The C++Now call for submissions is here:

http://cppnow.org/2017-conference/announcements/2017/01/06/call-for-submission.html

The deadline is soon (February third).

I look forward to meeting you in Aspen and discussing how we can get this into Boost.

If you have any questions or ideas about presenting at C++Now (or CppCon) please let me know and we can discuss by email (jonkalb at boost dot org) or by Skype.

Jon Kalb

C++Now Conference Chair

CppCon Conference Chair

Boost Steering Committee Chair

Most interesting…I have an interest in sorting algorithms and am reaching out to start a dialog if you’re interested via e-mail (will.f.gilreath@gmail.com) I’ve created the hash sort and binar sort (among other algorithms) which I’ve written a research monograph about.

Check out:

https://arxiv.org/abs/cs/0408040 hash sort

https://arxiv.org/abs/0811.3448 binar sort

Cheers, Will

]]>What I meant originally was actually quite dreary in comparison, which was seeing that ska_sort has a large improvement over std::sort in many cases, do you think part of reason this improvement (which I called variation) was so large in some cases (as you say yourself, as large as 30%), is because std::sort is *already* bad/inefficient in comparison to the methods used in large data keeping systems and the like?

]]>And yes, I didn’t invent the idea of using radix sort on other data. I do provide sources for where I have the generalizations from. However as far as I can tell nobody has ever

a) generalized radix sort to work on generic collections and tuples, (with those two generalizations you can sort pretty much anything)

b) done the work of providing a single sorting algorithm that dispatches to the right behavior automatically.

Even the boost radix sort algorithm can only sort ints, floats and strings. And boost is usually known for writing code that’s way too general.

]]>The idea of using the radix sort to sort things more than unsigned ints *is* quite old. 10-15 years ago, I was sorting megabytes of 32-bit floats this way.

One of the details you seem to be missing is the question of whether this is a “stable” sort or not. Admittedly, I didn’t read the whole thing with consistent assiduity, but I *did* do a word search and didn’t see the word stable referring to the nature of the sort. Can you comment on whether this is a stable sort or not?

]]>Anyway, if you want to reproduce the algorithm I used to generate the benchmark above, here are the steps:

* Take the latest version of random-access vergesort: https://github.com/Morwenn/cpp-sort/blob/master/include/cpp-sort/detail/vergesort.h#L168-L322

* Drop the comparison and projection logic

* Replace the calls to pdqsort by calls to ska_sort

Note that I try the new optimizations in my cpp-sort project first. It’s a handy framework to try and compare sorting algorithms, and it’s got a somewhat decent testsuite (which is also run through Valgrind and ubsan when I push changes to GitHub). To be honest, I was already wondering how I could integrate ska_sort into the project since it is interesting and the licenses are compatible :p

]]>