IWTB (in Lisp performance) - a followup
The previous article had some serious traffic yesterday and today.
Due to abuse of the moderation system on some sites I’ve been forced to present my response here separately.
To the SBCL contributor : Relax, SBCL is still a great Lisp compiler. It always has been and I’m sure it always will be. Your work on the compiler is appreciated. It’s unfortunate that you seem to fit the pontificating arm-chair theoretician stereo-type. You do have a few valid points to share in spite of your mostly incorrect presumptions regarding the Lisp code. The single worthwhile and correct contribution that you made to the discussion was your suggestion on type declarations in the critical path of the code. Thanks for that, you were one of only two people that has actually addressed the performance of the code so far. I posed a challenge to you because of your claims of being able to easily bring the code within under 3x the slowness of the C version. Your dismissive refusal to substantiate your claims with any real evidence speaks volumes to a wide ranging audience now.
The challenge to you stands indefinitely …
anonymous coward : Thank you for the practical optimization suggestions on the code. You know what you’re talking about.
You present facts objectively and we all gain from your experience in this area.
In spite of all the fury and rage (signifying nothing ?), I have recieved exactly ZERO submissions to beat the
performance of the original Lisp code. That’s ZERO submissions up until now ! I’d rather not have the final word on this.
If someone does decide to submit a contender then feel free to use whatever data structures you like but if
you algorithmically optimize your Lisp code against the C version then that’s just missing the point completely now isn’t it ?
I’d love to be shown up with a pure Lisp version that matches the performance of C (or that at least beats my Lisp version) …