On 10/7/2011 5:08 PM, Andrei Alexandrescu wrote:
On 10/07/11 13:50, Xinok wrote:
On 10/7/2011 2:20 PM, Andrei Alexandrescu wrote:
On 10/7/11 12:23 PM, Xinok wrote:
http://www.neowin.net/forum/blog/422/entry-3737-sort-algorithm-complete/
This is interesting. What do the numbers in the benchmark represent?
Andrei
I'll just post the code I used for benchmarking. Simply put, smaller
numbers are faster.
[snip]
Thanks. It would be great if you wanted to contribute your stable sort
to Phobos via a pull request.
Also, which version of D are you using? I'm seeing that
std.algorithm.sort (introSort) performs quite badly; for example, it's
twice as slow on shuffled data against quickSort, and it also deals
badly with already sorted data. Generally it does much worse than the
quickSort baseline. Wonder why.
Andrei
I'm not familiar with the process. What all would I need to do to
contribute my sort to Phobos?
D's standard library is on github:
https://github.com/D-Programming-Language/phobos. Anyone can fork it,
modify it as they please, and then submit the changes for review via a
so-called "pull request". Here's an example of a pull request with
comments and all:
https://github.com/D-Programming-Language/phobos/pull/272. There is
documentation available on github.com about how to use the site. It's
some work but it's time well invested - the kin of git and github are
here to stay. Would anyone in the community be willing to shepherd Xinok
through the first pull request?
Andrei
Thanks, I'll look into it when I have a little more time.
If my algorithm were to be implemented in Phobos, the template arguments
for std.algorithm.sort would need to be modified. For my algorithm to be
stable, it requires defining two conditions, both a 'less' and 'greater'.
