself.

--Jon
May 22, 2017
On 05/20/2017 04:52 PM, Jon Degenhardt wrote:
> On Thursday, 18 May 2017 at 15:14:17 UTC, Andrei Alexandrescu wrote:
>> The implementation is an improved version of what we now have in the D standard library. I'll take up the task of updating phobos at a later time.
>>
>> https://www.reddit.com/r/programming/comments/6bwsjn/fast_deterministic_selection_sea_2017_now_with/ 
>>
>>
>>
>> Andrei
> 
> Very nice! Is this materially faster than what is currently in Phobos (PR 4815)? That update was a substantial performance win by itself.
> 
> --Jon

The code per the paper is faster for more data sets. For the NLP data you tested on, speed should be about the same. -- Andrei