Op 9 feb 2014, om 13:19 heeft big stone het volgende geschreven:
Congratulations, *E.Pasma* !
You indeed won the "brute force sudoku speed" competition for sqlite
3.8.3
by a factor of x2.5 on my PC !
D. Richard Hipp is still tuning the CTE implementation, so
competition may
be less slow with 3.8.4.
To get clother to python speed (http://norvig.com/sudoku.html), we
would
need to implement :
- an efficient Peter Norvig's method,
- or an efficient Dancing Links method.
==> See below link for "slower than brute force" dancing links
implementation in pgSQL.
http://www.postgresql.org/message-id/caj-6ajqpxxex8ftljksngksvbjrn0h3hydw9xpsvu837gsn...@mail.gmail.com
regards,
I see that there is much competition. The version at http://norvig.com/sudoku.html
solves our example in 0.04 seconds. At least that is what it
reports. A timer within Python made it 0.1 seconds. Still that is at
least 10 times faster than my SQLite version. On my machine I have the
following timings now:
http://www.sqlite.org/lang_with.html#sudoku - 13 seconds
Bigstone's https://www.mail-archive.com/sqlite-users@sqlite.org/msg81630.html
- 4 sceonds
My bitmapped version - 1.5 seconds
Peter Norvig's http://norvig.com/sudoku.html - 0.1 second
By the way I could not run your postgresql version. Psql reports that
function recommands does not exist.
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users