Re: Timing Comparison?
"David McClain (as dbm at refined-audiometrics dot com)" <[email protected]>
| Newsgroups | gmane.lisp.lispworks.general |
|---|---|
| Message-ID | <[email protected]> |
So, just as an aside, what started this whole thing? I had a need for an embedded Forth implementation of a trade secret audio processing algorithm. Now patented, the details are in the public domain. But honestly, writing code in Forth is an exercise in stepping backwards by 50 years in computing. It takes me, even with as much as, now 50 years of Forth experience, on and off, it takes me 10x longer to produce a good encoding of algorithms, compared to Lisp. So I have an interest in a Scheme compiler emitting Forth as the intermediate language. I need that Forth code for obfuscation purposes, like always, but that obfuscation must be executable. And I really would rather code the algorithms in Scheme than in Forth. When I noticed that SBCL was running my large Scheme tests 30x faster than LWM, it caused me real concern. I didn’t believe for a moment that LWM is 30x slower than SBCL. And then I noticed, even after tracking down the reason for the speed disparity, that now very large numbers were still slower in LWM than in SBCL, and the slowdown grows with integer size above 2^6900. That was an interesting finding. Nothing to be said about it other than that it is comparing two different implementations of Lisp Bignums - neither of which were likely coded and compiled in Lisp. _______________________________________________ Lisp Hug - the mailing list for LispWorks users [email protected] http://www.lispworks.com/support/lisp-hug.html