Subject: Re: A new Sudoku xslt implementation (Was: Re: Sudoku - A solution in XSLT 2)
From: "andrew welch" <andrew.j.welch@xxxxxxxxx>
Date: Sat, 11 Mar 2006 14:11:47 +0000
|
> I tested your new stylesheet on the following "fiendish" board, and it
> performs almost 5 times better than the previous one:
>
> <board>
> <row>0,0,0,0,0,5,0,0,0</row>
> <row>0,0,0,0,2,0,9,0,0</row>
> <row>0,8,4,9,0,0,7,0,0</row>
> <row>2,0,0,0,9,0,4,0,0</row>
> <row>0,3,0,6,0,2,0,8,0</row>
> <row>0,0,7,0,3,0,0,0,6</row>
> <row>0,0,2,0,0,9,8,1,0</row>
> <row>0,0,6,0,4,0,0,0,0</row>
> <row>0,0,0,5,0,0,0,0,0</row>
> </board>
>
> The results:
>
> AW1 AW2
> =============================
>
> 113016 14.8MB 24407 35MB
>
>
> My results on this board are:
>
> 6688 10MB
Hi Demitre,
I ran the fiendish board with both stylesheets and have different
results to you!
I have:
AW1 AW2 DN
52.5 10.7 15.75
50.5 10.3 15.81
49.5 10.5 15.9
The tests were run using SaxonB 8.7 from the command line with the -3
option to run the transform 3 times.
One difference to note is that my stylesheet is a "standalone"
stylesheet which needs the -it switch, whereas yours is the more
traditional one which uses an input XML. I doubt this is that
significant but Saxon does produce different output for the two types
when using the -t switch (no memory info for standalone).
Either way, when using the "fiendish" board you've provided, I place
mine at just over 50% faster than yours at the moment :)
|