------- Comment #13 from tege at swox dot com 2006-08-04 11:57 ------- Subject: Re: suboptimal 'division by constant' optimization
I looked briefly into the suggested change, without trying to assess its correctness. But since these sort of transformations are of a very mathematical nature, I trust the author of the suggested changes has proven the algorithms? I am aware of that Peter Montgomery's and my algorithms do not generate optimal sequences for v/C where C is close to the largest possible value of v. We deemed it to have very little practical use to optimize these cases. If it is decided that these cases are indeed important enough to require optimization, the choose_multiplier function should course be amended, of course. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28417
