Re: [boost] [sorting] Any feedback for the multikey quicksort?

From: Sam Schetterer (samthecppman_at_[hidden])
Date: 2007-03-26 21:57:24

On 3/25/07, Lewis Hyatt <lhyatt_at_[hidden]> wrote:
>
> > Ok. I have just re-compiled with gcc and have fixed all of the reported
> > errors. Also, I found a few bugs and have fixed those too. I am posting
> this
> > working version on vault in the sorting directory with 3 test programs
> > showing radixsort, radix quicksort, and multikey quicksort. Suprisingly,
> > they are all faster than std::sort, mostly by a factor of three and
> more,
> > for floats, radix sort is faster by a factor of five.
>
> I was able to compile your radixsort test file on gcc+cygwin, and got
> that it was faster than std::sort by a factor of (5, 1.1) for (float,
> int) in a few trials. I could not compile any of your other test files,
> though, because of various errors. I also tried to compile your code on
> Linux, but was unable to because your Timer classes are windows-specific.
>
> You should use boost::progress_timer to do the timings, since it is
> already designed to be portable. You also should not have "stdafx.h",
> TCHAR, tmain, or any other microsoft mumbo-jumbo anywhere in any of your
> code.
>
> I might suggest, since you are on Windows, that you just download cygwin
> and make sure you can compile all your code on gcc in cygwin before you
> post it. You said you compiled these examples with gcc, but I don't see
> how, unless your gcc had access to microsoft-specific header files maybe?
>
> In any case, I would be very interested to try out the
> multi-key-quicksort and other routines if you can post a working example
> and an explanation of how to use it.
>
> Also, just to make sure, you are putting the files in the vault in
> /Sorting, not /Algorithms/Sorting, right?
>
> -Lewis
>
> _______________________________________________
> Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
>
>

Yes. The files are all in /Sorting. I have redone the program on gcc so that
there are no operating system dependencies, and also I have fixed all of the
errors, and it is uploaded to vault. If you are running a low-memory
machine, then you should consider modifying the maxn variable so that it is
lower.