On Wed, 9 Oct 2024 14:28:53 GMT, fabioromano1 <d...@openjdk.org> wrote:
>> Seems like `BigInteger`s division either takes care of removing common >> powers of 2 in the dividend and divisor (Knuth division), or removing them >> has no practical benefit (Burnikel-Ziegler). >> >> Thus, I'm wondering if using powers of 10 rather than powers of 5 in the >> stripping logic could simplify the code while retaining comparable >> performance. >> >> The powers of 10 are already indirectly kept in the class by method >> `bigTenToThe()` (although they affect resident memory footprint). However, >> since that method is there to stay and used in other places as well, it >> might be worthwhile to profit from it. > > @rgiulietti > - About powers of two, there is a difference here: all powers of two are > removed, not only the common, to use numbers as small as possible, and the > powers are removed all in once, while if not they were removed at each > division and also in the divisor, which is a power of 10, slowing the time of > every iteration. > - About `bigTenToThe()`: its shortcoming is that the powers are calculated > multiplying by 10 each time, rather than using repeated squares trick, until > the cache is completely filled. > > These are the main reasons why I decided not to use `bigTenToThe()`. Oh right, `expandBigIntegerTenPowers()` indeed computes and stores a lot of values. I'm thinking of opening an issue to rewrite it to just keep the values that are actually computed as intermediate results of a repeated squaring implementation. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/21323#discussion_r1793763450
