It's good to be wrong about Lisp performance

Wed, May 6, 2009

The last article ended with a challenge hoping that someone would
come up with an improved version of the boggle solver that beats my version.

Well someone did, they also proved that with this particular algorithm
the Lisp version can match C within about 20%. Very good !

Here are the new graphs (note that the first is no longer on a logarithmic scale).

The patches have been merged into the
boggle repo.

As for the problems encountered with the C version :

I’ve not done much testing on anything besides 64-bit linux.
It’s perfectly stable there for me and a few other folks.
I don’t have a 32-bit box to figure out what the issue is.

The response is much appreciated. The author also bullet points specific issues with
the original code that everyone can benefit from.

It’s good to be wrong about Lisp performance.