Speed

[Geoff Canyon]

On Sun, Aug 24, 2014 at 11:09 AM, Beat Cornaz <B.Cornaz at gmx.net> wrote:

> I am a bit at a loss, as I would be surprised if my way would be like 25
> times faster than the fastest known algorithm. Did I make a mistake in
> implementing Dicks code (although
> Dick also reports 2 minutes to do the job, as my way does it in 5.6 secs).
> What is happening here.
> I even suspect, that I can improve more on my improved way (by only
> calculating half of the perms and swapping the 1's and 2's. I might apply
> that principle more (like only calculating 1/4 of the perms and do some
> swapping, or even better). I will look into that later.
>

The fastest algorithm doesn't take into account the difference between
LiveCode's native C speed based on a single command vs. the much slower
execution of a substantial amount of transcript/livecode/<whatever we call
the code in livecode these days>. For example, a sorting algorithm written
in LC would be *much* slower than the sort command, even if it used a more
efficient algorithm.

My permutations code (and perhaps yours) makes extensive use of the replace
command specifically because it's a single command that for our purposes
here gets a lot done. That's key in speeding things up.

gc