From: <lut...@fr...> - 2011-08-02 22:33:17
|
Hi folks, I have done some work on implementing an algorithm found in "Division by Invariant Integers using Multiplication", 1994 by Torbjörn Granlund and Peter L. Montgomery, to optimise SBCL's TRUNCATE and friends. It works nicely: cl-bench's SEARCH-SEQUENCE -- which, as the name suggests, tests mostly the speed of (MOD <(INTEGER 0 999999)> 1000) -- runs three times as fast on my machine with these changes, and an integer implementation of a CMRG (L'Ecuyer's "Combined Multiple Recursive Random Number Generator" that consists of several multi- plications and modulo reductions) that runs at one seventh the speed of a C implementation in current SBCL catches up to run at half the speed of C. I would like to send a patch for review, hoping I have piqued interest, and just wanted to ask: Should I send it to the mailing list or open a launchpad item? Thanks so far, Lutz |